Cape Sharp Tidal Venture Environmental Effects Monitoring Program – Annual Report

January 2018

Page | 1

Contents Executive Summary .......................................................................................................................................... 1

1.0 Introduction ............................................................................................................................................ 3

2.0 Operational Update .............................................................................................................................. 4

3.0 Environmental Effects Monitoring Program ......................................................................................... 5

3.1 Context .............................................................................................................................................. 5

3.1.1 Adaptive Management ............................................................................................................. 6

3.2 Scope ................................................................................................................................................. 6

3.3 Monitoring Devices ........................................................................................................................... 7

3.3.1 icListen Smart Hydrophone ...................................................................................................... 8

3.3.2 Gemini Sonar............................................................................................................................. 8

3.2.3 Autonomous Multichannel Acoustic Recorder ......................................................................... 9

3.2.4 Video Camera .......................................................................................................................... 10

3.2.5 Acoustic Doppler Current Profilers ......................................................................................... 10

3.4 EEMP Objectives ............................................................................................................................. 10

4.0 EEMP Update ...................................................................................................................................... 13

4.1 Monitoring Devices Update ............................................................................................................ 13

4.1.1 icListen Hydrophones ............................................................................................................. 13

4.1.2 Gemini Sonar........................................................................................................................... 14

4.1.3 AMARs ..................................................................................................................................... 15

4.1.4 Video Camera .......................................................................................................................... 15

4.2 2016/2017 Monitoring Results ....................................................................................................... 15

4.2.1 Fish and Marine Mammals ..................................................................................................... 15

4.2.2 Operational Sound .................................................................................................................. 21

5.0 Contingency Monitoring Program ...................................................................................................... 23

5.0 Additional Items .................................................................................................................................. 25

5.1 Updated EEMP ................................................................................................................................ 25

5.1 Data Management .......................................................................................................................... 25

Page | 1

Executive Summary From November 8, 2016 to April 21, 2017 the Cape Sharp Tidal (CST) demonstration in-stream tidal turbine (the Project) produced electricity onto the Nova Scotia Power Inc. grid and gathered important monitoring data at the Fundy Ocean Research Center for Energy (FORCE) tidal energy test site. On June 15, 2017 the turbine was successfully and safely retrieved and taken to Saint John, New Brunswick where it continues to be subject to a detailed inspection and evaluation. The results of the evaluation have been used to improve the next turbine to be deployed. The CST Environmental Effects Monitoring Program (EEMP), initiated in November 2016 upon deployment of the in-stream tidal turbine, aims to test environmental monitoring devices and data management processes, and to monitor and better understand potential environmental effects and interactions of specific environmental components (fish, marine mammals, operational noise) in the near-field area (i.e., 0-100 metres) of the OpenHydro Open-Center Turbine (in-stream tidal turbine). The overall objective of the monitoring program is to verify the accuracy of environmental effect predictions made in the 2009 Environmental Assessment (EA) report completed for the FORCE test site. The monitoring program requires quarterly reports to be submitted to regulators to provide interim updates on the Project, including operational information on the turbine and the monitoring devices. This Annual Report has been submitted to provide a status update for the fourth quarter (October-December) of 2017, research results from the 2016/2017 deployment, and a year-end summary for 2017. Insights for this report include:

• The focus of operations during this reporting period (October – December 2017) continued from Q3 and has involved the ongoing evaluation and inspection of the retrieved turbine and all associated monitoring devices. The learnings from the first deployment (e.g., operations, turbine functioning, monitoring devices etc.) are being integrated into the second turbine. CST will continue to take full advantage of having the recovered turbine in port to carefully evaluate and understand how the first turbine performed while deployed, and to determine how to continually improve the operation and increase the efficiency of the technology.

• Some mitigative measures identified and discussed in Q3 for the monitoring devices have been implemented during this reporting period including new cabling and the design of a protective measure for the top mounted hydrophone and new bracketing for the video camera.

• To deliver an improved EEMP during the next deployment, a commissioning plan for all monitoring devices has been developed by CST for implementation in 2018. The purpose of the commissioning plan is to manage and document the improvements to all devices, ensure the completion of a suite of tests prior to deployment, and confirm proper functioning prior to deployment.

• Final results analyses for monitoring data collected during the 2016/2017 deployment including: Gemini imaging sonar data (fish); icListen hydrophone data (marine mammals – harbour porpoise); and acoustic recorder (turbine operational sound) were completed in Q4 by independent researchers. Reports for each environmental component were submitted to CST

Page | 2

detailing the results and the analyses and providing recommendations. • The following achievements were made for the 2016/2017 monitoring program:

o early indication of the seasonal frequency of occurrence of harbour porpoise and patterns associated with time and current velocity;

o early indications of fish patterns and abundance in both flood and ebb tides; and o phase 1 analysis of turbine sound; the levels of ambient sound in Minas Passage

and a successful methodology of monitoring turbine sound using a moored device. • Updates on identification of process improvements for data management, data transfer/sharing

and storage. The development of new protocols is ongoing with the objective of having a faster and more efficient method for monitoring data collection during the next deployment. This will facilitate use of data and prompt analyses.

• CST, FORCE and regulators are working on the development of a revised contingency monitoring program which will be finalized prior to the next deployment.

• Continued engagement with stakeholders remains a priority for CST and we continue to explore ways to improve this process. Updates to the CST website, including the addition of updated frequently-asked questions (FAQs), was also done this last quarter to address/answer a number of specific questions. These updates will be ongoing as the Project continues.

• A second deployment is planned for 2018; however a date has not yet been confirmed. The information gathered builds upon and complements the 93 scientific reports and related documents that have been completed to date on tidal power topics for the Bay of Fundy as well as the international body of research on in-stream tidal energy. We continually focus on research and development and are applying what we learned from our first deployment as we refine aspects of the turbine technology and the monitoring technologies. Whether our turbine is in the water or not, our team continues testing and modifying our technology.

Page | 3

1.0 Introduction Cape Sharp Tidal Venture (CST), a joint venture between Emera Inc. and OpenHydro, a Naval Energies company, deployed one, 2 megawatt (MW) in-stream tidal energy turbine at the Fundy Ocean Research Center for Energy (FORCE) site near Parrsboro, Nova Scotia, on November 7, 2016. The Project is the beginning of a 4MW, two-turbine demonstration project (the Project). The turbine was deployed for a six-month period and was retrieved to address minor repairs and upgrades. The FORCE site was approved under a joint federal and provincial Environmental Assessment (EA) in 2009. This EA considered:

• Multiple subsea turbine generators; • Subsea cables connecting the turbines to land-based infrastructure; • An onshore transformer substation, and; • Power lines connecting the substation to the local power transmission system..

The documents related to this process are available on the FORCE website at: http://fundyforce.ca/environment/enviromental-assesment/. The approval was in accordance with Section 13(1)(b) of the Environmental Assessment Regulations, pursuant to Part IV of the NS Environment Act and stated that any adverse effects or significant environmental effects could be adequately mitigated through compliance with specific conditions. Turbine developers who were awarded ‘berths’ within the FORCE site are required to meet the relevant conditions of the EA Approval, including the development and implementation of an Environmental Effects Monitoring Program (EEMP), for potential near-field effects to specific environmental components. The CST Project was further reviewed by the federal department of Fisheries and Oceans Canada (DFO) through an application made under the Fisheries Act prior to deployment. The Fisheries Act focusses on conservation and protection of fish habitat essential to sustaining freshwater and marine fish species and prohibits serious harm to fish [subsection 35(1)]. The Project application was also reviewed under the Species at Risk Act (SARA) to determine whether it would adversely impact listed aquatic species at risk and contravene sections 32, 33 and 58 of SARA. It was determined that this demonstration-scale tidal Project would not result in serious harm, as defined under the Fisheries Act, to fish and fish habitat, or cause negative effects to marine mammals, and that the Project would not contravene sections 32, 33 or 58 of SARA. The Fisheries Protection Program (FPP) at DFO acknowledged the technological and environmental challenges that need to be addressed through adaptive management measures to improve the understanding of interactions between aquatic resources and in-stream tidal devices. It was acknowledged that the adaptive management approach to environmental monitoring would address the information gaps raised by the department. DFO therefore issued CST a Letter of Advice and a set of recommendations to be implemented within the CST EEMP as the Project progressed. As part of the recommendations, and under the mandate of an adaptive management approach to

Page | 4

environmental monitoring, DFO recommended that CST submit interim reports on a quarterly basis to provide regulators and stakeholders with: an update on the turbine and the monitoring devices; issues or concerns; any available preliminary results; and a discussion of how the objectives of the EEMP are being demonstrated. The scope of the annual monitoring report (Q4 Report) includes an update for the last quarter, final data results for the year, up to the end of the third quarter1, and plans for following year. In addition to the annual and quarterly reports, CST meets with representatives of the FPP and NS Environment (NSE) on a regular basis to provide updates and to share information regarding ongoing data acquisition, analyses and management, as well as discussions on any data or monitoring issues and how they are being addressed and mitigated. This adaptive approach was developed to allow for ongoing review of the EEMP by the FPP and NSE, and for adjustments and constant improvements to the EEMP ensure that monitoring and management strategies are modified as appropriate. As noted above, this Annual Report is the final report to NSE and the FPP (DFO) for 2017 and provides:

• A follow-up to the Q1, Q2, Q3 Reports (attached in Appendix A); • An update on activities related to the turbine and the monitoring devices for the reporting

period (October - December); • Final data results for the 2016/2017 deployment; and • An update on the contingency program and updates to the EEMP for 2018.

Additional information about the EEMP, as well as quarterly reports for the 2016/2017 year, are available here (http://capesharptidal.com/eemp/), on the Cape Sharp Tidal website.

2.0 Operational Update CST deployed one, 2 megawatt (MW) in-stream tidal energy turbine at the FORCE site on November 7, 2016. The turbine was deployed for a six-month period and was disconnected from the FORCE subsea cable in April 2017 and retrieved June 2017. Following retrieval, the turbine and subsea base were towed to port facilities in Saint John, New Brunswick. Details of the marine operations around the retrieval were provided in the Q2 and Q3 Reports (Appendix A). The focus of operations during this reporting period (October – December 2017) has been the evaluation and inspection of the retrieved turbine and all associated monitoring devices. The learnings from the first deployment (e.g., operations, turbine functioning, monitoring devices etc.) are being integrated into the second turbine in preparation for deployment at the FORCE site, at Berth D, in 2018 (the date for this deployment has not yet been confirmed). CST will continue to take full advantage of having the recovered turbine in port to carefully evaluate and understand how the first turbine performed while deployed, and to determine how to continually improve the operation and increase the efficiency of the technology.

1 Monitoring activities can only be conducted when there is a turbine deployed and connected to the FORCE subsea cable. This 2017 Annual Report therefore only includes data results for the deployment that extended from November 2016 to April 2017.

Page | 5

The retrieval of the turbine also provided an opportunity for CST and researchers involved in the monitoring studies, to inspect all monitoring devices, evaluate and improve device protection where needed, and to implement learnings that were gained during the six-month deployment. All monitoring devices were removed from the turbine for a full inspection to confirm the condition of the instruments, and to troubleshoot issues with communications and data acquisition. The results will be used to adjust and improve the use and positioning of the monitoring instruments on the next deployment.

3.0 Environmental Effects Monitoring Program

3.1 Context As required by the conditions of the FORCE EA Approval (2009), the CST EEMP (the Program) was developed in collaboration with experts in the field of in-stream tidal energy and with input from government agencies, including DFO and NSE, as well as other in-stream tidal energy interests including the Offshore Energy Research Association of Nova Scotia (OERA), FORCE, and FORCE’s independent Environmental Monitoring and Advisory Committee (EMAC). The CST EEMP forms a component of FORCE’s EEMP commitment under the FORCE Environmental Management Plan. Both EEMPs have been designed to be complementary in order to achieve the most meaningful examination of potential effects, and in consideration of the 93 baseline studies and reports conducted on Fundy tidal power topics since 2008.2 Additional information is available in the CST EEMP document available on the CST website: http://capesharptidal.com/eemp/. The FORCE EEMP is available on the FORCE website: http://fundyforce.ca/wp-content/uploads/2012/05/FORCE-EEMP-2016.pdf. The CST EEMP cornerstone is an adaptive management approach to collecting, analyzing and evaluating data and making informed, science-based decisions to adjust technology and monitoring methods, assessing mitigation measures and addressing concerns as necessary. This approach is necessary because this is a new area of research and development and there are difficulties inherent with gathering data in harsh tidal environments, such as the Minas Passage. The Program allows for adjustments and constant improvements to be made as knowledge about the system and knowledge and understanding of environmental interactions grows. The overall purpose of the CST monitoring program is to better understand potential effects and interactions of the specific environmental components (i.e., fish, marine mammals, operational sound) in the near-field (~0-100m). The program also assesses new monitoring technologies, and processes for data collection and analyses for studies in tidal environments. The CST EEMP is part of the overall FORCE monitoring program that considers a broader area. As more data is collected, results of studies and lessons learned completed by FORCE and CST will be integrated as

2 The FORCE website provides a list of reports and studies completed to date: http://fundyforce.ca/environment/research/

Page | 6

part of a collaborative process to increase knowledge and to monitor and understand the potential effects for the FORCE test area. 3.1.1 Adaptive Management An adaptive management approach is used to evaluate monitoring data and make informed, science-based decisions to modify monitoring and assess mitigation measures as necessary. This approach is necessary due to the unknowns and difficulties inherent with gathering data in tidal environments such as the Minas Passage and allows for adjustments and constant improvements to be made as knowledge about the system and environmental interactions become known. Outcomes will be reviewed continuously with DFO, FORCE’s EMAC and others and, where required, approaches and methodologies will be revised on the basis of accumulated experience and observed progress toward achieving the monitoring objectives. This approach will assist with resolving gaps in the knowledge of the potential effects of the Project and usefulness of mitigation measures. The approach will also facilitate the implementation of new or modified monitoring strategies.

3.2 Scope The overall purpose of the CST EEMP is to better understand potential environmental effects and interactions of specific environmental components in the near-field area (i.e., 0-100 metres) with the Open-Center in-stream tidal device at the FORCE test site. The overall research objective of the CST EEMP is to verify the accuracy of environmental effect predictions made in the FORCE 2009 EA which stated that, with the implementation of the proposed mitigation measures, including development and implementation of a detailed monitoring plan, adverse residual environmental effects are predicted to be not significant for all valued ecosystem components. The EEMP will also inform future monitoring plans, increase overall knowledge about monitoring methods and analysis, help to identify possible mitigative measures, and contribute to the building of local technical knowledge and expertise of the tidal industry. The environmental components that CST was instructed to focus on by regulators are: fish; marine mammals; and turbine operational sound. Research programs for each of these components were initiated for the 2016/2017 deployment and data was collected for all three environmental components. The final research reports are attached as Appendices B, C and D. A summary of each report is provided in Section 4.2. The CST EEMP is reviewed continuously with regulators and FORCE and modified on the basis of accumulated experience and observed progress toward achieving the monitoring objectives. This adaptive approach assists with resolving knowledge gaps related to potential effects of the Project and will also facilitate the design and implementation of new or modified monitoring strategies.

Page | 7

CST is working with industry-leading local and international experts in marine technology in tidal environments to collect and interpret this data. Quarterly update reports are provided to regulators (NSE and DFO) on April 1, July 1, and October 1 of each year. Each quarterly report provides an interim update on Project operations and the EEMP, including any preliminary analyses that may have taken place. The Annual Report provides a quarterly update for the last quarter (October-December), a summary of the year’s data results and includes full research papers for each study completed over the year. The CST EEMP, implemented in 2016/2017, is available on the Project website here: http://capesharptidal.com/wp-content/uploads/2016/03/CSTEEMP.pdf. As noted in the first letter of recommendations submitted to NSE by the FPP (DFO) in June 2016, the EEMP will be updated in 2018.

3.3 Monitoring Devices The EEMP combines both passive (“listening”) and active (“sonar”) monitoring devices. The sensors are co-located on the turbine to explore an integrated monitoring system that collects data specific to marine mammals, fish and turbine operational sound. The ability to integrate monitoring sensors in this type of environment is an important part of this demonstration Project. Lessons learned are key to understanding the environment and the potential interactions of in-stream tidal turbines with the environment, but are also a fundamental element to the ongoing improvement of monitoring programs within this emerging industry. Data from all the monitoring devices on the turbine is transmitted continuously through a fibre optic data cable contained within the subsea power cable. Data is logged on-shore to hard drives and remotely saved to an OpenHydro server. As noted above in the Operational Update, the retrieval of the turbine in June provided an opportunity for CST and researchers involved in the monitoring studies (refer to Section 4.2); to inspect all monitoring devices, evaluate and improve device protection where needed, and to implement learnings that were gained during the six-month deployment. All monitoring devices were removed from the turbine for a full inspection to confirm the condition of the instruments, and to troubleshoot issues with communications and data acquisition. The results will be used to adjust and improve the use and positioning of the monitoring instruments on the next deployment. The following sections provide details about each of the monitoring devices and the approach taken for the CST EEMP.

Page | 8

3.3.1 icListen Smart Hydrophone There are four hydrophones located on the turbine. The objective of these passive devices is to detect harbour porpoise vocalizations to determine the seasonal frequency of this species, and other vocalizing marine mammals (i.e., whales), and to supplement data results associated with the sonar (the active device) on how marine wildlife interacts with the turbine. A hydrophone is an underwater microphone which converts sound energy to electrical energy. The icListen Smart Hydrophone, by Ocean Sonics, is a digital hydrophone which eliminates the need to add amplifiers and other electronics to the instrument to collect underwater sound. Data is processed and can be streamed to a receiver onshore and/or recorded within the instrument. Triggers, such as harbour porpoise clicks and whistles, are detected as specific acoustic events (sound events) and analyzed using specialized software. Other sounds in the environment are also recorded which provide context and can increase understanding of the overall sound in the environment. Using four devices for the Project creates an opportunity to:

• compare various locations for devices on the turbine; • investigate the potential for integration with other monitoring devices; and • examine the potential for localization of porpoise sounds under various tidal and operational

conditions. An additional benefit with the redundancy of hydrophones is so that in the event that one or more of the hydrophones is damaged, sufficient data to meet monitoring requirements (i.e., porpoise detection) can still be collected from one or two units to meet the monitoring requirements. As recommended by DFO in a letter to NSE (April 2017), the hydrophones mounted on the turbine will provide data regarding marine mammal activity in the area and that data will be used to undertake multi-year sampling at Berth D to allow for inter-annual comparisons. 3.3.2 Gemini Sonar Passive acoustic devices (e.g., the icListen hydrophones) detect the vocalizations of harbour porpoises and other cetaceans; they cannot detect the various noises made by fish. In order to detect fish therefore, an active or visual sonar device is required. To achieve this, one of the instruments that CST utilized was the Tritech Gemini imaging sonar; an active acoustic device. The Gemini is mounted on the turbine’s subsea base, within a protective framework, and faces the ebb tide 3. The device monitors an area up to 60 m in front of the turbine and approximately 104 m in width. Fish detection abilities can range from a lower limit of approximately 10 cm in length and upwards to larger sized fish (e.g., 1 m). The purpose of this sonar is to investigate the potential for integration with other monitoring devices and to track marine wildlife approaching the turbine. Data will be

3 The ebb tide was the direction chosen for the sonar as baseline studies showed a greater likelihood of species detection during this time and less turbulence which can affect detection.

Page | 9

used to better understand marine wildlife movements and, eventually behavior to understand potential interactions. The Gemini is a high-frequency multi-beam sonar technology that uses reflected sound (similar to an echo) to build up a picture of the underwater environment. Images created by these high-frequency sonars are low resolution when compared with contemporary video technologies such as video cameras; however when there is insufficient light or turbid conditions [cloudiness or haziness of water caused by suspended solids (sand)] video cameras lose the ability to create a clear image. Subsea environments have limited light and the Minas Passage is very turbid, but these factors are not as much of a problem for high-frequency sonars. A multi-beam, high-frequency sonar typically sends out a ‘ping’ (an acoustic pulse transmission) up to 30 times per second. Each ping is used to create a visual representation of the different intensities of the reflected sound as it bounces back. Light and turbidity therefore do not affect the sonar’s ability to detect objects. The Gemini sonar performs detection based on reflection of sound from objects in the water and then uses a ‘geometric approach’, meaning it focuses on size, shape and movements of each object. A specialized software program, SeaTec, is then used to examine the raw sonar data and can extract moving targets before performing a classification (i.e., concluding that an object is likely to be a fish or other marine wildlife or seaweed or moving rocks). These early stages of working with the Gemini data from the Minas Passage are aimed at developing an understanding of marine life around the turbine by comparing marine life detections by the SeaTec automated algorithms4 to those by a human observer, since the best way to verify automatic target detections in these early stages is human observation of the sonar data, to ensure that the software is making the correct determinations. Automatic algorithms can then be used to isolate a subset of data that contains interesting results (i.e., movements of schools of fish or larger individual species) for human observers experienced with sonar data to study in more detail. Furthermore, human analysis of the sonar data without automatic tracking provides a reference or a validation for the automatic algorithm performance. The automatic algorithms are used to identify subsets of data that may contain “interesting” targets (possible marine wildlife), which can then be used to focus the attention of human observers on times where marine life may be present. Comparing human observations to the automated detections then aids in distinguishing true detections of marine life from false ones (e.g., debris moving with the current). Although low-frequency sonars have a greater range (distance), this Project utilizes the Gemini high- frequency sonar to achieve better resolution (i.e., clearer images) at distances, allowing CST to investigate the potential for interactions as marine wildlife approach the turbine. 3.2.3 Autonomous Multichannel Acoustic Recorder Autonomous multichannel acoustic recorder (AMAR) units are bottom-mounted hydrophone units. These devices were chosen as the method to measure turbine operational sound. The particular AMAR unit,

4 An algorithm is a procedure or a formula for solving a problem. The problem is solved by conducting a sequence of specified actions.

Page | 10

designed by JASCO, is moored to the sea bottom to allow researchers to avoid the difficulties experienced with drifting hydrophones that, although typically reliable in high flow environments, will only provide a snapshot of the ambient noise and turbine sound as the instrument passes the turbine. A bottom-mounted hydrophone is also favourable in high flow environments as the instrument can reduce flow noise due to water pressure since flow speeds at the bottom of water columns are generally lower than elsewhere in the column. This flow-induced ‘noise’ can be further minimized by shielding the hydrophone. JASCO incorporates both considerations into their design, resulting in a streamlined design called the High-Flow (HF) Mooring. Two AMARs were deployed to perform a Sound Source Characterization (SSC) over a period of three months while the turbine was deployed (November - January). The SSC measures underwater sound levels from the turbine during various tidal regimes and operating modes. One AMAR was deployed approximately 100m from the turbine and a second AMAR, a control unit, was deployed approximately 680 m away from the turbine. Data can be used to compare turbine sound to the ambient or natural sound created by the environment and to understand how turbine operational sound differs during different tides (i.e., ebb vs flood). 3.2.4 Video Camera An SAIS IP-CAM HD Ethernet underwater video camera was positioned on the subsea base, facing the rotor, to explore the viability of this technology (i.e., its ability to produce an image in the tidal environment) and, if successful, to record movements of marine wildlife in close proximity of the turbine rotor. 3.2.5 Acoustic Doppler Current Profilers There are three Acoustic Doppler Current Profilers (ADCPs) located on the turbine. These devices are hydro-acoustic current meters that are used to measure water current velocities (speeds) at a number of points using the “Doppler effect” of sound waves scattered back from particles within the water. The ADCPs provide data on flow regimes within the Minas Passage. This information on current velocities is used to support analysis for all other monitoring devices and assist CST with understanding the operating efficiency of the unit. The results from these instruments are not reported as part of the EEMP, but do supplement the data from the other instrumentation to understand flows during specific time periods of interest.

3.4 EEMP Objectives The collection and processing of data relating to marine life in tidal energy sites has been limited to individual use of either active or passive acoustic sensor technologies. That past data has therefore been constrained by the limitations of each individual sensor, resulting in the need for further data collection and increased processing time. Successful real-time integration of data streams for both active and passive acoustic detections, however, would vastly improve the ability to obtain necessary high-quality environmental information, increasing understanding of the marine environment and potential

Page | 11

interactions with in-stream tidal devices. CST is utilizing this new and innovative approach to address the specific objectives of the EEMP. The approach will be implemented through the development of an interface between the two acoustic technologies (i.e., Gemini sonar and the icListen hydrophones) which would combine the strengths of each sensor type and enable the efficient collection of high quality, synergistic data for environmental monitoring at high energy sites. The EEMP objectives were developed as part of an adaptive approach designed to achieve the purpose of the EEMP and in consideration of ongoing worldwide monitoring developments in the field of tidal energy. Specific studies for this program were developed to achieve these objectives and in consideration of the environmental components (i.e., fish, marine mammals and sound). By utilizing an adaptive approach, the studies allow for:

• Improvements to existing sensor technology software that will maximize individual sensor capability;

• Integration of two complementary sensor technologies to improve ability to detect, classify, localize and track marine mammals and fish in real-time; and

• Testing of the individual sensor capability and integrated system effectiveness in a high energy site. Table 1 provides a summary of the objectives for the CST EEMP. Table 2 provides a summary of how the EEMP objectives were addressed in this first year of monitoring. Table 1. Summary of Objectives for the 2016/2017 CST EEMP

EEMP Component EEMP Objectives

Fish Determine the seasonal frequency of occurrence of fish within the near-field environment of the turbine. Integrate data-sets into a strike risk model for fish. This is a long-term objective requiring multiple years of data.

Marine Mammals

Determine the seasonal frequency of occurrence of harbour porpoise within the near-field environment of the turbine. If the numbers of other vocal cetaceans are high enough to use in a data set, results will be used to determine the seasonal frequency of these species (e.g., white-sided dolphins). Determine the relationship between harbour porpoise occurrence and turbine operations. Integrate data-sets into a strike risk model for marine mammals. This is a long term objective requiring multiple years of data.

Operational Sound

Characterize operational turbine sound to assess the effect of turbine operations on the noise profile of the site. Determine flow at specific noise levels

Table 2. Summary of Achievements of the CST EEMP – Year 1 (2016/2017 Deployment)

EEMP Objectives Achievements in Year 1 Fish Determine the seasonal frequency of occurrence of fish

The position of the Gemini sonar resulted in a view of the near-bottom water and the sea floor. The assessment of fish presence and movements in relation to the turbine therefore focused on the area of the water column 0-15m directly in front of the turbine. The data collected was biologically and technically relevant and useful for:

• assessing general trends/patterns in target abundance with temperature, day vs. night, and flood vs. ebb tides;

• assessing directional movements of fish during flood vs. ebb tides; • developing a better understanding of the performance and potential of the Gemini

sonar; and • development and validation of automated Gemini data processing techniques.

Page | 12

EEMP Objectives Achievements in Year 1 In addition, the experience during the first deployment provided a foundation for developing and refining the methodologies to be implemented for future Gemini data collection, processing, analysis, presentation and interpretation.

Integrate data-sets into a strike risk model for fish.

This is a long-term objective requiring additional years of data collection.

Marine Mammals Determine the seasonal frequency of occurrence of harbour porpoise.

Some operational issues were observed with the hydrophones, however results indicated some seasonal frequency and preference of habitat use related to current velocity:

• porpoise vocalizations (and therefore presence in the near-field area) on multiple days of every month during the deployment period from November 2016 to April 2017;

• March had the most porpoise detections – porpoises were detected on 15 days of the month;

• April also had high detections – porpoises were detected on 8 of the first 12 days of that month (data was not collected past April 13 as the turbine was being prepared for retrieval); and

• Some indication that there was a preference for current velocities around 30% of peak tidal flow in the ebb tide.

Determine the seasonal frequency of other vocal cetaceans (e.g., white-sided dolphins).

No other cetaceans were detected during the monitoring period.

Determine the relationship between harbour porpoise occurrence and turbine operations.

The hydrophones successfully detected the presence of harbor porpoises. With the operational issues addressed, localization using multiple hydrophones will be one of the focus areas for the next deployment. New cables and protective mechanisms for each of the hydrophones have been implemented to re-address this objective for the next deployment.

Integrate data-sets into a strike risk model for marine mammals.

This is a long-term objective requiring additional years of data collection.

Operational Sound Characterize operational turbine sound.

The sound study is being completed in two phases. Phase 2 will be completed in 2018. For the Phase 1 analysis, turbine sound was recorded for two months during the commissioning phase and low frequency sound was characterized. A study of the methodology was also done to understand the usefulness of a moored device to capture this type of data. For Phase 1, although recovery of the control unit was not yet achieved, the unit located 100m from the turbine was recovered. Data was analyzed and the Phase 1 results indicate:

• Turbine sound was indistinguishable from flow noise below 60 Hz; • There was a band of frequencies in the 60-300 Hz range where the sound characteristics

were likely from the turbine; • The operational sounds are different for when the turbine is free-spinning compared to

when it is generating; and • A moored device is very capable of gathering data in this type of high flow environment.

Although some understanding of turbine operational sound was achieved for low frequency, further examination of the data set is required to better understand the turbine and how it operates at high/fast currents (e.g., peak tidal flow) in order to address the objective. Phase 2 of the study will include:

• A full comparison of the data from all hydrophones (i.e., turbine-mounted units and drifters used by FORCE) to provide a more detailed analysis of the soundscape in the Crown Lease Area.

• Development of a model of sound emitted by an operating turbine as a function of frequency, current speed and direction, and turbine operating state; and

• A discussion of the impacts of the turbine sound levels on the local soundscape, including distance at which the turbine sound exceeds the background or ambient noise; and estimates of the ranges at which different groups of fish and marine mammals may hear the turbine.

CST will also work collaboratively with FORCE on the development of a model for cumulative effects of multiple turbines on mid and far field marine environment

Determine flow at specific noise levels.

Page | 13

4.0 EEMP Update The EEMP was initiated following deployment of the turbine on November 7, 2016. The turbine was disconnected from the FORCE subsea cable in April 2017 and retrieved in June 20175. Since retrieval, the turbine and subsea base have been located at the port of Saint John. During this reporting period (October – December) the monitoring program has focused on finalizing inspections and evaluations of the devices, implementing corrective/mitigative measures identified in Q3, and reviewing and analyzing the data results from the 2016/2017 deployment. As part of an overall mitigative measure towards an improved EEMP during the next deployment, a commissioning plan for all monitoring devices has been developed by CST for implementation in early 2018. The purpose of the commissioning plan is to track the improvements for all devices, ensure the completion of a suite of tests prior to deployment, and confirm proper functioning prior to deployment. Details on the performance of the various monitoring devices during the deployment, and inspections of the devices following retrieval, as well as identified issues/concerns and related mitigation, please refer to the Q3 Report (Appendix A). Updates for each of the devices are provided in the following sections.

4.1 Monitoring Devices Update 4.1.1 icListen Hydrophones In Q4, data recovery and archiving was completed for the four icListen hydrophones. This process revealed that data collection was consistent from only one hydrophone during the 2016/2017 deployment. A second hydrophone did not communicate at all and a third functioned for only a short period of time before losing connection early in the deployment. The issues appear to have been caused by problems with cables and communications. A fourth hydrophone, located on the top of the rotor, lost communication shortly after deployment due to damage it sustained. This unit has been replaced with a new device. The issues experienced with this component of the EEMP have been addressed as follows:

• Iron bar guards will be added to the hydrophone mounting structure located on the top of the turbine rotor to provide increased protection to this unit. Wider barriers around the other hydrophones will be added for increased protection of the other three devices.

• The connector was replaced for the hydrophone on the fore section of the subsea base. • The cables for all four hydrophones have been replaced to remedy any future communication

problems. Settings on all hydrophones have been adjusted to ensure synchronization. • The issue around the time series data not sampling at full bandwidth, which is needed for the

other two software programs, was corrected during the deployment allowing for all three

5 The presence of an entangled mooring line created a delay between the subsea cable disconnection and the retrieval of the turbine.

Page | 14

programs to be used on the last data set (March-April). This issue will be improved in the future by implementing set-up guidelines for the hydrophones to ensure proper sampling frequencies to ensure proper sampling frequencies.

Inspections for all four hydrophones were completed in Q4. All the units were synchronized and have undergone an initial 46 hours of consecutive operational testing in Saint John. A summary of the results of the hydrophone study are provided in Section 4.2. A final research report from Ocean Sonics is provided in Appendix B. 4.1.2 Gemini Sonar The Gemini sonar operated consistently throughout the deployment and was found to be in good shape following inspection in Saint John post-turbine retrieval. The orientation of the unit and an intermittent power supply created issues with obtaining data for the full water column and resulted in a lengthy data analysis process. These issues have been addressed and mitigation measures have been employed for the next deployment. Although the sonar will remain in the same location on the subsea base, the device has been repositioned within the protective frame to capture a more complete view of the water column at hub height and less of, or none, of the sea bottom. This will result in a greater clarity of the files and less false positives which will facilitate the human validation component of the data analysis and decrease the overall time required to obtain results. For data analysis of the 2016/2017 deployment, human observation of the data and validation of the automatic tracking results was performed to mitigate some of the issues related to the mounting angle and the communications interruptions. Researchers were able to analyze files covering an area of approximately 0m to 12m in front of the turbine where the sonar was monitoring the water column. Clarity in this region was much better and marine wildlife was noted. The issue with the power supply has been rectified by changing the cable assembly set-up to a shielded cable suitable for subsea ethernet communication. This will allow more consistent signal strength for data capturing and will give greater confidence in the automatic results. In preparation for redeployment, the Gemini sonar has undergone a series of preliminary tests. The unit was operated for 46 consecutive hours and achieved 100% link quality. A 77% link quality was maintained when the device was tested simultaneously with all four hydrophones. Link quality refers to the communication between the device and the sensor pod. While 77% is an adequate level to transfer data, and although consistent file sizes were produced in both cases, CST is working to improve this level for deployment. This will be accomplished by working with the research team and will be tested and confirmed prior to redeployment. A summary of the results of the Gemini study are provided in Section 4.2. A final research report from Acadia University is provided in Appendix C.

Page | 15

4.1.3 AMARs The AMAR unit closest to the turbine was successfully retrieved in January 2017 and results were used to analyze and characterize low (i.e., below 60Hz) and high-frequency operational sound. Results of an early analysis on sound, below 60 hertz, were provided in the Q1 Report (Appendix A). The control AMAR unit has not yet been recovered. Numerous attempts have been made to grapple the unit however the presence of boulders has hampered retrieval operations. Additional retrieval opportunities continue to be investigated. To mitigate the delay in accessing the data from the control unit, CST is working with FORCE baseline data collected prior to turbine deployment to provide the comparison to the results obtained from the AMAR positioned close to the turbine. The FORCE baseline data is appropriate as it provides information on ambient sound at the study site before any turbine was deployed. The sound study is being completed in two phases. A summary of the results of Phase 1 are provided in Section 4.2 and a final research report on Phase 1 from JASCO Applied Sciences is provided in Appendix D. Phase 2 will be completed in 2018. 4.1.4 Video Camera Communication with the video camera was unsuccessful during deployment. Upon inspection, it was confirmed that the camera had been damaged, likely early in the deployment. No video footage from this device is available. A different mounting and bracketing system has been developed in Q4 to provide better protection for this device for the next deployment. The camera was also tested successfully and is ready for installation.

4.2 2016/2017 Monitoring Results The following sections provide summaries of the final results reports submitted by third parties to CST for data collected during the 2016/2017 deployment. Full reports are provided in relevant appendices (Appendices B, C and D) as noted. 4.2.1 Fish and Marine Mammals icListen Smart Hydrophone – Passive Acoustic Study This report (Appendix B) addresses the marine mammal component of the EEMP; specifically harbour porpoise (Phocoena phocoena). Passive acoustic monitoring (hydrophones) was used to address the marine mammal component of the EEMP; specifically to gather information on the distribution and use of the near-field (i.e., < 100 m) area of the turbine by harbour porpoise and, if possible, other vocal cetaceans. The hydrophones sampled both waveform data (WAV) and processed spectral data (FFT). Both data sets

Page | 16

(WAV and FFT) were streamed from each icListen hydrophone device on the turbine to an onshore FORCE substation via a subsea cable. The data was processed using different harbour porpoise click detector software programs: PAMGuard, and Coda. The software were used to analyze data to determine when porpoises were detected in the Minas Passage. Visual inspection, screenshots of spectral data and third octave processing was also performed. Power spectral density plots were used to compare data before and after the CST turbine deployment at the FORCE site. Although it was initially thought that two hydrophones were consistently collecting data (due to two data streams recording) it became apparent during analysis that data collection was consistent from only one hydrophone during the full deployment. A second hydrophone did not communicate successfully, and a third functioned for only a short period of time before losing communication early in the deployment. These issues were associated with cabling problems. The fourth hydrophone located on the top of the rotor was damaged early on in the deployment but was still sending data resulting in the second data stream. Due to the damage done to the sensor of this unit, the data was not useable. Of the data that was collected, the results were analyzed first with Lucy (a PC Program) and then with PAMGuard and Coda software. These software programs were first used to locate porpoise clicks in the data and then the matches were used to find porpoise click trains. A porpoise click train is series of clicks, described by the time between clicks known as the inter-click interval (ICI) (i.e., 3 or more detected clicks within a small-time period such as 5 seconds). The click detectors used a minimum of 3 clicks and an ICI of 0.2 seconds to define a click train. The click train is used to minimize false positive detections by eliminating single clicks from other sources. Each click detector has different algorithms, and thus behaves differently under the full range of conditions. PamGuard settings were adjusted for the data but the high tidal flow noise and sonar signals caused many false positive and false negatives, with few true positive detections. Coda performed better in the high noise environment, so that software program was chosen as the preferred program for assessing porpoise presence. The Lucy program includes a porpoise click detector that uses intensity to indicate a porpoise click in the data. Data run through this program indicated porpoise vocalization (clicks) on multiple days per month during the deployment period from November 2016 to April 2017. March and April had the most porpoise detections (Table 3). Table 3. Days with Detected Porpoise Clicks as Detected with Lucy

Month Days of Porpoise Detections November 10, 11 December 23, 24, 25, 29 January 1, 4, 8, 29 February 2, 5, 23, 24, 26 March 1, 2, 3, 5, 10, 12, 13, 20, 21, 23, 25, 26, 28, 30, 31 April 1, 2, 6, 7, 9, 10, 11, 12

Coda was used to detect porpoise clicks for the tidal turbine data from March 25 to April 13, 2017. The porpoise clicks detections were noted for four days in March and 11 days in April (Table 4). The analysis focused on these dates because the hydrophone sampling rate was remotely increased on March 8, 2017

Page | 17

and then increased further on March 24, 2017. This increased sampling rate for the waveform data was needed because early analyses indicated that the original sampling rate was suboptimal for use in the click detector programs. March and April therefore utilized the optimal sampling rate and were the months focused on for further analysis. Table 4. Days with Porpoise Click Trains as Detected by Coda Month Days with Porpoise Train Detection March 25, 27, 28, 31 April 1, 2, 5, 6, 7, 8, 9, 10, 11, 12, 13 No other cetaceans were detected during the monitoring period. The results were then used to compare detections against time and current. Results indicate that the hours with the highest proportion of porpoise clicks were from 19:00 to 03:00 (Atlantic Daylight Time), suggesting greater use of the site at night. The most clicks were also noted at -30% current speeds. This means that more porpoise click trains were picked up when current speed was at 30 % of peak flow velocity in ebb tide direction suggesting a greater presence at that time. The results from the different software programs allowed for a comparative analysis Coda appears to be the best program due to its ability to better detect porpoises in the high noise intense environment of the Minas Passage. Ongoing analysis, during the next deployment will provide further opportunity to test this software program and compare results of this preliminary analysis. Issues with cabling and synchronization of the hydrophones meant that localization of identified porpoise clicks could not be completed during this monitoring period. These issues have been addressed for the upcoming deployment (see section 4.1.1). Although the icListen scope was reduced to one hydrophone during deployment, thereby affecting localization capabilities, the working hydrophone was still able to identify harbour porpoise presence in the near-field environment between November 2016 and April 2017. The results from this first deployment have contributed to a further understanding of when porpoises appear to use the Passage and have also provided some early indication of the seasonal frequency of occurrence of harbour porpoise. The icListen did not detect other vocal cetaceans, although it is recognized that a lack of observed sound detected from other vocal cetaceans does not mean total absence. It was also noticed, when comparing results to the FORCE CPOD results, that there was a discrepancy between the results of the CPOD devices and the icListen devices, with the CPODs recording a greater number of clicks during the same sampling period. There is a difference in sensitivity between these two detection systems so it is difficult to compare data results directly. However, trends detected within each system could be compared (i.e., detect trends with CPOD or icListen data first, and then compare the relative trends between units). FORCE and CST will investigate this further in the next deployment to better understand monitoring technologies in high current environments and marine mammal usage of the marine environment in the Crown Lease Area.

Page | 18

It is clear that this marine mammal component requires additional investigation for a better understanding of seasonal frequency of marine mammals and to determine the relationship between harbour porpoise occurrence and turbine operations. The 2018 EEMP will address these specific questions around harbour porpoise detection and will compare the icListen results with FORCE data results to understand discrepancies between CPODs and icListen devices and how the data can be used together for a better understanding of marine mammals in the Crown Lease Area. For future monitoring programs, Ocean Sonics recommended the following:

• Increased protection of icListen hydrophones mounted on the turbine; • Improved on-site data management plan; • Further investigation of the PAMGuard and Coda porpoise click detectors to determine the best

programs for marine mammal detection; • Scheduled real-time access to hydrophones for ongoing diagnostics; and • Further data acoustic analysis to better understand turbine operations and detection programs.

CST will implement these recommendations as part of the 2018 monitoring program as well as a greater collaborative effort with FORCE in data comparison for marine mammals. Gemini Sonar – Active Acoustic Study Acadia University was contracted to review and analyze the data from the Gemini sonar. The report (Appendix C) addresses the fish component of the EEMP. Active acoustic monitoring was used to gather information on the occurrence of fish within the near-field (i.e., < 100 m) area of the turbine. To achieve this, a Gemini multi-beam imaging sonar was mounted on the turbine structure and used to monitor marine life in the near-field area during the 2016/2017 deployment. The Acadia Centre for Estuarine Research at Acadia University was contracted to analyze the data collected by the sonar and address the specific objectives under the CST EEMP to determine the seasonal frequency of fish. The goals of the Gemini sonar study were to increase understanding of potential interactions of marine life with in-stream tidal turbines, including the use of the site by wildlife, as determined by target detection and tracking, and to further develop monitoring methodology as it relates to Gemini data collection, processing, analysis, and presentation. The work described in this report used manually-processed Gemini data to:

• assess trends in target abundance within the sampled volume, over short and long-time scales and with respect to tidal stage and current speed;

• characterize target movement with respect to current direction; and • identify targets that may be fish schools.

Page | 19

Data collected were from the near-field area immediately in front of the turbine (facing into the current during ebb tide). Given the sonar’s downward-angled orientation, the volume sampled was below the depth of the turbine rotor. A subset of the sonar dataset was manually processed to identify and track targets in the volume of water column sampled. This involved a human observer reviewing five-minute long video clips at two-hour intervals for one full day per week, for the full five-month period of data collection. The observer searched for ‘targets,’ which are defined as objects moving independently of the seafloor background that could be marine life. The time of detection and the net movement direction was recorded for each target. Target abundance was found to decrease with falling winter temperatures, which is consistent with other biological surveys of this area. Target abundance did not differ significantly between day and night for the duration of the dataset, but abundance of targets in the volume sampled was consistently lower during the flood tide than the ebb tide, possibly due to effects on the flow field in the area sampled by the sonar as flood tide waters moved through and around the turbine. Target movement direction exhibited patterns that reflected the flow environment, with most targets moving in the same general direction as the current. However, variation in movement direction of targets within the sampled volume was greater during the flood tide, when targets were downstream of the turbine, than during the ebb tide, when targets were upstream (approaching the device). Again, this difference could be related to the physical effect of the turbine on the flow regime in the near-field; examination of fine-scale hydrodynamics upstream and downstream of the device would be needed to determine wake effects. The results of target detection and tracking presented here are encouraging for the future use of the Gemini sonar to monitor marine life presence and behavior at turbine rotor height in the near-field of the CST turbine. Efficiency and extent of sonar data processing will increase greatly with further development and validation of automated Gemini data processing techniques. Assessment of the near-field hydrodynamics, and examination of the data provided by FORCE’s mobile and stationary active acoustic surveys of fish, will be important for the interpretation of Gemini sonar data collected from turbine rotor height in future studies. The potential for an overall improved dataset from a re-oriented Gemini sonar and a longer (planned) deployment of a turbine will provide an opportunity to obtain data with increased spatial and temporal coverage. This will help to clarify the results presented and discussed in this report, improve understanding of year-round presence and spatial distributions of marine animals, and therefore help to meet the overall objective of understanding how fish and marine mammals might interact with the CST in-stream turbine. The purpose of the Gemini sonar was to determine the seasonal frequency of occurrence of fish in the near-field environment of the turbine. Although the orientation of the Gemini sonar created difficulties with an optimal view plane, researchers were able to identify fish throughout the deployment period and make early predictions on fish patterns and abundance. The lessons learned as part of the data analysis will

Page | 20

also be valuable to inform the next deployment as CST develops a better understanding of the how to analyze large amounts of data from this sonar. Acadia identified that there are aspects of data collection, processing, and analysis that can be improved in future applications:

• Data collection: Re-orientation of the sonar will result in high-quality data amenable to automated processing by providing a view of only the water column and reducing interference from solid structures. Further, the modelled current speed and direction data were useful for examining general trends, but high-resolution current speed and direction information, up- and down-stream of the turbine, would help with separating active and passive target behaviors. Communication with the sonar must be high-quality and consistent, which can be difficult when multiple instruments are communicating with shore via the same cable. To address this, pre-deployment testing is key to ensuring that the new cabling and set-up provides improved data acquisition.

• Data processing: Validated, automated processing methods are needed for the Gemini sonar if it is to be used for monitoring purposes. The development of suitable processing algorithms will be facilitated when the Gemini’s view is reoriented to cover only the water column. A certain amount of manual (human) processing will be necessary to validate the results of an automated system, to quantify its error rate relative to a human observer, and to ensure its continued functionality over time. Manual data processing with the Gemini SeaTec software can be improved substantially by allowing measurements to be taken in the magnification window. This would allow many more of the small targets within the first 10 m of the Gemini to be located and measured for inclusion in the dataset.

• Data analysis and interpretation: To detect turbine effects, it will be important to continue to assess temporal and spatial variation in metrics extracted from Gemini data. Combining these data with information on the physical environment (e.g., wake characteristics) will improve identification of active and passive target behaviors and which of those may be responses to the device. Due to the sonar orientation, almost all targets identified were within 10m of the device. With the sonar correctly oriented, future assessments will include tracking of target movements over a wider range of distances from the device. Additionally, with more targets detected throughout the sampled volume, a finer grid could be applied to generate summary statistics and images that are more useful for turbine effect assessment. As automated processing methods are implemented, new metrics will become available for use in behavioural analyses. It will also be important to assess the detection probability of targets throughout the Gemini’s field of view under a range of environmental conditions—for example, at low to high current speeds. This is necessary for understanding potential sources of bias in the results.

• Mid- to Far-Field Monitoring: All results obtained from the near-field of the CST device must be considered in the larger context of Minas Passage and the fish populations that utilize it. Mid- and far-field monitoring are well outside of the CST EEMP requirements, but information from broader-scale studies of the area will help with Gemini data interpretation.

CST will implement these recommendations as part of the 2018 monitoring program and continue to focus

Page | 21

on potential fish interactions with the turbine to demonstrate the usefulness of the Gemini as a monitoring option in high current environments. In addition, CST will implement a greater collaborative effort with FORCE to integrate the Gemini fish data with the data from the FORCE fish surveys for a more comprehensive understanding of the Crown Lease Area. 4.2.2 Operational Sound AMAR Units JASCO Applied Sciences was contracted to complete a data synthesis for an Autonomous Multichannel Acoustic Recorder (AMAR). The AMAR is an underwater acoustic data recorder that, in this study, was moored 100 m from the turbine from November to January to record turbine sound as well as all other available data (i.e., from the icListen hydrophone). The purpose was to develop a model of the turbine source levels (sound) and then use the model to compare the turbine sound levels to tidal current, tidal direction, and turbine operating state to understand at what conditions can a turbine be heard over the ambient sound and how might this relate to marine wildlife. Due to the amount and types of data collected this analysis had to be expanded to be completed over two phases with the second phase planned for 2018. As part of the overall study, JASCO also assessed the methodology of a moored acoustic device (i.e., the AMAR) to drifting devices, to better understand the usefulness of these two techniques for measuring turbine sound in high flow environments. This is important for informing future cumulative sound studies at the Crown Lease Area. The Phase 1 report is attached as Appendix D. Knowledge of turbine operational sound is important to increase understanding of the potential effects to marine wildlife from in-stream tidal turbines and guiding the development of mitigation measures. Since acoustic measurements of tidal turbine sound have so far proven problematic due to contamination from flow noise in high-flow environments, information on methodologies for collecting sound data is also important. Flow-noise contamination can be reduced by collecting acoustic data using drifting sound recorders, but this approach provides only brief measurements that preclude a strong statistical sampling of the emitted noise. Due to the exceptionally high current in the Minas Passage, hydrodynamic high-flow moorings (HFM) were used to evaluate the potential for these devices in future studies. Fixed acoustic recorders enable longer recordings, but the instrumentation is more susceptible to flow-noise contamination in comparison to drifting recorders. JASCO deployed two AMARs in the Minas Passage on November 18, 2016. The recorders were deployed on the seabed 100 m and 680 m from the turbine. The 680m recorder was intended to serve as a control measurement. The recorder at 100m was retrieved on January 19, 2017. The control has not yet been retrieved. The primary objective of the study was to perform a Sound Source Characterization (SSC) of a full-size in-stream turbine operating in different states and at various flows to understand when and at what frequencies the turbine might be distinguishable from the ambient background noise. To assist with the analysis, information on currents and turbine operation were incorporated. This included the normalized

Page | 22

flow speed, current direction, and operating state of the turbine (i.e., not spinning, free spinning, and generating). The combination of information allows for an association between the measured sounds and the turbine operating state. The JASCO analysis of turbine sound from the November 2016 – January 2017 AMAR deployments is on-going and will be completed in 2018. Phase 1 results indicate:

• Turbine sound was indistinguishable from flow noise below 60 Hz. The early analysis suggested that there was a band of frequencies in the 60-300 Hz range where the sound characteristics were likely from the turbine.

• The autonomous recorder 100 m from the turbine was able to distinguish turbine sounds from ambient noise for frequencies above 50 Hz, at all flow rates.

• The turbine operational sounds are different when free-spinning as compared to generating. In the free-spinning state the turbine is louder than the background at frequencies of 50-500 Hz for flow speeds of 20-60% of the full flow. In the generating state, the turbine is louder than the background at frequencies of 50-12500 Hz for flow speeds of 20-60% of full flow. The sound produced by the turbine while generating is dominated by two wide bands, one centered near 1250 Hz and the other centered near 4000 Hz. The 4000 Hz band was higher than background levels at all flow speeds.

• A preliminary comparison to the sound levels near the outer Bay of Fundy shipping lanes was performed. When the turbine was not spinning and currents were below 20% of full flow, the Minas Passage was 15-25 dB quieter in the band of 25-320 Hz, and 5-15 dB quieter for the band of 320-10000 Hz. As the flow speed increased ambient noise levels in the Minas Passage also increased, at all frequencies. For frequencies below 100 Hz the increases were due to flow-noise over the recorder. Above 100 Hz the source of the noise was primarily sediment movement (i.e. pebbles and gravel striking each other). For current speeds above 80% of full flow, the levels measured in Minas Passage were 10-30 dB higher than the lower Bay of Fundy at all frequencies analyzed (10-12500 Hz).

• Above 315 Hz the free-wheeling and not-spinning sound levels are very similar, indicating that the ambient environment is the source of noise in these states. The sound levels increase with frequency as the flow rate increases due to sediment interaction noise (pebbles striking each other). This appears to be the dominant source of noise above 10 kHz even when the turbine is generating. Further work through comparisons with the icListen hydrophones and drifting hydrophones is needed to verify this assessment.

• The threshold for possible temporary hearing loss, such as humans experience at loud music concerts, is also measurable for marine life including harbour porpoise. That level when exposed to continuous sounds from human activity is 153 dB re 1 µPa²·s weighted SEL. The AMAR noted that this level was exceeded on 52 of 63 days in the Minas Passage and when analyzed shows that most of that noise, within the porpoise hearing range, is a result of sediment movement. These results match results of another study completed in Admiralty Inlet. Since the Minas Passage is considered regular habitat for porpoise, it can be theorized that their hearing is not affected, and by extension the temporary hearing loss threshold for continuous human generated sound in the marine

Page | 23

environment may not be appropriate for noise created by sediment movement in high flow environments. While these early results seem positive, the noise created by the environment and by an operating turbine still requires further investigation. This will continue in 2018.

It should be noted that CST turbine was under engineering evaluation during the sound monitoring period. During this time the turbine was engaged a number of operating parameters as CST completed various tests. The results are therefore indicative of this commissioning phase rather than of a true operational stage. Phase 2 of the study will be completed in 2018 and will complete the analysis of the source level of the turbine as a function of the operation state and flow speed. The Phase 2 report will include a full comparison of the data from all hydrophones (i.e., turbine-mounted units and drifters used by FORCE) to provide a more detailed analysis of the soundscape in the Crown Lease Area. The analysis results will be used towards the development of a model of sound emitted by an operating turbine as a function of frequency, current speed and direction, and turbine operating state. The Phase 2 analysis will include a discussion of the impacts of the turbine sound levels on the local soundscape, including:

• The distance at which the turbine sound exceeds the background or ambient noise; and • Estimates of the ranges at which different groups of fish and marine mammals may hear the

turbine. The first phase of the study was important in helping confirm that the turbine sound was not distinguishable at low frequencies. However, the analysis also identified the need for more in-depth analysis of the turbine at different states of operation and during different states of flow. The Phase 1 study also highlighted the high levels of noise created by the ambient environment, especially at high velocities. While more work will be required to enhance the acoustic recorder performance and address the objective of understanding turbine sound, the monitoring was successful in confirming the best method to collect sound data from a deployed turbine over various tidal and operational states. The results from the first deployment indicate that this type of moored, long-term, static monitoring allows accurate characterization of the overall soundscape. Seasonal and tidal trends can be determined, and once the moorings are deployed, the devices can record throughout all weather conditions. The method of static moorings is the best method for measurement and analysis, but the other methods, such as drifting hydrophones, do add to the knowledge about the overall soundscape and should not be discounted. Further knowledge of the soundscape for the Crown Lease Area and how it may change with instream turbine installations will be explored in greater detail in collaboration with FORCE during the next deployment.

Page | 24

5.0 Contingency Monitoring Program The purpose of a contingency monitoring program is to ensure that original commitments under the EEMP are met during times when data from the monitoring devices (e.g., Gemini sonar, icListen hydrophones, etc.) are not accessible. This could happen if devices malfunction or are damaged or if the subsea cable is disconnected for a lengthy (i.e., > 2 weeks) or an unknown period of time, thereby disabling the monitoring devices. Although the present version of the CST EEMP includes a contingency plan around damage to monitoring instrumentation, the lessons learned during the retrieval operations in Spring 2017 identified the need for an enhanced plan. Formal discussions between CST, NSE, DFO and FORCE were undertaken in November to address this matter and are currently on-going. During the time leading up to the next deployment CST will continue to identify potential contingency options, evaluate these options, and work through the details towards ensuring that EEMP commitments are met. A draft updated contingency program is being developed in collaboration with FORCE and will be provided to NSE and DFO for discussion early in 2018 and will be finalized prior to deployment. The updated contingency program will address monitoring during the following unexpected events:

• Damage or loss of environmental devices (detectable through remote monitoring); • Short-term gaps in monitoring caused by preparatory activities related to deployment and

retrieval operations; and • Longer-term gaps in monitoring caused by delays in deployment and retrieval operations.

Once agreed that the plan is appropriate, CST will execute the necessary steps to ready that plan for implementation as part of the planning process for the next deployment. A new section will also be added to the updated CST EEMP to address contingency planning as part of future operations.

Page | 25

6.0 Additional Items

6.1 Updated EEMP In recognition of recommendations provided by DFO in a letter to NSE (April 2017) CST is committed to continuous evaluation and analyses of the positioning of the Gemini sonar device. The device has been repositioned and will be tested prior to the next deployment. Results of those tests will be discussed with regulators and the research team. CST is also committed to investigating and discussing supplemental monitoring components to complement the present monitoring program to better understand potential interactions. An updated EEMP is currently under development for 2018 and a draft version will be submitted for review to NSE, DFO and the NS Department of Energy prior to the next deployment.

6.2 Data Management An improved data management plan is under development for the Project. Lessons learned from the previous deployment highlighted the need to address this element of the EEMP. The improved data management plan will include:

• Identification of practices or workflows to help manage data throughout the deployments; • Data storage and backup; • Sharing data including access; and • Data preservation and long-term access.

The data management plan will be continuously updated as needed, as CST continues to better understand the amounts of data created from the Project and how data sets can be transferred and stored. In recognition of the need for improving this process, CST has engaged IT specialists to assess the process and provide potential solutions that can be implemented for the next deployment to create a more streamlined process for delivering data to researchers for preliminary and final data analysis. This involves working with FORCE to understand access at the test site and how this may be facilitated; exploring other possible integrations, and sending data to a secure web-based site for access by the researchers.

Appendix A 2017 Quarterly Reports (Q1, Q2 and Q3)

Cape Sharp Tidal Venture:

Environmental Effects Monitoring Program – Q1 Report

April 1, 2017

CSTV EEMP: 2017 Q1 Update Report

Page 2 of 17

Executive Summary

Since November 8, 2016, the Cape Sharp Tidal demonstration turbine has been producing electricity for export to the Nova Scotia Power grid and gathering important environmental data at the FORCE site. The Environmental Effects Monitoring Program (EEMP) developed for the Project aims to monitor and better understand potential environmental effects and interactions of specific environmental components (fish, marine mammals, operational noise) in the near-field area (i.e., 0-100-metres) of the OpenHydro Open-Center instream tidal turbine. The overall research objective of the monitoring program is to verify the accuracy of environmental effect predictions made in the 2009 Environmental Assessment (EA) report that tidal demonstration will not result in serious harm1 to fish or negative effects to marine mammals. The purpose of this interim report is to provide an update for the first quarter of the project during its initial commissioning phase (November 2016 – January 2017). This update includes information on the monitoring equipment being tested, preliminary results and mitigative measures. The annual environmental monitoring report (due January 1, 2018) will reflect a year long period of data collection in which behavioural and species analysis will be provided in greater detail. We are working with industry-leading local and international experts in marine technology in tidal environments to collect and interpret this data. First-quarter insights include:

• Deployment of the first turbine has enabled us to test the suite of environmental and turbine sensors together for the first time, and in turn develop innovative methods to advance data collection and correlation.

• In this first period of data collection, analysis for fish and marine mammals has focused on understanding and improving near-field localization, decreasing interference from the environment and other instrumentation, and refining algorithms to achieve data clarity.

• The Gemini Tritech sonar has operated successfully since turbine deployment. Data from this imaging sonar will be used to determine the seasonal frequency of marine fish and marine mammals in the turbine’s 100-metre vicinity, and results will analyze how marine wildlife interacts with the turbine. Early results indicate positive identification of marine life within the sonar view plane. A longer period of study with greater numbers of marine life detections is required to interpret behaviours and species, or draw scientific conclusions about how these organisms are interacting with the turbine.

• The four OceanSonics icListen hydrophones have produced acoustic (sound) data that will be used to determine the seasonal frequency of occurrence of harbour porpoise in the near-field area as well as to determine the relationship between harbour porpoise occurrence and turbine operation. Early results indicate that harbour porpoise clicks were detected in November and December. No other marine mammals were acoustically identified from November to January.

1 Department of Fisheries and Oceans reviewed the CSTV project application under the Fisheries Act and the Species at Rick Act, and determined a demonstration-scale tidal project would not result in serious harm, as defined under the Fisheries Act, to fish and fish habitat.

CSTV EEMP: 2017 Q1 Update Report

Page 3 of 17

• All four acoustic Doppler current profilers (ADCP), used to support analysis for other devices and record current velocities over a depth range, have been functioning and providing data on flow regimes at the turbine location.

• An autonomous multichannel acoustic recorder (AMAR) has been monitoring turbine operational sound for two months. This data will be compared to data from a second control AMAR unit, and other devices, to understand turbine sound at varying tidal states and current speeds. One AMAR has already been recovered and preliminary results suggest that turbine operational sound is indistinguishable from flow noise below 60 Hz. In the case of the Minas Passage therefore, this could mean that the sound produced by the turbine is lower than the ambient or natural noise for most of the tidal cycle.

• A video camera was positioned on the turbine in front of the rotor to assess the capability of an optical camera at the demonstration site. We have been unable to communicate with that device since deployment.

• The turbine is performing well during commissioning. In April, we’ll temporarily retrieve the device from the FORCE site to make minor repairs and upgrades to some Turbine Control Centre2 (TCC) components, which will also provide us opportunity to examine and adjust, where needed, the monitoring instrumentation. This will be the right opportunity to reposition the sonar in particular, investigate why we can’t communicate with the optical camera, and inspect the contingency hydrophones.

These are early days in this important research initiative that will build on and complement the existing 100+ baseline studies completed at the FORCE site and an international body of research on in-stream tidal energy. We remain committed to drawing scientific conclusions about the monitoring program from the detailed data analysis and providing empirical evidence to support and demonstrate those findings.

2 The TCC is an electrical component sub-system attached to the subsea base and connected to the turbine which allows for the transformation of raw electrical power from the generator into grid-compatible AC power, as well as transmit operational and monitoring data in real time. By treating information coming from the multiple sensors located within the turbine system, the TCC is used to optimize the power output in any operating conditions. This is the first time OpenHydro’s pioneering technology has been used anywhere in the world. Its design has been a critical step forward in being able to generate electricity from multiple turbines at sea and export to shore via a single export cable.

CSTV EEMP: 2017 Q1 Update Report

Page 4 of 17

1.0 Introduction Cape Sharp Tidal Venture (CSTV), a joint venture between Emera Inc. and OpenHydro, deployed one, 2 megawatt (MW) instream tidal energy turbine at the Fundy Ocean Research Center for Energy (FORCE) site near Parrsboro, Nova Scotia on November 7, 2016 (the Project). The Project is the beginning of a 4MW, two-turbine demonstration phase (Phase 1) of an ongoing tidal energy initiative in Nova Scotia.

2.0 Operational Update CSTV is currently in its commissioning phase. During this time, OpenHydro has focussed on ramping up the power production of the machine and establishing normal operating parameters through its control system. Tests have also been done on the monitoring technologies (i.e., sonar, hydrophones, video camera). In the first three months, the turbine has logged approximately 1500 hours of operation. In Spring 2017, OpenHydro will temporarily retrieve this first turbine to perform minor repairs and upgrades to some of the Turbine Control Center (TCC) components. The work will take place in Saint John, New Brunswick, and then the turbine will be redeployed at the FORCE site. The retrieval will provide an opportunity to take monitoring learnings from the first quarter of operations and to examine and make adjustments or modifications to the monitoring devices located on the turbine. In conjunction with FORCE, CSTV is also working to integrate, test, qualify and deploy a second Gemini imaging sonar on one of the FORCE Fundy Advanced Sensor Technology (FAST) platforms. Details about testing will follow. The device will be deployed with the FAST platform and positioned near a turbine, providing a view plane of the side of the turbine structure. The purpose of this initiative is to explore the potential for monitoring the close vicinity of one side of the turbine rotor.

3.0 Environmental Effects Monitoring Program

3.1 Context

The Environmental Effects Monitoring Program (EEMP) developed for the Project aims to monitor potential environmental effects in the near field area (i.e., 0-100-metres of the turbine) to better understand potential effects and interactions of specific environmental components (fish, marine mammals, operational noise) with the OpenHydro Open-Center in-stream tidal turbine. The overall research objective is to verify the accuracy of environmental effect predictions made in the FORCE 2009 Environmental Assessment (EA) report, but the monitoring program will also assist with increasing knowledge about monitoring methods and analysis, development of mitigative measures, and building technical knowledge within the local tidal industry. The FORCE EA Report and subsequent 2010 FORCE EA Addendum are available here: http://fundyforce.ca/environment/enviromental-assesment/. As required by the conditions of the FORCE EA Approval (2009), the CSTV EEMP was developed in collaboration with experts in the field of instream tidal energy and with input from government agencies, including Fisheries and Oceans Canada (DFO) and Nova Scotia Environment (NSE), as well as other instream tidal energy interests including the Offshore Energy Research Association of Nova Scotia (OERA), FORCE, and FORCE’s independent Environmental Monitoring and Advisory Committee (EMAC).

CSTV EEMP: 2017 Q1 Update Report

Page 5 of 17

The CSTV EEMP forms a component of FORCE’s EEMP commitment under the FORCE Environmental Management Plan. Both EEMPs have been designed to be complimentary in order to achieve the most meaningful examination of potential effects and in consideration of the extensive baseline studies that have taken place at the FORCE Crown Lease Area since 20083. The EEMP’s cornerstone is an adaptive management approach to evaluate data and make informed, science-based decisions to adjust technology and monitoring methods, assess mitigation measures and address concerns as necessary. This approach is necessary because of the unknowns and difficulties inherent with gathering data in harsh tidal environments, such as the Minas Passage. It allows for adjustments and constant improvements to be made as knowledge about the system and environmental interactions become better known. This is important as we start to analyze the first data sets. Feedback from the sensors is providing critical information on performance under harsh conditions, amount of data being gathered, and where improvements and advances can be implemented. Although we are at the very early stage of learning about the operating conditions and data acquisition ability of the sensors, we are already increasing our understanding of marine wildlife behaviour in the vicinity of the turbine. The EEMP is reviewed continuously with regulators and FORCE and where required, approaches and methodologies will be revised on the basis of accumulated experience and observed progress toward achieving the monitoring objectives. This adaptive approach will assist with resolving knowledge gaps of the potential effects of the Project and usefulness of mitigation measures. The approach will also facilitate the design and implementation of new or modified monitoring strategies. Additional information about the scope of the EEMP, methodologies and background on the monitoring technologies is available in the EEMP document available on the CSTV website: http://capesharptidal.com/eemp/. Table 1 provides an updated summary of the overall objectives of the CSTV EEMP and how each objective aligns with similar objectives implemented as part of the updated FORCE EEMP. Additional components specific to the FORCE EEMP are available on the FORCE website: http://fundyforce.ca/wp-content/uploads/2012/05/FORCE-EEMP-2016.pdf.

3 The FORCE website provides a list of baseline studies: http://fundyforce.ca/environment/research/

CSTV EEMP: 2017 Q1 Update Report

Page 6 of 17

Table 1. EEMP Summary of Objectives for CSTV and FORCE Responsibility Objective Scope Methodology Timing Marine Fish FORCE 1. To quantify fish

distributional changes that reflect behavioural responses to the presence of a deployed tidal instream energy converter (TISEC) device.

2. To estimate the

probability of fish encountering a device.

Fish density and vertical distribution. Estimate probability of fish encountering a device.

Down-looking, vessel- towed hydroacoustic echosounder along parallel transects. Before-After-Control-Impact (BACI) study design; and Multivariate analysis (Hotellings T2 tests) of fish vertical distributions. Use of an encounter probability model.

Surveys will be distributed over the year. Each survey to be completed over a full tidal and diel cycle (i.e., 25 hours).

CSTV 1. Determine the seasonal frequency of occurrence of fish within the near-field environment of the turbine(s).

2. Integrate data-sets into a

strike risk model for fish.

Use of fish detection algorithms for fish frequency Analyze data to describe track trajectories for use in the strike risk model.

Active Acoustic Monitoring (AAM).

Data analysis software and use of seasonal data from AAM data-set, as well as a model including track trajectory data.

Start date coincided with operations and runs continuously. Data analysis initiated. Data analysis will be initiated as soon as possible – exact date to be determined

Marine Mammals FORCE 1. Assess direct effects of

operational turbine sound: attraction or avoidance.

2. Assess indirect effects due to changes in prey distribution and abundance: attraction or avoidance.

Determination of possible permanent avoidance of the mid-field study area. Change in distribution of a portion of the population: large scale (~50%) decreases or increases in relative occurrence as measured via echolocation activity levels across the mid-field study area, including in the vicinity of operating turbines.

Deployment of C-PODs to provide a comparative ‘after’ data set compared with data collected pre-turbine deployment. Deployment of one C-POD 100+m from any occupied berth. Beach walks and observations of marine mammals in proximity to site.

Deployments throughout the year. Each deployment is anticipated to be roughly 3-4 months in length.

CSTV 1. Determine the seasonal frequency of occurrence of harbour porpoise within the near-field environment of the turbines.

Identification of harbour porpoise clicks compared to baseline detection rates [Note: a data quality assessment will be performed to ensure use of acoustic data from the hydrophone

Integrated AAM and Passive Acoustic Monitoring (PAM).

Start date coincided with operations and runs continuously. Data analysis initiated.

CSTV EEMP: 2017 Q1 Update Report

Page 7 of 17

Responsibility Objective Scope Methodology Timing 2. Determine the

relationship between harbour porpoise occurrence and turbine operations.

3. Determine the seasonal

frequency of other vocal cetaceans (e.g., white-sided dolphins).*

4. Determine the seasonal

frequency of occurrence of marine mammals within the near-field environment of the turbines.

5. Integrate data-sets into a

strike risk model for marine mammals.

with the highest quality (i.e., least flow/sediment noise) by comparison of detection data from all hydrophones over 1 month]. Detections obtained by identifying click trains and compared to rates of porpoise detection at 4 representative speed categories using current or RPM data from the turbine (Note: categories to be determined). Identification of vocalizations. Use of marine mammal detection algorithms for marine mammal frequency to discern different species, frequency and evasion behavior. Analyze data to describe track trajectories for use in the strike risk model.

Integrated AAM and PAM. Integrated AAM and PAM. Data analysis software and use of seasonal data from integrated PAM/AAM data-sets, as well as a model including track trajectory data.

Start date coincided with operations and runs continuously. Data analysis initiated. Start date coincided with operations and runs continuously. Data analysis initiated. Data analysis will be initiated as soon as possible – exact date to be determined. Data analysis will be initiated as soon as possible – exact date to be determined.

Sound FORCE Establish pre-deployment

baseline ambient noise conditions. Use the noise data to verify the EA predictions that suggest turbine sound will not negatively affect marine biota.

Measurement of ambient noise within the CLA.

Undertake simultaneous drifting hydrophone measurements for comparison and data validation.

Surveys conducted before and after device installation to measure changes in the soundscape.

CSTV Characterize operational turbine sound to assess the effect of turbine operations on the noise profile of the site.

Comparison of sound levels will be completed for a control site and in the near-field environment to identify the frequencies at which the turbine sound is discernable. Received sound levels will be correlated with tidal state and current

A high-flow mooring design using 2 fixed autonomous acoustic recorders to collect acoustic measurements will be used. One recorder will be placed at 100 m range from the turbine and the other at a control location (approximately 680 m away).

Following deployment for a period of two months to provide data collected in a variety of sea conditions and during different turbine operational phases.

CSTV EEMP: 2017 Q1 Update Report

Page 8 of 17

Responsibility Objective Scope Methodology Timing Determine flow at specific noise levels.

speed to assist with characterization of sound levels attributed to flow noise. Flow data will be collected simultaneously.

Flow data to be collected using Acoustic Doppler Current Profilers (ADCPs) located on the turbine structure.

* It is anticipated that detection rates for other vocal cetaceans will be too low for detailed current speed analyses.

4.0 Q1 Monitoring Update The monitoring program was initiated following deployment on November 7, 2016. An update is provided for the first three months (November 7, 2016 to January 27, 2017) in the following sections.

4.1 Monitoring Technology As noted in Table 1, the monitoring devices on the turbine are a combination of passive acoustic (PAM) (Ocean Sonics icListen hydrophones) monitors and active acoustic (AAM) (Tritech Gemini imaging sonar) monitors. These sensors are co-located on the turbine to create an integrated monitoring system that collects data specific to marine mammals and fish. In addition to these devices, four acoustic doppler current profilers (ADCPs) are mounted on the turbine to provide data on flow, and a video camera, mounted on the subsea base, and facing the rotor, was added in an attempt to film the turbine rotor. Data from all monitoring devices is collected continuously through a fiber optic data cable contained within the subsea cable. Data is logged on-shore to hard drives and remotely saved to an OpenHydro server. Turbine operational sound was monitored by a separate deployment of a high flow mooring design of an autonomous multichannel acoustic recorder (AMAR). This device was deployed 100 metres from the turbine and data results will be compared to another identical device deployed at a control location, 680 meters away, where no turbine sound is present. Data storage is achieved within the instrumentation of these units. Current data collected simultaneously by the ADCPs will allow for sound levels to be compared between the two AMARs to identify the frequencies at which the turbine sound is discernable. Received sound levels will be correlated with tidal state and current speed to assist with characterization of sound levels attributed to flow noise.

CSTV EEMP: 2017 Q1 Update Report

Page 9 of 17

4.1.1 icListen Hydrophones Four icListen hydrophones (PAM devices) developed by Ocean Sonics are positioned on the turbine. The use of four devices created an opportunity to investigate various positions on the turbine, the potential for integration with other monitoring devices and how much potential exists for localization of porpoise sounds under various tidal and operational conditions. An additional benefit is the redundancy. In the event that one or more of the hydrophones is damaged, sufficient data is still collected to meet monitoring requirements. Communication with and data collection from the two devices located on the starboard and port sides of the turbine have been functioning well and consistently during the first quarter while a third device on the subsea base is delivering data intermittently. The fourth device, located on top of the turbine has not responded to communication and this will be investigated during the retrieval activities (refer to Section 5). Ocean Sonics has begun assessment of the first data sets from the hydrophones to prepare for further analysis.

4.1.2 Gemini Imaging Sonar The Tritech Gemini sonar was mounted within a protective enclosure slightly forward of the turbine and pointed towards the ebb tide direction4. This device monitors an area up to 60 m in front of the turbine that is approximately 104 m in width. Species detection ranges from a lower limit of approximately 10 cm in length and upwards. The sonar has operated consistently throughout the first three months of deployment. Initial data results indicate that an adjustment to the present position would provide a broader view plane of the water column and so the sonar will be repositioned during the upcoming retrieval activities (refer to Section 5). Tritech Ltd. will complete the raw data analysis to prepare the data for a later biological analysis. This first step in data analyses is required to confirm identifications of marine mammals and/or fish and to remove those identified as ‘false positives’ (i.e., objects within the water column that are not marine wildlife).

4.1.3 Video Camera Communication with the SAIS IP-CAM HD Ethernet underwater video camera has been unsuccessful since deployment; therefore no data has yet been logged from this device. This will be further investigated during the upcoming retrieval activities. This camera was a supplemental component of the monitoring program and given the harsh environmental conditions it may not be suitable to continue its use. A camera of this same model was used during deployment operations and has shown that given the level of sediment and depth of the turbine, visibility at the site is less than three feet. As such, we do not expect a video camera to be able to provide meaningful benefits or results; however, installing a video

4 The ebb tide was the direction chosen for the sonar as baseline studies showed a greater likelihood of species detection during this time and less turbulence that may affect detection.

CSTV EEMP: 2017 Q1 Update Report

Page 10 of 17

camera was a commitment made to stakeholders before deployment, and we will continue to investigate opportunities and technologies to enhance our visibility of the rotor.

4.1.4 ADCPs All four ADCPs have been functioning and providing data on flow regimes within the Minas Passage. This information on current velocities will be used to support analysis for all other devices.

4.1.5 AMARs The AMAR units were planned for a two-month deployment to capture turbine operational sound during various tidal regimes and operating modes. Both units were deployed on November 18, 2016 and one (closest to the turbine) was successfully retrieved at the end of January 2017. Weather conditions have prevented safe retrieval of the second (control) unit. When it is successfully retrieved, JASCO Applied Sciences (JASCO) will complete a detailed sound analysis once both data sets are available. 4.2 Q1 Preliminary Results The ability to gather data from monitoring devices in real time allows scientists, turbine technology manufacturers and developers to better understand technological and data requirements to support environmental monitoring. The large amounts of data generated so far has given the project team an opportunity to refine the process and develop protocols to efficiently manage, store and share that data. This data collection and subsequent analysis is ongoing. Understanding the interactions between marine organisms and instream tidal turbines is key to developing tidal as a safe, renewable energy source. Monitoring results will inform CSTV on the types of potential effects that may occur to fish and marine mammals. The results will also provide the basis by which future instream tidal energy projects can evaluate cumulative effects of turbine operations on the environment which will facilitate the development of effective mitigation, if necessary.

4.2.1 Marine Fish and Marine Mammals Data is collected from all devices and saved continuously via a data cable. The majority of data analysis for fish and marine mammals in the first quarter has been focused on understanding and improving near-field localization, decreasing areas of interference from the environment and from other instrumentation, and refining algorithms in order to achieve data clarity and, in time, improve the overall data collection and monitoring program. This is required in order to coordinate and integrate the passive and active acoustic devices in to fully understand potential near-field interactions with the turbine. This data will then be considered in describing evasion behaviour and subsequent incorporation into future strike risk models (refer to Table 1). The monitoring devices are collecting large amounts of data which requires ongoing data analysis. Results are considered preliminary and will be refined to present an analysis in the final annual environmental report.

CSTV EEMP: 2017 Q1 Update Report

Page 11 of 17

The Gemini Tritech imaging sonar has operated successfully during the first three months of turbine operation. Data from this imaging sonar will be used to determine the seasonal frequency of marine fish and marine mammals in the near field and, later on, will be used to analyze how marine wildlife interacts with the turbine. Raw data from the Gemini unit analyzed by an automatic detection software and then examined by human observation to determine what parts of the data are false positives; identifications that are actually non-life such as general debris in the water column (e.g., sticks etc.). Preliminary analyses of results from November indicate positive identification of marine life within the view plane of the sonar. The data from November 2016 has been analyzed by the software and a subset of this data has been further analyzed by Tritech so that a complete 24 hour period is available for the time period of November 17 to 30, 2016. This subset of data (November 17-30) was selected for closer analysis because these days are when the maximum and minimum number of detections in the water column occurred (Table 2). The total number of detections during this time period is 2613 but the number is high due to feedback or interference from the seabed. Results so far indicate that too much of the sea bed is visible in the sonar view. This creates a high level of false positive detections, especially around the time the elevation of water level changes (Table 2). Since some areas of the sea bed are of a particular topography that creates a lot of turbulence in the water and this turbulence is detected by the algorithms causing a vast majority of false positives to occur in these areas. Table 2. Summary of number of detections by Gemini Sonar from November 17-21, 2016*.

Date # of Detections During Water Level Elevation Change # of Detections during other times Total

Detections 17/11/2016 647 19 666 18/11/2016 243 11 254 19/11/2016 52 7 59 20/11/2016 22 4 26 21/11/2016 12 6 18 22/11/2016 59 6 65 23/11/2016 75 8 83 24/11/2016 84 11 95 25/11/2016 139 5 144 26/11/2016 324 7 331 27/11/2016 165 6 171 28/11/2016 132 7 139 29/11/2016 143 12 155 30/11/2016 394 13 407

*The dates 17th and 21st of November were chosen because they contain the most and least number of overall detections, respectively.

Further analysis of the results focused on two particular time periods; both chosen in the context of the number of detections and in reference to the changing water elevation. The first period is from November 17 from 06:00 to 14:00; it contains a high level of detections and covers a period that starts

CSTV EEMP: 2017 Q1 Update Report

Page 12 of 17

immediately after a tide elevation change, encompassing the following change cycle, and finishes just before a second change. The second period is November 21 for 24 hours; the results from this day contained the fewest number of detections consistently across the whole period. There are a number of false positives that occur during this time, almost all of which are located around the same turbulence caused by seabed areas around the times of sea level change (Table 3). Marine life has been identified in a number of these time periods for both days but classification as to species or indeed distinction between fish or marine mammal is ongoing.

Table 3. Summary of hourly segments for November 17 and 21, 2016. Date Time Detections Analyzed False Positives Potential Marine Life November 17, 2016 06:00 7 1 5

07:00 10 7 3 08:00 40 39 1 09:00 7 7 0 10:00 2 1 1 11:00 1 0 1 12:00 1 0 1 13:00 2 0 2 totals 70 55 14

November 21, 2016 00:00 1 0 1 02:00 1 0 1 04:00 2 0 2 05:00 2 0 2 06:00 3 3 0 07:00 0 0 0 11:00 1 0 1 13:00 1 0 1 18:00 1 1 0 19:00 4 4 0 20:00 1 1 0 23:00 1 1 0 totals 18 10 8

As noted above, ongoing analysis will be completed to determine whether positive identifications are fish or marine mammals and how these organisms are interacting with the turbine. When the turbine is retrieved, we will take the opportunity to reposition the sonar with the objective of reducing the amount of seabed in the sonar view. This will decrease interference from the seabed and broaden data collection. This is also expected to facilitate data analysis as the numbers of false positives (i.e., identifications of non-living detections in the water column) are decreased.

CSTV EEMP: 2017 Q1 Update Report

Page 13 of 17

Deployment of the Ocean Sonics icListen hydrophones has been successful during the first three months of turbine deployment. As noted in Section 4.1.1 there has been no communication with one of the two contingency hydrophones, which will be investigated at the time of turbine retrieval. The main objective of these devices is to detect harbour porpoise vocalizations. As the most commonly occurring marine mammal at the FORCE test site, this species is the primary marine mammal species of interest in relation to instream tidal operations (Porskamp et. al., 2015). Data from the hydrophones will be used to determine the seasonal frequency of this species, and other vocalizing marine mammals (i.e., whales) in the near field and will support data results associated with the Gemini imaging sonar to understand how marine wildlife interacts with the turbine. During the first quarter the icListen hydrophones positively identified harbour porpoise vocalizations in the Minas Passage on eight different occasions. These identifications were made on the following days: November 10 and 11, and December 23, 24 and 25, 2016. A total of 17 identifications were made with the majority of identifications occurring in November. Although we know that harbour porpoise utilize the Minas Passage throughout the year, this data will increase our understanding of their seasonal frequency – what times, or seasons, of the year is this species more likely to be present in the Minas Passage and what the abundance might be. Data will need to be compared among the units and to the Gemini imaging sonar for further analysis. Although other vocalizing marine mammals such as fin whales have been seen in the Minas Passage, none were acoustically identified during the first quarter. Although some positioning could be refined with input from the other two icListen units, data collection from the two hydrophones is providing sufficient data to make these determinations. Opportunities for improvements to decrease interference from other instrumentation on the turbine are currently being investigated. 4.2.2 Operational Sound JASCO deployed two acoustic recorders on November 18, 2016 to perform a Sound Source Characterization (SSC). The SSC measured underwater sound levels from the sound source; the instream tidal turbine. The goal of this study is to characterize the sound from the turbine operation specifically, which includes mechanical and vibrational sound, and compare that sound to the ambient (natural) sound created by the environment. Since the natural noise in a high flow environment is likely to be quite loud, it is possible that turbine operation sound might be masked or insignificant in comparison. In addition, knowledge of the sound profiles of individual turbines will contribute to understanding about cumulative, or collective, sounds. The AMARs were moored at fixed locations 100 m and 680 m (control site) from the turbine on the seabed. The recovered recorder captured sounds from the turbine, mechanical vibrations, and ambient noise. The recordings provide an assessment of the variability in the sound levels from the turbine over its daily operating cycle and in different current and sea states. Sound is most commonly described using the sound pressure level (SPL) metric. Underwater sound amplitude levels are commonly measured in decibels (dB) relative to a fixed reference pressure of p0 = 1 μPa. The root-mean-square (rms) SPL is used to quantify the sounds generated by the turbine. Preliminary results from the recovered unit, from near the turbine, show that low frequency noise dominated during the entire study period (18 Nov 2016 to 19 Jan 2017) and that the differences over a

CSTV EEMP: 2017 Q1 Update Report

Page 14 of 17

short time, shown in the band level plot, indicate that large variations were due to changes in tide state (Figure 1).

Figure 1. In-band sound pressure levels (top) and spectrogram (bottom) for recovered AMAR unit.

The sound exposure level (SEL) plot (Figure 2) shows the total and vessel-based SEL for each day of the study. As indicated in the figure, vessel detections were not frequent and did not contribute significantly to the overall sound exposure level.

Figure 2. Total and vessel-associated daily sound exposure levels (SEL) and equivalent continuous noise levels (Leq) from recovered unit. Note: Peaks in the SEL plot coincide with full moons on the 13 Dec 2016 and 12 Jan 2017.

CSTV EEMP: 2017 Q1 Update Report

Page 15 of 17

To compare tidal state noise and turbine operation for a specific date, band level and spectrogram plots were generated for November 18 to November 20, 2016 (Figure 3). Each tidal cycle can be seen clearly by large increases in the low-frequency noise due to flow and noise created as rocks and sediment are shifted by the extreme currents. Horizontal lines in the spectrogram, from 60–300 Hz are tones, or sound, believed to be associated with turbine operation. These tones increased and decreased slightly throughout each tidal cycle, likely associated with changes in the turbine rotation rate as the current increased and decreased from high tide to slack tide, respectively. The tones were distinctly different from the more broadband, atonal flow noise visible in the spectrogram as deep red up to 60 Hz. If turbine noise exists below 60 Hz, it is difficult to distinguish it from the flow noise. For comparison, the noise from vessel Nova Endeavour at the beginning of this time period is apparent as spikes in the 10–100 Hz and the 100–1000 Hz bands (these spikes are absent in subsequent tidal cycles).

Figure 3. In-band sound pressure levels (top) and spectrogram (bottom) for 18–20 Nov 2016. These preliminary results suggest that turbine sound is indistinguishable from flow noise below 60 Hz. In the case of the Minas Passage therefore, this could mean that the sound produced by the turbine is lower than the ambient or natural noise for most of the tidal cycle. In the absence of data from the second recorder, at this time, and to provide a comparison of turbine sound at high flows to other sound contributors, it is possible to estimate the source level of the turbine sound by back-propagating the measured sound levels from 60–300 Hz. JASCO performed this analysis and for reference, the highest source levels were compared to the source levels of a 65 m long survey vessel the Setouchi Surveyor transiting at full speed (Hannay et. al., 2004). Results showed that the

CSTV EEMP: 2017 Q1 Update Report

Page 16 of 17

highest source levels from the operating turbine would only be reached in the loudest 5% of conditions, and that these source levels would still not be as loud as a transiting survey vessel. This indicates that in times of increased flow noise (i.e., from 60 Hz to 300 Hz) sound likely to be associated with turbine operation during these higher velocities can be detected but that turbine sound is still creating less noise in that frequency range when compared to other sound contributors in the area. These preliminary results are promising for understanding the potential effects of turbine sound in the Minas Passage and for understanding the sound profile of the turbine during various flow regimes. However, additional analysis will need to be completed once the second AMAR unit is recovered and with flow data from the ADCPs to provide a final characterization of operational sound, confirm how it changes with flow speed and how levels compare to those of levels of natural noise and other sound contributors specific to the Minas Passage. 5.0 Mitigation A number of mitigative measures have been identified to adapt the present monitoring program moving forward:

i. The ability to remotely access some of the instrumentation provides researchers with the ability to adjust frequency settings and explore options to avoid interference with other instrumentation. This work is currently ongoing.

ii. With the opportunity to access the turbine and subsea structure during the planned temporary retrieval, the following activities are planned:

• repositioning the Gemini sonar; • examination, issue identification, and repair (if needed) of two of the hydrophones

and associated cables and connectors; and • examination of the video camera and associated cables and connectors.

6.0 Additional Items Several other initiatives are underway to complement the present monitoring program and to continue to meet the commitments made to stakeholders prior to deployment:

• Participating in FORCE’s beach walk monitoring program; and • Providing input to a workshop led by with FORCE and Acadia University, regarding fish detection

and population modelling; • Continuing to invest in new marine technologies and innovation such as partnering with FORCE

to explore deploying a second Gemini sonar; • Continuing to engage with fishery stakeholders to create improved consultation processes; • Staggering a second turbine deployment to allow additional opportunity for data monitoring,

collection and analysis.

CSTV EEMP: 2017 Q1 Update Report

Page 17 of 17

7.0 References Hannay, D., A. MacGillivray, M. Laurinolli, and R. Racca. 2004. Sakhalin Energy: Source Level Measurements from 2004 Acoustics Program. Version 1.5. Technical report prepared for Sakhalin Energy by JASCO Applied Sciences.

Porskamp, P., A.M. Redden, J.E. Broome, B. Sanderson and J. Wood. 2015. Assessing marine mammal presence in and near the FORCE Lease Area during winter and early spring – addressing baseline data gaps and sensor performance. Final Report to the Offshore Energy Research Association and the Fundy Ocean Research Center for Energy. ACER Technical Report No 121, 35 pp, Acadia University, Wolfville, NS, Canada.

Cape Sharp Tidal Venture

Environmental Effects Monitoring Program – Q2 Report

July 1, 2017

Executive Summary

Since November 8, 2016, and throughout the second quarter of the Project, the Cape Sharp Tidal demonstration turbine has been producing electricity for export to the Nova Scotia Power grid and gathering important environmental data at the FORCE site. The Environmental Effects Monitoring Program (EEMP) developed for the Project aims to monitor and better understand potential environmental effects and interactions of specific environmental components (fish, marine mammals, operational noise) in the near-field area (i.e., 0-100-metres) of the OpenHydro Open-Center instream tidal turbine. The overall research objective of the monitoring program is to verify the accuracy of environmental effect predictions made in the 2009 Environmental Assessment (EA) report. The purpose of this Q2 interim report is to provide a status update for the second quarter of the Project (February 2017 – April 2017). The annual environmental monitoring report (due January 1, 2018) will reflect data collection and learnings from 2017. We are working with industry-leading local and international experts in marine technology in tidal environments to collect and interpret this data. Second-quarter insights include:

• The turbine continued to perform well during the Q2 commissioning period. • In the beginning of April operations were initiated to retrieve the turbine to make minor repairs

and upgrades to some Turbine Control Centre1 (TCC) components. This will also provide us opportunity to examine and adjust, where needed, the monitoring instrumentation.

• Testing and data collection from monitoring sensors continued throughout Q2. • As in Q1 there was no communication from the video camera and one hydrophone. A second

hydrophone was intermittent. Two hydrophones and the Gemini sonar continued to operate until the sub-sea cable was disconnected in preparation for turbine retrieval.

• This second quarter has been focussed on identifying areas where the data collection process can be refined and to develop protocols to efficiently manage, store and share that data. This has included software improvements and successful adjustments to monitoring instrument frequencies for improved data collection. Additional work has also been completed on preliminary analysis and data interpretation.

• A contingency monitoring program was developed to be implemented during times of disconnection but when the turbine remains deployed.

These are early days in this important research initiative that will build on and complement the existing 93+ baseline studies completed at the FORCE site and the international body of research on in-stream tidal energy. We remain committed to drawing scientific conclusions about the monitoring program from the detailed data analysis and providing empirical evidence to support and demonstrate those findings.

1 The TCC is an electrical component sub-system attached to the subsea base and connected to the turbine which allows for the transformation of raw electrical power from the generator into grid-compatible AC power, as well as transmission of operational and monitoring data in real time. By treating information coming from the multiple sensors located within the turbine system, the TCC is used to optimize the power output in any operating conditions. This is the first time OpenHydro’s pioneering TCC technology has been used anywhere in the world. Its design has been a critical step forward in being able to generate electricity from multiple turbines at sea and export to shore via a single export power cable.

1.0 Introduction Cape Sharp Tidal Venture (CSTV), a joint venture between Emera Inc. and OpenHydro, deployed one, 2 megawatt (MW) instream tidal energy turbine at the Fundy Ocean Research Center for Energy (FORCE) site near Parrsboro, Nova Scotia on November 7, 2016 (the Project). The Project is the beginning of a 4MW, two-turbine demonstration phase (Phase 1) of an ongoing tidal energy initiative in Nova Scotia. The Project was reviewed by Fisheries and Oceans Canada (DFO) through an application made under the Fisheries Act. This Act focusses on conservation and protection of fish habitat essential to sustaining freshwater and marine fish species and prohibits serious harm to fish [subsection 35(1)]. The Project application was also reviewed under the Species at Rick Act (SARA) to determine whether it would adversely impact listed aquatic species at risk and contravene sections 32, 33 and 58 of SARA. It was determined that this demonstration-scale tidal Project would not result in serious harm, as defined under the Fisheries Act, to fish and fish habitat or negative effects to marine mammals and that the Project would not contravene sections 32, 33 or 58 of SARA. DFO therefore issued CSTV a Letter of Advice as opposed to an Approval. In the Letter of Advice, DFO recommended the adoption of an adaptive management approach to monitoring to allow for adjustments and constant improvements to be made to the monitoring program as results become known, and that the results be shared and reviewed on an annual basis with the Fisheries Protection Program (FPP) to ensure that monitoring and management strategies could be modified as appropriate. This Q2 Report is part of that regular update to the FPP by providing a status update on the Project and a follow-up to the Q1 Report. The Q1 Report is available on the Cape Sharp Tidal website: http://capesharptidal.com/eemp/.

2.0 Operational Update In the second quarter of the Project (February 2017 – April 2017), CSTV remained in a commissioning phase. During this time, OpenHydro continued to focus on ramping up the power production of the machine and establishing normal operating parameters through its control system. In March, FORCE began upgrades to the substation at which point power production ceased. Tests and data collection also continued on the monitoring technologies (i.e., sonar, hydrophones). This was continuous from deployment until disconnection of the subsea power cable on April 21, 2017 in order to prepare for a temporary retrieval of the turbine. This temporary retrieval was scheduled in order to perform minor repairs and upgrades to some of the Turbine Control Center (TCC) components, which will take place in Saint John, New Brunswick. The turbine will then be redeployed at the FORCE site, again in Berth D. The retrieval will also provide an opportunity to take monitoring learnings from the first six months of operations and to examine and make adjustments or modifications to the monitoring devices located on the turbine.

3.0 Environmental Effects Monitoring Program

3.1 Context

The Environmental Effects Monitoring Program (EEMP) developed for the Project aims to monitor potential environmental effects in the near field area (i.e., 0-100-metres of the turbine) to better understand potential effects and interactions of specific environmental components (i.e., fish, marine mammals, operational noise) with the OpenHydro Open-Center in-stream tidal turbine. The overall research objective is to verify the accuracy of environmental effect predictions made in the FORCE 2009 Environmental Assessment (EA) report, but the monitoring program will also assist with increasing knowledge about monitoring methods and analysis, development of mitigative measures, and building technical knowledge within the local tidal industry. The FORCE EA Report and subsequent 2010 FORCE EA Addendum are available here: http://fundyforce.ca/environment/enviromental-assesment/. As required by the conditions of the FORCE EA Approval (2009), the CSTV EEMP was developed in collaboration with experts in the field of instream tidal energy and with input from government agencies, including DFO and Nova Scotia Environment (NSE), as well as other instream tidal energy interests including the Offshore Energy Research Association of Nova Scotia (OERA), FORCE, and FORCE’s independent Environmental Monitoring and Advisory Committee (EMAC). The CSTV EEMP forms a component of FORCE’s EEMP commitment under the FORCE Environmental Management Plan. Both EEMPs have been designed to be complimentary in order to achieve the most meaningful examination of potential effects and in consideration of the extensive baseline studies that have taken place at the FORCE Crown Lease Area since 20082. The CSTV EEMP cornerstone is an adaptive management approach to evaluate data and make informed, science-based decisions to adjust technology and monitoring methods, assess mitigation measures and address concerns as necessary. This approach is necessary because of the unknowns and difficulties inherent with gathering data in harsh tidal environments, such as the Minas Passage. It allows for adjustments and constant improvements to be made as knowledge about the system and environmental interactions become better known. The CSTV EEMP is reviewed continuously with regulators and FORCE and modified on the basis of accumulated experience and observed progress toward achieving the monitoring objectives. This adaptive approach will assist with resolving knowledge gaps of the potential effects of the Project and will also facilitate the design and implementation of new or modified monitoring strategies. Additional information is available in the CSTV EEMP document available on the CSTV website: http://capesharptidal.com/eemp/. The FORCE EEMP is available on the FORCE website: http://fundyforce.ca/wp-content/uploads/2012/05/FORCE-EEMP-2016.pdf.

2 The FORCE website provides a list of baseline studies: http://fundyforce.ca/environment/research/

4.0 Q2 Monitoring Update The monitoring program was initiated following deployment of the turbine on November 7, 2016. There were no changes to the operation of the monitoring devices on the turbine since Q1. A summary update is provided for the second quarter of operation (February 2017 – April 2017) in the following sections.

4.1 Monitoring Technology Operations The four passive acoustic monitoring (PAM) (Ocean Sonics icListen hydrophones) sensors and the active acoustic monitoring (AAM) (Tritech Gemini imaging sonar) sensor are co-located on the turbine to create an integrated monitoring system that collects data specific to marine mammals and fish. In addition to these devices, four acoustic doppler current profilers (ADCPs) are mounted on the turbine to provide data on flow, and a video camera was positioned on the subsea base to record the turbine rotor. Data from all the monitoring devices on the turbine is transmitted continuously through a fiber optic data cable contained within the subsea power cable. Data is logged on-shore to hard drives and remotely saved to an OpenHydro server. As noted in Section 2.0, the turbine was disconnected from the subsea power cable in preparation for retrieval so data collection was halted at this time. A contingency program for monitoring was discussed with NSE and DFO and implemented on May 1, 2017. In recognition of the potential need for contingency planning as part of future operations, a new section on contingency monitoring was added to the CSTV EEMP. Turbine operational sound was monitored by a separate deployment of a high flow mooring design of an autonomous multichannel acoustic recorder (AMAR). Data storage is achieved within the instrumentation of this sea bed mounted unit.

4.1.1 icListen Hydrophones There are four hydrophones located on the turbine. The main objective of these devices is to detect harbour porpoise vocalizations to determine the seasonal frequency of this species, and other vocalizing marine mammals (i.e., whales), and to support data results associated with the Gemini imaging sonar on how marine wildlife interacts with the turbine. Using four devices creates an opportunity to compare various locations for devices on the turbine; the potential for integration with other monitoring devices; and to examine how much potential exists for localization of porpoise sounds under various tidal and operational conditions. An additional benefit was the redundancy so that in the event that one or more of the hydrophones is damaged, sufficient data can still be collected from one or two units to meet monitoring requirements. Data collection was consistent in Q2 from two hydrophones. Communications from a third hydrophone has been intermittent and a fourth has not communicated since the turbine was deployed. All hydrophones will be inspected while the turbine is in Saint John for inspection and upgrades (refer to Section 5).

4.1.2 Gemini Imaging Sonar The Tritech Gemini imaging sonar faces the ebb tide3 and monitors an area up to 60 m in front of the turbine that is approximately 104 m in width. Species detection ranges from a lower limit of approximately 10 cm in length and upwards. The purpose of this sonar is to investigate the potential for integration with other monitoring devices and to track marine wildlife approaching the turbine in order to better understand potential interactions. The imaging sonar has operated consistently throughout the second quarter of deployment. An inspection and readjustment to the positioning of the unit is planned after the turbine is retrieved (refer to Section 5).

4.1.3 Video Camera Communication with the SAIS IP-CAM HD Ethernet underwater video camera has been unsuccessful since deployment; therefore no data has yet been logged from this device. The device will be inspected after the turbine is retrieved (refer to Section 5).

4.1.4 ADCPs All four ADCPs have been functioning and providing data on flow regimes within the Minas Passage. This information on current velocities will be used to support analysis for all other devices. The results from these instruments are not reported as part of the EEMP, but rather will supplement the data from the other instrumentation to understand flows during specific time periods of interest.

4.1.5 AMARs The AMAR was deployed approximately 100m from the turbine to perform a Sound Source Characterization (SSC) over a period of 3 months. The SSC measured underwater sound levels from the turbine. The goal of this study was to characterize low frequency (i.e., below 60Hz) sound from the turbine operation and compare it to the ambient (natural) sound created by the environment. A second recorder was deployed as a control unit approximately 2km from the turbine. The AMAR unit deployed closest to the turbine was successfully retrieved in January 2017. The control unit has not yet been recovered due to weather issues and ongoing competing marine operations at the FORCE site.

3 The ebb tide was the direction chosen for the sonar as baseline studies showed a greater likelihood of species detection during this time and less turbulence that may affect detection.

4.2 Q2 Preliminary Results This second quarter has been focussed on identifying areas where the data collection process can be refined and to develop protocols to efficiently manage, store and share that data. Additional work has also been completed on preliminary analysis and data interpretation.

4.2.1 Marine Fish and Marine Mammals As with Q1, the majority of data analysis for fish and marine mammals in Q2 continued to focus on understanding and improving near-field localization, decreasing areas of interference from the environment and from other instrumentation, and refining algorithms in order to achieve data clarity and, in time, to improve the overall data collection and the monitoring program itself. This is required in order to coordinate and integrate the passive and active acoustic devices to fully understand potential near-field interactions with the turbine. Gemini Imaging Sonar Targets identified by the Gemini sonar are classified as one of four possibilities:

i. True positive: a fish or marine mammal; ii. True negative: an object drifting with the tide (e.g. seaweed or driftwood, etc.);

iii. False negative: a fish or marine mammal that is not detected when it ought to be; or iv. False positive: an object, turbulence or noise is erroneously detected as a marine mammal or

fish Data collection, management and raw data analysis from the Gemini sonar unit continued in Q2. This raw data analysis is required to prepare the data for a later biological analysis. Large amounts of data are being received resulting in a lengthy process for running the software and performing validations. This is in part due to the orientation of the imaging unit which has a larger area of the seabed in the view-plane than expected. This has led to increased identifications of false positives. Data collection and analysis will resume after the turbine has been re-deployed. The data management and analysis is expected to be more efficient with an improved sonar orientation. In Q2, Tritech also focussed on software development, in particular improving individual target identification and target tracking ability, as well as synchronizing the imaging sonar with the hydrophones. The work in Q2 specifically involved:

• Application of a number of different image processing techniques to create a new filter to identify targets in the data. This included smoothing and de-noising techniques and identification techniques to evaluate improvements in identifying single targets.

• Implementation of a new algorithm to test with fish data and to compare to existing marine mammal algorithms.

• Looking at ways to change how small targets (potentially fish) are dealt with along with optimization methods to maintain real time operations.

• Work on development of fish measurement techniques which have proven extremely successful for later biological analyses.

Ongoing work will focus on improvements to target tracking abilities and improvements to group classification as well as cross referencing and human validation of automatically detected targets. While the turbine is in Saint John, we will take the opportunity to reposition the Gemini sonar with the objective of reducing the amount of seabed in the sonar view. This will decrease interference from the seabed and is expected to facilitate data analysis by decreasing the numbers of false positives. icListen Hydrophones During the second quarter, data from two hydrophones was continually collected and preliminary processing began. Processing will continue with using automated click detectors. On March 27, 2017, the recording settings on one hydrophone (1404) were changed to record the full bandwidth data to improve detection of harbour porpoise clicks without as much interference from other equipment on the turbine. The changes were done remotely at the FORCE site. The second hydrophone (1405) continues to record data but is receiving background noise from other equipment close to the hydrophone that may mask the frequencies at which harbour porpoises click are found. Analysis of this relationship is ongoing. Early observations of the data found frequent porpoise detections over many days during the Q2 period. Most porpoise clicks were found during slack tide, possibly due to the masking of porpoise clicks during peak tide. The hydrophones have also detected vessel presence. Hydrophone data will continue to be analyzed for comparison among the units and to the Gemini imaging sonar. During the retrieval the team will consider moving hydrophone 1405 to a quieter area on the turbine where there will be less noise from other equipment. Future tasks will also involve troubleshooting the communication issues related to that of the other two non-functioning hydrophones; providing quantitative data for days of harbour porpoise visits; and understanding turbine operation better to have context for additional data processing. 4.2.2 Operational Sound A preliminary analysis of the sound data from the recovered recorder (closest to the turbine) was provided in the Q1 Report. Those results suggested that turbine sound is indistinguishable from flow noise below 60 Hz. In the case of the Minas Passage therefore, this could mean that the sound produced by the turbine is lower than the ambient or natural noise for most of the tidal cycle. CSTV will work with JASCO to continue to try and recover the control unit which is still deployed in the Minas Passage in order to compare data from the two units.

5.0 Mitigation A number of mitigative measures have been identified to adapt the present monitoring program moving forward:

i. The ability to remotely access some of the instrumentation provides the ability to adjust frequency settings and explore options to avoid interference with other instrumentation. This work was ongoing in Q2.

ii. With the opportunity to access the turbine and subsea structure during the planned temporary retrieval, the following activities are planned:

• repositioning of the Gemini sonar; • examination, issue identification, and repair (if needed) of two of the hydrophones

and associated cables and connectors; and • examination of the video camera and associated cables and connectors.

iii. In consultation with DFO and NSE, CSTV is developing an additional component to the monitoring program to address times when the turbine may be deployed but no longer connected to the subsea power cable, and therefore unable to collect data from the monitoring devices. This could occur during deployment or retrieval operations. The contingency program will be collaborative with FORCE.

6.0 Additional Items As noted in the Q1 report, CSTV is looking into additional monitoring components to complement the present monitoring program and to continue to meet the commitments made to stakeholders prior to deployment. This includes:

• Participating in FORCE’s beach walk monitoring program; • Continuing to engage with fishery stakeholders to create improved consultation processes; and • Staggering a second turbine deployment to allow additional opportunity for data monitoring,

collection and analysis. In addition, and in conjunction with FORCE and Open Seas Instrumentation Inc., OpenHydro is continuing work to integrate, test, qualify and deploy a second Gemini imaging sonar on one of the FORCE Fundy Advanced Sensor Technology (FAST) platforms. Wet tests are planned at the FORCE site while the turbine is in Saint John. The sonar will be deployed with the FAST platform and positioned near a turbine, to provide a view plane of the side of the turbine structure. The purpose of this initiative is to explore the potential for monitoring the close vicinity of one side of the turbine rotor. CSTV will also investigate the scope of the next stage of operational noise studies that will focus on measuring turbine operational noise above 60Hz and comparing results to ambient (natural) levels.

Cape Sharp Tidal Venture

Environmental Effects Monitoring Program – Q3 Update Report

October 18, 2017

Table of Contents

1.0 Introduction ...................................................................................................................................... 2

2.0 Operational Update .......................................................................................................................... 3

3.0 Environmental Effects Monitoring Program ..................................................................................... 4

3.1 Context .......................................................................................................................................... 4

3.2 EEMP Objectives ........................................................................................................................... 5

3.2.1 Fish and Marine Mammals .................................................................................................... 5

3.2.2 Operational Sound ................................................................................................................ 6

4.0 Q3 Monitoring Update ...................................................................................................................... 7

4.1 Monitoring Devices ....................................................................................................................... 7

4.1.1 icListen Hydrophones ............................................................................................................ 7

4.1.2 Gemini Imaging Sonar ........................................................................................................... 9

4.1.3 Video Camera ...................................................................................................................... 12

4.1.4 ADCPs .................................................................................................................................. 12

4.1.5 AMAR Units ......................................................................................................................... 13

4.2 Preliminary Results ..................................................................................................................... 14

4.2.1 iCListen Hydrophones ......................................................................................................... 14

4.2.2 Gemini Sonar ....................................................................................................................... 15

4.2.2 AMAR Units ......................................................................................................................... 20

4.3 Data Management ...................................................................................................................... 22

4.4 Contingency Monitoring Program .............................................................................................. 22

5.0 Additional Items .............................................................................................................................. 27

Executive Summary

From November 8, 2016 to April 21, 2017 the Cape Sharp Tidal Venture (CSTV) demonstration instream tidal turbine produced electricity onto the Nova Scotia Power Inc. grid and gathered important monitoring data at the Fundy Ocean Research Center for Energy (FORCE) tidal energy test site. On June 15, 2017 the turbine was successfully and safely retrieved and taken to Saint John, New Brunswick where it has been subject to a detailed evaluation as well as upgrades and repairs to the turbine control center1 (TCC) and monitoring equipment. The CSTV Environmental Effects Monitoring Program (EEMP), initiated in November 2016 upon deployment of the in-stream tidal turbine, aims to test environmental monitoring devices and to monitor and better understand potential environmental effects and interactions of specific environmental components (fish, marine mammals, operational noise) in the near-field area (i.e., 0-100-metres) of the OpenHydro Open-Center Turbine (instream tidal turbine). The overall research objective of the monitoring program is to verify the accuracy of environmental effect predictions made in the 2009 Environmental Assessment (EA) report completed for the test site. The EEMP requires quarterly reports to be submitted to provide regulators with interim updates on the Project, including operational information on the turbine and the monitoring devices. This Q3 report has been submitted to provide a status update for the third quarter of 2017 (May – September 20172). Third-quarter insights include:

• The turbine was disconnected on April 21, 2017 to prepare for on-site retrieval activities. However, early preparatory activities identified that additional work was needed to disentangle a mooring line from around the subsea base. This caused a delay in the retrieval process. The turbine was retrieved on June 15, 2017 during a low tide window. The process took 70 minutes.

• Disconnection from the subsea power cable occurred on April 21, 2017. Therefore no data was collected from equipment on the turbine from April 22 – June 15, 2017. A contingency environmental monitoring program was developed and implemented during this time.

• Turbine retrieval provided an opportunity to recover and inspect the monitoring devices; replace broken sensors where required, upgrade cables, and in keeping with the adaptive management approach of the EEMP, to mitigate issues and concerns from the first deployment to ensure a more robust program moving forward.

• During the retrieval, monitoring data collected during the six-month deployment was continually refined and underwent preliminary analyses. These early results show that porpoises were detected at the site, in the vicinity of the turbine site, during all months from November to April. The month with the most porpoise visits was March. No other marine mammals were detected during the six month deployment. The Gemini data shows clear patterns in the

1 The TCC is an electrical component sub-system attached to the subsea base and connected to the turbine which allows for the transformation of raw electrical power from the generator into grid-compatible AC power, as well as transmission of operational and monitoring data in real time. By treating information coming from the multiple sensors located within the turbine system, the TCC is used to optimize the power output in any operating conditions. This is the first time OpenHydro’s pioneering TCC technology has been used anywhere in the world. Its design has been a critical step forward in being able to generate electricity from multiple turbines at sea and export to shore via a single export power cable. 2 Initially the Project quarterly reports were based on the Project timeline which began in November 2016 with deployment. To align the reporting period with FORCE’s reporting and with the calendar year, the CSTV quarterly reporting dates will be: April 1, July 1, and October 1. An annual report will be submitted on January 1. In order to address the change, this Q3 report covers a reporting period from the end of the Q2 report (May) to the end of the third quarter in the calendar year (September).

Page | 1

automatically-detected activity in the data that correspond to the tidal current and water elevation

• Development of protocols to manage, store and share monitoring data is ongoing with the objective of having a faster and easier method for the next deployment. This will facilitate data access and early analyses.

• A study to examine high frequency sounds of turbine operation was initiated during this reporting period.

• An update to the CSTV website, including the addition of updated frequently-asked questions (FAQs), was also initiated to address/answer a number of specific questions about the monitoring program including how the monitoring devices function and how analyses is being completed at these early stages. These updates are ongoing but are planned for completion in next quarter.

These are still early days in this important research initiative. The information gathered now will build upon and complement the 93 scientific reports and related documents that have been completed to date on Fundy tidal power topics as well as the international body of research on in-stream tidal energy.

Page | 2

1.0 Introduction Cape Sharp Tidal Venture (CSTV), a joint venture between Emera Inc. and OpenHydro, deployed one, 2 megawatt (MW) instream tidal energy turbine at the Fundy Ocean Research Center for Energy (FORCE) site near Parrsboro, Nova Scotia, on November 7, 2016 (the Project). The Project is the beginning of a 4MW, two-turbine demonstration phase (Phase 1) of an ongoing tidal energy initiative in Nova Scotia. The turbine was deployed for a six-month period and was retrieved to address minor repairs and upgrades (refer to Section 2.0). The FORCE site was approved under a joint federal and provincial environmental assessment (EA) in 2009. This EA considered:

• Multiple subsea turbine generators; • Subsea cables connecting the turbines to land-based infrastructure; • An onshore transformer substation, and; • Power lines connecting to the local power distribution system as part of the process.

The documents related to this process are available on the FORCE website at: http://fundyforce.ca/environment/enviromental-assesment/. The approval was in accordance with Section 13(1)(b) of the Environmental Assessment Regulations, pursuant to Part IV of the NS Environment Act and stated that any adverse effects or significant environmental effects could be adequately mitigated through compliance with specific conditions. Turbine developers who were awarded ‘berths’ within the FORCE site are required to meet the relevant conditions of the EA Approval, including the development and implementation of an Environmental Effects Monitoring Program (EEMP), for potential near field effects to specific environmental components. The CSTV Project was further reviewed by the federal department of Fisheries and Oceans Canada (DFO) through an application made under the Fisheries Act prior to deployment. The Fisheries Act focusses on conservation and protection of fish habitat essential to sustaining freshwater and marine fish species and prohibits serious harm to fish [subsection 35(1)]. The Project application was also reviewed under the Species at Risk Act (SARA) to determine whether it would adversely impact listed aquatic species at risk and contravene sections 32, 33 and 58 of SARA. It was determined that this demonstration-scale tidal Project would not result in serious harm, as defined under the Fisheries Act, to fish and fish habitat, or cause negative effects to marine mammals, and that the Project would not contravene sections 32, 33 or 58 of SARA. The Fisheries Protection Program (FPP) at DFO acknowledged the technological and environmental challenges that need to be addressed through adaptive management measures to improve the understanding of interactions between aquatic resources and instream tidal devices. It was acknowledged that the adaptive management approach to environmental monitoring would address the information gaps raised by the department. DFO therefore issued CSTV a Letter of Advice and a set of recommendations to be implemented within the CSTV EEMP. As part of the recommendations, and under the mandate of an adaptive management approach to environmental monitoring, DFO recommended that CSTV submit interim reports on a quarterly basis to provide regulators and stakeholders with an update on the turbine and the monitoring devices; issues or

Page | 3

concerns; any preliminary results that are available; and a discussion of how the objectives of the EEMP are being demonstrated. The scope of the annual monitoring report (Q4), to be submitted on January 1, will include a summary of all data results for the year, up to the end of the third quarter, and will provide additional insight into analysis of the results including general patterns of distribution and movements of marine wildlife in the Minas Passage as well as recommendations moving forward. In addition to the annual and quarterly reports, CSTV meets with representatives of the FPP and NSE on a regular basis to provide monthly updates and to share information regarding ongoing data acquisition, analyses and management, as well as discussions on any data or monitoring issues and how they might be mitigated. This adaptive approach was developed to allow for ongoing review of the EEMP by the FPP and NSE, and adjustments and constant improvements to ensure that monitoring and management strategies could be modified as appropriate. This Q3 Report is part of the regular update to NSE and the FPP for the 2016/2017 deployment and provides:

• A follow-up to the Q1 and Q2 Reports; • A status update on the turbine and the monitoring devices; • A discussion of issues and how they are being addressed and/or mitigated; • Preliminary data results; and, • An update on the CSTV EEMP objectives.

The Q1 report is available here (http://capesharptidal.com/wp-content/uploads/2017/04/20170401-EEMP-Q1-Report-FINAL.pdf) and the Q2 Report is available here (http://capesharptidal.com/wp-content/uploads/2017/07/20170701-EEMP-Q2-Report-FINAL.pdf), on the Cape Sharp Tidal website. Additional information about the EEMP can be found at: http://capesharptidal.com/eemp/.

2.0 Operational Update The focus of operations during this reporting period (May 2017 – September 2017) for the Project has been retrieval of the turbine and subsequent evaluation and inspection of the turbine and all associated monitoring devices. Where required, monitoring equipment has been replaced. As noted in the Q2 report, the first turbine was disconnected on April 21, 2017 to prepare for on-site retrieval activities. However, early preparatory activities identified that additional work was needed to disentangle a mooring line from around the subsea base. This meant that the focus of the operation turned to removal of the line rather than recovery of the turbine. To better understand the location of the mooring line, which is required for holding station during turbine disconnection and retrieval, high resolution sonar and video surveys were undertaken to confirm the extent and exact location of the entanglement. A new mooring was installed at this time. After the line was successfully removed during marine operations that took place over several days during different tidal windows, the turbine was recovered during a 70-minute window on June 15, 2017.

Page | 4

Inspection of the recovered turbine is currently ongoing in Saint John, New Brunswick. The second turbine is also in Saint John and what has been learned from the first deployment (e.g., operations, turbine functioning, monitoring devices etc.) is being integrated into this second turbine in preparation for deployment at the FORCE site, in Berth D. A date for this deployment has not yet been confirmed. OpenHydro will continue to take full advantage of having the recovered turbine in port to carefully evaluate and understand how the first turbine performed while deployed, and to determine what improvements can still be made. The retrieval of the turbine also provided an opportunity for CSTV and the research team to inspect all monitoring devices on the turbine and to implement learnings that were gained during the six-month deployment. All monitoring devices were removed from the turbine for a full inspection and a site visit was completed by members of the research team in August. The purpose of the site visit was to examine the monitoring devices to confirm the condition of the instruments, troubleshoot issues with communications and data acquisition and discuss the future redeployment of the monitoring instruments on the next turbine. Details on these inspections are provided in specific sections in Section 4.0.

3.0 Environmental Effects Monitoring Program

3.1 Context

The EEMP developed for the Project aims to monitor potential environmental effects in the near-field area (i.e., 0-100-metres of the turbine) to better understand potential effects and interactions of specific environmental components with the OpenHydro Open-Center Turbine, an in-stream tidal turbine at the FORCE test site. The environmental components that CSTV was instructed to focus on by regulators included fish, marine mammals, and turbine operational noise. The overall research objective of the CSTV EEMP is to verify the accuracy of environmental effect predictions made in the FORCE 2009 EA which stated that with the implementation of the proposed mitigation measures, including development and implementation of a detailed monitoring plan, adverse residual environmental effects are predicted to be not significant for all valued ecosystem components. The monitoring program will also assist with increasing knowledge about monitoring methods and analysis, development of mitigative measures, and building technical knowledge within the local tidal industry. The FORCE EA Report and subsequent 2010 FORCE EA Addendum are available here: http://fundyforce.ca/environment/enviromental-assesment. As required by the conditions of the FORCE EA Approval (2009), the CSTV EEMP was developed in collaboration with experts in the field of instream tidal energy and with input from government agencies, including DFO and NSE, as well as other instream tidal energy interests including the Offshore Energy Research Association of Nova Scotia (OERA), FORCE, and FORCE’s independent Environmental Monitoring and Advisory Committee (EMAC). The CSTV EEMP forms a component of FORCE’s EEMP commitment under the FORCE Environmental Management Plan. Both EEMPs have been designed to be complementary in order to achieve the most meaningful examination of potential effects, and in consideration of the 93 baseline studies and reports conducted on Fundy tidal power topics since 20083.

3 The FORCE website provides a list of reports and studies completed to date: http://fundyforce.ca/environment/research/

Page | 5

The CSTV EEMP cornerstone is an adaptive management approach to collecting, analyzing and evaluating data and making informed, science-based decisions to adjust technology and monitoring methods, assessing mitigation measures and addressing concerns as necessary. This approach is necessary because this is a new area of research and development and there are difficulties inherent with gathering data in harsh tidal environments, such as the Minas Passage. It allows for adjustments and constant improvements to be made as knowledge about the system and environmental interactions become better known. The CSTV EEMP is reviewed continuously with regulators and FORCE and modified on the base of accumulated experience and observed progress toward achieving the monitoring objectives. This adaptive approach will assist with resolving knowledge gaps of the potential effects of the Project and will also facilitate the design and implementation of new or modified monitoring strategies. CSTV is working with industry-leading local and international experts in marine technology in tidal environments to collect and interpret this data. Quarterly update reports are provided to regulators (NSE and DFO) on April 1, July 1, and October 1 of each year. The annual environmental monitoring report (due January 1, 2018) will summarize data collection and learnings from the 2016/2017 deployment up until the end of September (end of Q3). Additional information is available in the CSTV EEMP document available on the CSTV website: http://capesharptidal.com/eemp/. The FORCE EEMP is available on the FORCE website: http://fundyforce.ca/wp-content/uploads/2012/05/FORCE-EEMP-2016.pdf. 3.2 EEMP Objectives The following table provides a summary the objectives of the CSTV EEMP. EEMP Component

Study Objectives

Fish Determine the seasonal frequency of occurrence of fish within the near-field environment of the turbine using Gemini imaging sonar. Integrate data-sets into a strike risk model for fish. This is a long-term objective.

Marine Mammals

Determine the seasonal frequency of occurrence of harbour porpoise within the near-field environment of the turbines. Determine the relationship between harbour porpoise occurrence and turbine operations. If the numbers of other vocal cetaceans are high enough to use in a data set, results will be used determine the seasonal frequency of these species (e.g., white-sided dolphins). Integrate data-sets into a strike risk model for marine mammals. This is a long term objective.

Operational Sound

Characterize low frequency operational turbine sound to assess the effect of turbine operations on the noise profile of the site. Characterize high frequency operational turbine sound to assess the effect of turbine operations on the noise profile of the site.

3.2.1 Fish and Marine Mammals The study objectives above were developed to achieve the objectives of the EEMP using an adaptive approach and in consideration of ongoing worldwide monitoring developments in the field of tidal energy. The collection and processing of data relating to marine life in tidal energy sites has so far been limited to individual use of either active or passive acoustic sensor technologies. The data collected is

Page | 6

therefore constrained by the limitations of each individual sensor, resulting in the need for further data collection and increased processing time. Investigating the use and integration of active and passive acoustic sensors as a way to obtain real time detection, classification, localization and tracking of fish and marine mammals in high energy sites is key to understanding interactions of marine wildlife with instream tidal technologies. CST’s integrated monitoring system incorporates two different sensors; the Tritech Gemini Imaging Sonar (active acoustic) and the Oceansonics icListen smart hydrophone (passive acoustic). These sensors will be co-located on a common platform (the turbine and subsea base) and a data interface will be created to allow data from each sensor to be combined into an integrated fish and marine mammal data set. Tritech’s Seatec object detection and tracking software will be further developed to interpret this data and to provide real-time automated detection, classification, localization and tracking of fish and marine mammals. The goal is to not only improve how data from each individual sensor is processed and interpreted, but to also develop an interface between the two technologies (i.e., sonar and smart hydrophones) which would combine the strengths of each sensor type and enable the efficient collection of high quality, synergistic data for environmental monitoring at high energy sites. By utilizing an adaptive approach, the objectives allow for:

• Improvements to existing sensor technology software that will maximize individual sensor capability;

• Integration of two complementary sensor technologies to improve ability to detect, classify, localize and track marine mammals and fish in real-time; and

• Testing of the individual sensor capability and integrated system effectiveness in a high energy site.

3.2.2 Operational Sound Increased noise in the marine environment has the potential to directly or indirectly adversely affect marine mammals and fish species. The extent of this issue, in particular the relationship with tidal turbines, is not well understood and data remains sparse. Due to this lack of published data related to noise monitoring of marine turbines during operation and to understand the environment in which the turbines operates, acoustic information is required in order to reliably assess the significance of sound as a risk factor. This information will serve to further evaluate the potential effects of noise in order to determine the nature soundscape, assess acoustic effects, and develop effective mitigation, if required.

Page | 7

4.0 Q3 Monitoring Update

The monitoring program was initiated following deployment of the turbine on November 7, 2016. A summary update is provided for this reporting period in the following sections.

4.1 Monitoring Devices The four passive acoustic monitoring (PAM) (Ocean Sonics icListen hydrophones) sensors and the active acoustic monitoring (AAM) (Tritech Gemini imaging sonar) sensor are co-located on the turbine to create an integrated monitoring system that collects data specific to marine mammals and fish. In addition to these devices, four acoustic Doppler current profilers (ADCPs) are mounted on the turbine to provide data on flow. A video camera was positioned on the subsea base to record the turbine rotor to try and understand the passage of fish immediately at the rotor location. Data from all the monitoring devices on the turbine is transmitted continuously through a fibre optic data cable contained within the subsea power cable. Data is logged on-shore to hard drives and remotely saved to an OpenHydro server. As noted in Section 2.0, the turbine was disconnected from the subsea power cable, on April 21, 2017, in preparation for retrieval so data collection was halted at this time. A contingency program for monitoring was discussed with NSE and DFO and implemented in May 2017 (refer to Section 4.4). In recognition of the potential need for contingency planning as part of future operations, a new section on contingency monitoring was added to the CSTV EEMP. All environmental monitoring devices have been inspected to assess for damage, biofouling and/or issues with cables. All pins and cables were also inspected for corrosion and abrasion. Turbine operational sound was monitored by a separate deployment of a high flow mooring design of an autonomous multichannel acoustic recorder (AMAR). Data storage is achieved within the instrumentation of this sea bed mounted unit. The following sections provide an update on the performance of the various monitoring devices used to inform the CSTV EEMP, as well as issues/concerns and related mitigation. Preliminary data results are provided in Section 4.2.

4.1.1 icListen Hydrophones There are four hydrophones located on the turbine. The main objective of these devices is to detect harbour porpoise vocalizations to determine the seasonal frequency of this species, and other vocalizing marine mammals (i.e., whales), and to support data results associated with the Gemini imaging sonar on how marine wildlife interacts with the turbine. A hydrophone is an underwater microphone, also referred to as a passive acoustic monitoring device, which converts sound energy to electrical energy. The icListen Smart Hydrophone, by Ocean Sonics, is a digital hydrophone which eliminates the need to add amplifiers and other electronics to the instrument to collect underwater sound. Data is processed and can be streamed to a receiver onshore and/or

Page | 8

recorded within the instrument. Triggers, such as harbour porpoise clicks and whistles, are detected as specific acoustic events (sound events). Other sounds in the environment are also recorded to provide context and increase understanding of the overall sound environment. Using four devices for the Project creates an opportunity to compare various locations for devices on the turbine; the potential for integration with other monitoring devices; and to examine how much potential exists for localization of porpoise sounds under various tidal and operational conditions. An additional benefit was the redundancy of hydrophones so that in the event that one or more of the hydrophones is damaged, sufficient data can still be collected from one or two units to meet the monitoring requirements. Update As noted in the Q2 report, data collection was consistent from two hydrophones during the six-month deployment. Communication from a third hydrophone was intermittent and a fourth, located on the top of the rotor, did not communicate shortly after deployment. All four hydrophones were inspected, synchronized and tested in Saint John following recovery of the turbine. The intermittent hydrophone, located on the fore (front) section of the subsea base, was found to have a damaged cable. The hydrophone located on top of the rotor was sheared off. Data was recovered and archived and a final inspection will take place prior to deployment. Testing of the hydrophones is planned to be conducted in Saint John Harbour and again in the Minas Passage just prior to deployment. Issues and Mitigation During the first deployment and through inspections of the retrieved turbine, there were a few problems identified. Although the device on the top of the rotor provides the best position for uninterrupted data collection, it is also the location where the device is the most exposed to currents and any debris that may be moving with the tide. Ocean Sonics has provided suggestions for protecting the device but while maintaining the listening capacity of the instrument. The four hydrophones on the turbine were meant to be synchronized to allow for the option of localizing sound sources that were detected (i.e., determine the position of a vocalizing animal in relation to the turbine). During the recent deployment however, only one usable channel was acquired so localization couldn’t be performed. This was due to communication issues that caused problems in recording data. Third octave processing was performed for different episodes in the data and a visual inspection of the data was also performed, including screenshots of spectral data. Harbour Porpoise click detectors were run on three different software programs: Lucy; Coda; and PAMGuard; however data recorded before March 24, 2017 was processed only with Lucy because the time series data was not set to sample at full bandwidth, which is needed for the other two software programs. Some of these problems were addressed during the deployment and additional mitigation has been implemented for the next deployment as follows:

Page | 9

• The ability to remotely access hydrophones during the last deployment provided an opportunity to adjust frequency settings and explore options to avoid interference with other instrumentation.

• Iron bar guards will be added to the hydrophone mounting structure located on the top of the turbine rotor to provide increased protection to this unit.

• Wider barriers around the other hydrophones will be added for increased protection of the other three devices.

• The connector was replaced for the hydrophone on the fore section of the subsea base. • While the turbine has been in port, the cables for all four hydrophones have been replaced to

remedy any future communication problems. During the inspections in Saint John, all hydrophones were serviced appropriately for the next deployment and the settings were adjusted to ensure synchronization. All hydrophones were set to synchronize to an external time source, pulse per second (PPS) (a signal that synchronizes the hydrophones).

• The issue around the time series data not sampling at full bandwidth, which is needed for the other two software programs, was corrected during the deployment and all three programs were used on the last data set (March-April) for reference and reliability. This issue will be improved in the future by providing set-up guidelines for the hydrophones to ensure proper sampling frequencies. This information has been provided to OpenHydro.One issue was identified during the dry testing related to a single intermittent cable connection. OpenHydro has addressed the cable problems and will perform a final check prior to deployment. Set-up guidelines have been provided by Ocean Sonics for the hydrophones to ensure proper sampling frequencies.

• Overall, data transfer and management processes are under discussion to see how the process can be improved for the next deployment (refer to Section 4.3). There is a possibility of installing and running the software remotely at data archive site. Remote access to the hydrophones to check status and settings will also be implemented during the next deployment.

4.1.2 Gemini Imaging Sonar The Tritech Gemini imaging sonar is mounted on the subsea base and faces the ebb tide4. The device monitors an area up to 60 m in front of the turbine that is up to approximately 104 m in width. Species detection ranges from a lower limit of approximately 10 cm in length and upwards. The purpose of this sonar is to investigate the potential for integration with other monitoring devices and to track marine wildlife approaching the turbine in order to better understand potential interactions. The Gemini imaging sonar is an active acoustic device. It is a high frequency multi-beam sonar technology that uses reflected sound (similar to an echo) to build up a picture of an underwater environment. Images created by these high frequency sonars are low resolution when compared with contemporary video technologies; however when there is insufficient light or high turbidity [cloudiness or haziness of water caused by suspended solids (sand)] video cameras lose the ability to create a clear image. Subsea environments have limited light and the Minas Passage is very turbid, but these factors are not as much of a problem for high-frequency sonars.

4 The ebb tide was the direction chosen for the sonar as baseline studies showed a greater likelihood of species detection during this time and less turbulence that may affect detection.

Page | 10

A multi-beam, high frequency sonar typically sends out a ‘ping’ (an acoustic pulse transmission) up to 30 times per second. Each ping is used to create an image that visualizes the different intensities of the reflected sound as it bounces back. Light and turbidity therefore do not affect the sonar’s ability to detect objects. The Gemini sonar performs detection based on reflection of sound from objects in the water and then uses a ‘geometric approach’, meaning it focuses on size, shape and movements of each object. A specialized software program, SeaTec, examines the sonar data and extracts moving targets before performing a classification (i.e., concluding that an object is likely to be a fish or other marine wildlife or seaweed or moving rocks). These early stages of working with the Gemini data from the Minas Passage are aimed at developing an understanding of marine life around the turbine by comparing marine life detections by the SeaTec automated algorithms to those by a human observer to ensure that the software is making the correct determinations. The best way to verify automatic target detections is human observation of the sonar data. Automatic algorithms can be used to isolate a subset of data that contains interesting results (i.e., movements of schools of fish or individual species) for human observers experienced with sonar data to study in more detail to address the EEMP objectives. Furthermore, human analysis of the sonar data without automatic tracking provides a reference for the automatic algorithm performance. The automatic algorithms are used to identify subsets of data that may contain “interesting” targets (possible marine wildlife), which can then be used to focus the attention of human observers on times where marine life may be present. This human validation process involves viewing five minute files twice, and focusing on different ranges. Comparing human observations to the automated detections then aids in distinguishing true detections of marine life from false ones (e.g., debris moving with the current). Although low frequency sonars have a greater range (distance), this project utilizes the Gemini high-frequency sonar to achieve better resolution (i.e., clearer images).

Update The imaging sonar operated consistently throughout the six-month deployment and was found to be in good shape following inspection in Saint John post-turbine retrieval. The sonar was visually inspected and has undergone multiple tests in a tank at the port facilities. A final inspection will take place prior to deployment and testing is planned to be conducted in Saint John Harbour and again in the Minas Passage immediately prior to deployment. Issues and Mitigation During the first deployment, there were three issues identified with the Gemini sonar on the CST turbine.

1. Sonar orientation: Too much of the sea bed, rather than the mid-water column in front of the centre of the turbine, was in the sonar’s field of vision. This resulted in the detection of a high number of false positives5 by the software. These false positives are due to marine flora (seaweed), moving rocks, water turbulence etc., created by the currents stirring up the sea

5 False positives are targets that incorrectly trigger alert on the automatic target tracking that identifies the target as swimming marine life. False positives are most commonly due to marine flora, rocks, water turbulence, noise, etc. False positives in automatic detection are more common when the sonar field of view includes the sea bed and where a high volume of noise is present in the data.

Page | 11

bottom during ebb flows. The reason for a high number of false positives is because the target detection algorithms in SeaTec examine sonar data for patterns of reflection indicating marine life. The algorithms work best when there is a contrast between a target and the background, therefore mid-water column views present the ideal environment for automatic identification and tracking. Conversely, the sea bed and surface of the water are good reflectors of sound and therefore create high intensity returns that can interfere with target identification and tracking. This is because there is considerably less contrast between a swimming target and background when the sea bed reflects the sonar signal. Furthermore, target tracking is more difficult because the seabed is a characteristically high activity area due to marine flora, fauna, rocks, water turbulence, noise, etc. Movement may be due to marine life, seaweed, or changing sea bed topography so false positives in automatic detection are more common when the sonar field of view includes the sea bed.

Although the detection of some false positives is normal and can be expected in the water column, the number of false positives in the data set was extensive due to the fact that the sonar was orientated towards the sea bottom. In some of the worst cases false positives were detected in this area at the rate of 600-700 per hour, most of which coincide with the change in water elevation, at low and high tide, when the movement of the water is stirring up the sea bottom and moving objects, such as rocks, around. Due to the high numbers of false positive, human validation of these files took much longer than expected .

2. Continuity of the data transmission: The power supply created interference with sonar communications between the Gemini and the shore-based computer and resulted in data transmission that was not continuous and timely and resulted in very small data files being created and jumps in time between adjacent frames. This was found to be related to the communications between the sonar and PLC due to cabling. The transmission affected the analyses because identification of presence and movement of marine life is hard to discern for the automatic SeaTec algorithms and for humans. When the sonar communication is operating properly, data files contain a consistent number of frames per second and have continuous time stamps. Each file is approximately 307MB.

3. Lengthy and time consuming preliminary data analysis: These two issues above, created a third

issue with the preliminary data analysis by creating a lengthy and time consuming process. Human identification of targets is also more challenging due to the number of moving objects created by view plane dominated by the sea floor and through the creation of many, small files.

These issues have all been taken into consideration for the next deployment and mitigation measures have been developed as follows:

• Human observation of the data and validation of the automatic tracking results has been performed to mitigate some of the problems related to the mounting angle and the communications interruptions. Researchers were able to analyze some files covering an area of approximately 0m to 12m in front of the turbine where the sonar was monitoring the water column. Clarity in this region was much better and marine wildlife was noted during the deployment (refer to Section 4.2 for preliminary results).

Page | 12

• In areas where false positives were numerous and affecting the ability to acquire a preliminary analysis, an exclusion zone was created around the problematic area to inform the algorithms not to perform target identification and tracking there.

• Although the sonar will remain in the same location on the subsea base, the device will be repositioned to capture a more complete view of the mid-water column and less of, or none, of the sea bottom. This will result in a greater clarity of the files and less false positives which will facilitate the human validation component of the data analysis and decrease the overall time required to obtain results. False positive reduction will instead focus on turbulence and better classification of moving targets.

• The issue with the power supply was rectified by changing the cable assembly set-up to a shielded cable suitable for subsea ethernet communication. This will allow more consistent signal strength for data capturing will give greater confidence in the automatic results.

• The amount of data generated by a sonar operating 24 hours a day, every day, continues to create data management concerns related to data transfer time, analysis and storage. As noted above, the lengthy time required for analysis should be remediated with a repositioning of the device and different cables. The time to transfer data and development of an adequate storage system remains ongoing. CSTV is currently working with information technology experts to improve this process for the next deployment (refer to Section 4.3).

4.1.3 Video Camera Communication with the SAIS IP-CAM HD Ethernet underwater video camera was unsuccessful since deployment; therefore no data was logged from this device. Update Upon inspection, it was confirmed that the video camera had been damaged, likely early in the deployment. No video footage from this device is available. Issues and Mitigation An inspection of the unit and the mounting brackets suggests that a different mounting system should provide better protection for this device. A new camera unit will be installed with a different bracketing system for the mount for the next deployment. 4.1.4 ADCPs All four ADCPs have been functioning and providing data on flow regimes within the Minas Passage. This information on current velocities will be used to support analysis for all other devices. The results from these instruments are not reported as part of the EEMP, but rather will supplement the data from the other instrumentation to understand flows during specific time periods of interest.

Page | 13

4.1.5 AMAR Units AMAR units are bottom-mounted hydrophone units and were chosen as the method to measure turbine operational sound. This particular AMAR unit, designed by JASCO, allows researchers to avoid the difficulties experienced with drifting hydrophones that, although display reliability in high flow environments, will only provide a snapshot of the ambient noise and turbine sound as the instrument passes the turbine. A bottom-mounted hydrophone is also favourable in high flow environments as the instrument can reduce flow noise due to water pressure since flow speeds at the bottom of water columns are generally lower than elsewhere in the column. This flow induced ‘noise’ can be further minimized by shielding the hydrophone. JASCO incorporates both considerations into their design, resulting in a streamlined design called the High-Flow (HF) Mooring. Two AMARs were deployed to perform a Sound Source Characterization (SSC) over a period of 3 months while the turbine was deployed (November - January). The SSC measures underwater sound levels from the turbine during various tidal regimes and operating modes. One AMAR was deployed approximately 100m from the turbine and a second AMAR, a control unit, was deployed approximately 680 m away from the turbine. The purpose for the control unit was to be able to compare the sound data from the turbine site to the ambient or natural sound created by the environment, at a location without a turbine. Update The AMAR unit closest to the turbine was successfully retrieved in January 2017. The control AMAR was not recovered (see below). A characterization of low frequency (i.e., below 60Hz) sound from the turbine during operation was completed to compare to the ambient (natural) sound created by the environment. Results were provided in the Q1 report and are summarized in Section 4.2. A second analysis for high frequency sound has been in Q3.

Issues and Mitigation The second AMAR has not yet been recovered. Numerous attempts have been made to grapple the unit however the presence of boulders has made this method of retrieval very difficult. Additional retrieval opportunities continue to be investigated. To mitigate the delay in accessing the data from the control unit, CSTV is looking into using baseline data collected at the FORCE site prior to turbine deployment to provide the comparison to the results obtained from the AMAR positioned close to the turbine. The FORCE baseline data is appropriate as it provides information on ambient sound at the study site before any turbine was deployed and can be used to compare low and high frequency sound.

Page | 14

4.2 Preliminary Results The following sections provide information on the preliminary data analysis and results for the first six-month deployment from November 2016-April 2017.

4.2.1 iCListen Hydrophones All data sets from each hydrophone are collected and data processing is performed for different time periods. This includes a visual inspection of the data, including screenshots of spectral data, and then harbour porpoise click detectors are run on three different software programs. The data is then compared to operational data of the turbine. During the six-month deployment the hydrophones recorded 200 kHz of processed spectral data and sent this data, and 200 kHz time series data, to an onshore computer. In addition to sounds from the environment, the hydrophones are also able to detect variations in turbine operation. A comparison of the hydrophone data to various stages of turbine operations will be completed for the annual report to provide context and increase understanding of the overall sound environment. The preliminary analysis from the hydrophones shows that porpoises were detected on the following days during the recent deployment between November and April:

Deployment Month Dates of Porpoise Detection November 10, 11 December 23, 24, 25, 29 January 1, 4, 8, 29 February 2, 5, 23, 24, 26 March 1, 2, 3, 5, 10, 12, 13, 20, 21, 23, 25, 26, 28, 30, 31 April 1, 2, 6, 7, 9, 10, 12

Porpoises were in the vicinity of the turbine site once or twice on the days stated above. As observed above, the month with the most porpoise visits was March with 15 days. No other marine mammals were detected during the six month deployment. Peak sound levels were detected using third octave processing. This is shown in Figure 1 for the tidal cycle on April 3, 2017.

State of Tide SPL at 1000 Hz SPL at 130000 Hz Slack Tide 79 dB re μPa 94 dB re μPa Flood Tide 135 dB re μPa 144 dB re μPa

Page | 15

Figure 1. Third Octave Graph showing the difference in sound intensity between slack tide and flood tide.

The increased noise levels of around 130 kHz is within the typical range of a porpoise click, which can cause some difficulty in detecting clicks. Noises of 130 kHz can come from multiple sound sources and during the tidal cycle the overall sound increases from the tide. There is also a possibility of noise from equipment on the tidal turbine to cause sound within this range. Any noise near 130 kHz can potentially cause a false positive identification of a porpoise click by a click detecting program. Porpoise click detectors must be able to differentiate between the sound in that bandwidth (other noises at the same frequency) and a porpoise click. Work continues to further develop a click detector program which has been successful in decreasing the number of false positives and false negatives for porpoise clicks that are reported. Another click detector software, PAMGuard, is still being refined for this data to minimize the amounts of false positives and false negatives and will continue into the next deployment. Additional work is also being done to compare data from the turbine operations to sound recorded by the hydrophones. This work will also continue into the next deployment.

4.2.2 Gemini Sonar Data analysis was performed by the automatic SeaTec algorithms and through human observation of sonar imagery (without target tracking enabled) to detect marine wildlife. In Q1, a subset of the Gemini data from November 17-30 was selected for closer analysis because these days are when the maximum and minimum number of detections in the water column had occurred for the quarter (Table 1). The total number of detections during this time period is 2613 but the number is high due to the false positives from the seabed.

40

60

80

100

120

140

160

20 200 2000 20000 200000

Inte

nsity

(dB

re u

Pa)

1/3 Octave Band (Hz)

Slack Tide

Flood Tide

Page | 16

Table 1. Summary of number of detections by Gemini Sonar from November 17-30, 2016*.

Date # of Detections During Water Level Elevation Change # of Detections during other times Total

Detections 17/11/2016 647 19 666 18/11/2016 243 11 254 19/11/2016 52 7 59 20/11/2016 22 4 26 21/11/2016 12 6 18 22/11/2016 59 6 65 23/11/2016 75 8 83 24/11/2016 84 11 95 25/11/2016 139 5 144 26/11/2016 324 7 331 27/11/2016 165 6 171 28/11/2016 132 7 139 29/11/2016 143 12 155 30/11/2016 394 13 407

*The dates 17th and 21st of November were chosen because they contain the most and least number of overall detections, respectively.

Further analysis of these results focused on two particular time periods; both chosen in the context of the number of detections and in reference to the changing water elevation. The first period is from November 17 from 06:00 to 14:00; it contains a high level of detections and covers a period that starts immediately after a tide elevation change, encompassing the following change cycle, and finishes just before a second change. The second period is November 21 for 24 hours; the results from this day contained the fewest number of detections consistently across the whole period. There are a number of false positives that occur during this time, almost all of which are located around the same turbulence caused by seabed areas around the times of sea level change (Table 2). Marine life has been identified in a number of these time periods for both days but classification as to species or indeed distinction between fish or marine mammal is ongoing.

Table 2. Summary of hourly segments for November 17 and 21, 2016. Date Time Detections Analyzed False Positives Potential Marine Life November 17, 2016 06:00 7 1 5

07:00 10 7 3 08:00 40 39 1 09:00 7 7 0 10:00 2 1 1 11:00 1 0 1 12:00 1 0 1 13:00 2 0 2 totals 70 55 14

November 21, 2016 00:00 1 0 1

Page | 17

Date Time Detections Analyzed False Positives Potential Marine Life

02:00 1 0 1 04:00 2 0 2 05:00 2 0 2 06:00 3 3 0 07:00 0 0 0 11:00 1 0 1 13:00 1 0 1 18:00 1 1 0 19:00 4 4 0 20:00 1 1 0 23:00 1 1 0 totals 18 10 8

Preliminary analysis in Q2 focussed on software development, in particular improving individual target identification and target tracking ability, as well as synchronizing the imaging sonar with the hydrophones. The work in Q2 specifically involved:

• Application of a number of different image processing techniques to create a new filter to identify targets in the data. This included smoothing and de-noising techniques and identification techniques to evaluate improvements in identifying single targets.

• Implementation of a new algorithm to test with fish data and to compare to existing marine mammal algorithms.

• Looking at ways to change how small targets (potentially fish) are dealt with along with optimization methods to maintain real time operations.

• Work on development of fish measurement techniques which have proven extremely successful for later biological analyses.

In Q3, an in-depth analysis of automatic target tracking was performed on days with both high and low levels of detections. There are clear patterns in the automatically-detected activity in the data that correspond to the tidal current and water elevation. The tidal current is near a minimum at times where water elevation has been detected to change sign (measured from an average); this identifies low tide and high tide. It is during these times that the highest number of automatic detections were recorded, and also the most false positives. Many of the targets are not moving in a predictable pattern and there is less consensus in velocity direction and speed. Approximately three hours later, when the tidal current is at the maximum and the elevation is changing more rapidly, there are fewer detections in general. Automated analysis was performed on subsets of the larger data set with the considerations of the sonar view and false positives in mind. Time periods for human observation were selected to include time periods with the highest and lowest number of automated detections. Time periods cover tidal current local maximum and minimum (i.e., at least 3 hours on either side of either a slack or flood tide). Manual processing was performed on a subset of data to identify possible targets (Appendix 1). Gemini data in which the automated tracking had detected targets were viewed without the tracking enabled (Figure 2).

Page | 18

Figure 2. Example of the larger view of the Gemini video, with the 1.5x magnification window used for viewing the first 15m shown at the right. Note: The first five meters, shown in white brackets, is water column. The blue brackets depict the sea floor. This allowed a human observer to identify moving objects which could be marine animals and record the detection time, location over time, and approximate size of the target. If a target could not be clearly identified (e.g., if there were increased background noise, or a broken path), the uncertain target positions were classified as “possible”6, “potential” or “probable” marine life.

Possible fish aggregations (Figure 3) were identified as a single unit, with a singular time stamp and range estimate of the general range of the fish aggregation. Any unusual properties of a target (e.g., a very bright and/or wide track) or any particular movements (e.g., movement relative to the turbine) were also noted.

6 ‘Possible’ targets)(depicted with green boxes on the screen) are the correct size and shape for marine life; ‘Potential' (depicted with an amber box) denotes upgraded ‘Possible’ targets that also have a path that suggests the object is not moving with tidal drift. ‘Probable’ (depicted with a red box) are targets that are upgraded from ‘Potential’ when they have a high probability of being marine life. Using this scheme also allows the software to eliminate a large number of false targets such as marine debris moving passively with the tide.

Page | 19

Figure 3. Example of a possible fish aggregation, marked by a box. Once the subset of Gemini data in which the automated tracking had identified targets was reviewed by a human observer, automated detections were compared to those detected manually and possible matches were identified. A match was determined using the time of automatic detection and the range, which were then compared to the manual notes. Only 2 of the 135 targets identified through the automated tracking were matched to the manually detected targets in the videos reviewed to date. As noted above, the automatic SeaTec algorithms had difficulty in detecting and tracking targets due to the interruptions in communications and also with respect to the orientation of the device. The algorithms require a minimum of 5 frames per second to identify targets using tracks built up from sequential images. If there are gaps in the data, the algorithms have difficulty in connecting steps and creating tracks over time. Furthermore, the algorithms are designed to operate in in mid-water column with contrast between high intensity targets and low intensity background. The orientation of the sonar resulted in ensonifcation7 of the sea bed. This means the sea bed, rather than the water column, received the sound waves resulting in a view of a cluttered and high intensity background with a large number of returns and backscatter from movement of water and flora, rocks etc. Fixing the issue that resulted in the interruption to the communications (refer to Section 4.1.2) will solve the problem with connecting target’s tracks over time. The new orientation of the sonar will also address this problem as

7 Ensonifcation refers to an area (or an object when using an ultrasound) to be flooded with carefully-controlled sound waves. In this case, the orientation of the sonar caused the seabed, rather than the water column, to be the receiver of the sound waves or ‘pings’ from the sonar.

Page | 20

it will allow the algorithms to work as intended. SeaTec has been used successfully on many real-time installations when connectivity is consistent and orientation is towards the mid-water column. Details of the manual processing reported as part of the Q3 results are provided as a separate report in Appendix 1. 4.2.2 AMAR Units The recovered AMAR captured sounds from the turbine, mechanical vibrations, and ambient noise. The recordings provide an assessment of the variability in the sound levels from the turbine over its daily operating cycle and in different current and sea states. Sound is most commonly described using the sound pressure level (SPL) metric. Underwater sound amplitude levels are commonly measured in decibels (dB) relative to a fixed reference pressure of p0 = 1 μPa. The root-mean-square (rms) SPL is used to quantify the sounds generated by the turbine. Preliminary results from the recovered unit, from near the turbine, show that low frequency noise dominated during the entire study period (18 Nov 2016 to 19 Jan 2017) and that the differences over a short time indicate that large variations were due to changes in tide state. To compare tidal state noise and turbine operation for a specific date, band level and spectrogram plots were generated for November 18 to November 20, 2016 (Figure 4). Each tidal cycle can be seen clearly by large increases in the low-frequency noise due to flow and noise created as rocks and sediment are shifted by the extreme currents. Horizontal lines in the spectrogram, from 60–300 Hz are tones, or sound, believed to be associated with turbine operation. These tones increased and decreased slightly throughout each tidal cycle, likely associated with changes in the turbine rotation rate as the current increased and decreased from high tide to slack tide, respectively. The tones were distinctly different from the more broadband, atonal flow noise visible in the spectrogram as deep red up to 60 Hz. If turbine noise exists below 60 Hz, it is difficult to distinguish it from the flow noise. For comparison, the noise from vessel Nova Endeavour (a vessel commonly used for work in the FORCE site) at the beginning of this time period is apparent as spikes in the 10–100 Hz and the 100–1000 Hz bands (these spikes are absent in subsequent tidal cycles).

Page | 21

Figure 4. In-band sound pressure levels (top) and spectrogram (bottom) for 18–20 Nov 2016. These preliminary results suggest that turbine sound is indistinguishable from flow noise below 60 Hz. In the case of the Minas Passage therefore, this could mean that the sound produced by the turbine is lower than the ambient or natural noise for most of the tidal cycle. In the absence of ambient (natural) noise levels and to provide a comparison of turbine sound at high flows to other sound contributors, it is possible to estimate the source level of the turbine sound by back-propagating the measured sound levels from 60–300 Hz. JASCO performed this analysis and for reference, the highest source levels were compared to the source levels of a 65 m long survey vessel (the Setouchi Surveyor) transiting at full speed. Results showed that the highest source levels from the operating turbine would only be reached in the loudest 5% of conditions, and that these source levels would still not be as loud as a transiting survey vessel. This indicates that in times of increased flow noise (i.e., from 60 Hz to 300 Hz) sound likely to be associated with turbine operation during these higher velocities can be detected but that turbine sound is still creating less noise in that frequency range when compared to other sound contributors in the area. These preliminary results are promising for understanding the potential effects of turbine sound in the Minas Passage and for understanding the sound profile of the turbine during various flow regimes. However, additional analysis will need to be completed to compare ambient noise and with flow data from the ADCPs mounted on the turbine to provide a final characterization of operational sound, understand how it changes with flow speed and how levels compare to those of levels of natural noise and other sound contributors specific to the Minas Passage.

Page | 22

4.3 Data Management

A plan for data management, including improved data transfer, data analysis, and data storage for the EEMP, is currently under development as CST continues to better understand the amounts of data created from the project and options for improving the data transfer and analysis processes in the next deployment. To date, data generated by the monitoring devices8 on the turbine [i.e., Gemini sonar, iCListen hydrophones (4)] were saved to two external hard drives at the FORCE site; an 8TB drive for hydrophone data and a 12TB drive for the Gemini data. Both drives were replaced every four weeks and data was uploaded to an ftp site and to a cloud storage site managed by OpenHydro. The drives had hardware level data encryption. Third party access to the data was accessed via the ftp site. The protocol that was developed and implemented for this process was documented in an internal Method Statement. For the Gemini sonar, raw data was processed by Tritech Ltd., using the SeaTec software. Files were then transferred to Acadia University where a further preliminary analysis took place involving human validation. Ocean Sonics completed all the preliminary analyses for the hydrophones. Due to the large amount of data generated by the monitoring devices, the main issue with the process was the time required to download the data (from the hard drives and from the ftp site) and to upload date to the ftp site. This ‘bottle neck’ created by the different stages of data transfer has been identified as one of the most important issues affecting the duration of the data analysis process and, in turn, the reporting process (refer to Sections 4.1.1 and 4.1.2). In recognition of the need for improving this process, CST has engaged IT specialists to assess the process and provide potential solutions that can be implemented for the next deployment to create a faster process for getting data to researchers for preliminary data analysis so that frequency of reporting can be increased. This will involve investigating access at the FORCE site and how this may be facilitated, exploring other possible integrations, or sending data to a secure web-based site for access by researchers. 4.4 Contingency Monitoring Program The purpose of this contingency monitoring program9 was to ensure that original commitments under the EEMP were met during the time period when the subsea cable was disconnected (during retrieval or deployment), thereby disabling the monitoring devices, but when the turbine remained deployed. Discussions between CSTV, FORCE, NSE and DFO were undertaken to identify potential contingency options and to provide regulators with immediate short and long-term plans for contingency monitoring. Components included visual surveys by CSTV and FORCE, focused on areas near the FORCE test site and high priority beaches, as well as marine mammal monitoring activities (CPOD studies) that were currently ongoing at the FORCE test site through FORCE’s EEMP10. 8 The video camera was damaged and did not communicate any data for the deployment from November-April. 9 Although the present version of the CSTV EEMP includes a contingency plan around damage to monitoring instrumentation, this contingency was specific to the disconnection of the subsea cable and therefore the data cable which transmits the monitoring data. 10 Results of FORCE surveys are provided in FORCE reports.

Page | 23

During the retrieval operations, CSTV implemented shoreline visual surveys of areas around the FORCE test site, including high priority beaches. The objective of these surveys was to observe and report any marine life behavior including unusual occurrences (i.e., strandings, mammal in distress, mortalities of any marine wildlife, or unusual concentrations of seabirds). Vessel surveys were also incorporated when observers could be safely accommodated on vessels. A total of 57 hours and 53 minutes of observer effort were spent between shoreline, stationary and vessel-based marine observation surveys took place. This comprised 10 shoreline surveys, 11 stationary surveys, and 3 vessel-based surveys. GPS tracklines were kept for both shoreline and vessel-based observation periods to generate coverage maps. There were 248 sightings of marine wildlife over the three survey types; 1 marine mammal was observed: a single harbour porpoise (Phocena phocena) passing between Black Rock and the FORCE beach; and a total of 241 sightings of marine birds (920 individuals) comprised of 17 different species were observed. Most behaviours for marine birds included sitting on rocks, swimming, diving around Black Rock and along the shoreline, and attending to passing fishing boats. There was no surface feeding or diving behaviour noted in proximity to the CSTV berth (i.e., the turbine location) during any survey. Several unknown small fish were observed jumping at the surface while travelling, on a vessel, around Cape Sharp. No indications of any mortalities of marine wildlife or unusual behaviour associated with the turbine operation were observed. An additional component of the surveys, during the retrieval operations, involved discussions with local beach walkers and other Parrsboro residents on an opportunistic basis regarding the marine observation program, and their recent experiences along the coastline and on the water. No strandings or abnormal animal behaviour was reported. In order to recognize this important consideration for the CSTV EEMP, a new section on contingency planning was drafted to be included in the CSTV EEMP document. An updated version of the EEMP with the new contingency planning section will be posted to the CSTV website prior to the next deployment. The contingency program addresses planning during the following unexpected events:

• Damage or loss of environmental devices (detectable through remote monitoring); • Short-term gaps in monitoring caused by preparatory activities related to deployment and

retrieval operations; and • Longer-term gaps in monitoring caused by delays in deployment and retrieval operations.

Table 3 provides a summary of actions to consider during unexpected events. Discussions are currently taking place with regulators and FORCE to explore all options and to reduce timelines for implementation. A finalized program will be developed prior to the next deployment.

Page | 24

Table 3. Contingency Table for Monitoring Devices EEMP Component

Method Summary

Potential Risk Scenario

Proposed Options for Contingency

Considerations for Proposed Contingency Options

Implementation Times for Proposed Contingency Options11

Marine Mammals/Fish

Passive and active acoustic sonars mounted on the turbine.

• Sensor malfunction

• Loss of instrument12

• Disconnection of monitoring devices during deployment or retrieval operations

• Vessel and shore-based observation program for marine mammals.

CSTV would apply a similar methodology as the FORCE marine mammal baseline study, but with modifications to suit near-field observations at Berth D.

Confirmation of survey scope with FORCE, EMAC, fishers, DFO and other stakeholders: 2 weeks Discussions with consultant: 1 month to confirm scope and costs. Vessel acquisition: There are a few vessels that could do this work. Expect 2 weeks to confirm.

• Deployment of C-POD units within Berth D for marine mammals.

CSTV would apply a similar methodology to the FORCE CPOD study. Locations would be closer to Berth D and include 2-3 units.

Confirmation of survey scope and unit locations with FORCE, EMAC, fishers, DFO and other stakeholders: 2 weeks Time to contract marine activities: 1 month. Time to acquire CPOD units: 2-3 weeks (done concurrently with discussions on the survey scope).

• Hydro-acoustic echosounder for near field fish movements.

CSTV would apply a similar scope as the FORCE fish monitoring component, but transects would involve multiple passes concentrated over Berth D.

Confirmation of survey scope with FORCE, EMAC, fishers, DFO and other stakeholders: 2 weeks Time to contract marine activities: 1 month. Time to acquire echosounder: 2 weeks Vessel acquisition: There are a few vessels that could do this work.

• Independent deployments or use of FORCE FAST platforms for sonar units.

Although the icListen units (hydrophones) have been deployed successfully on individual platforms, the Tritech sonar has not been used with a battery and would suffer from limited data storage. It is therefore recommended that if this contingency option is considered for the Tritech unit that the unit be deployed on a cabled FAST platform in the vicinity of Berth D or that a cabled platform be considered for any long-term use.

2 months to acquire instruments 1 week to prepare FAST-3 with new instruments (dependent on FAST-3 availability). 1-2 days for deployment of FAST-3 as an autonomous option. *If deemed necessary (i.e., battery power insufficient for useful lengths of data collection) cabling (fibre optic) for this platform can be considered: cable acquisition requires about 4-6 weeks.

11 It is assumed that the time to analyze the data and obtain results is the same as the originally planned monitoring methodology. 12 Redundancy has been provided for the icListen units. Four of these units are mounted on the subsea base – allowing for the failure of three units before data is compromised.

Page | 25

EEMP Component

Method Summary

Potential Risk Scenario

Proposed Options for Contingency

Considerations for Proposed Contingency Options

Implementation Times for Proposed Contingency Options11

Hydrophones can be deployed with battery/storage on an independent platform (i.e. non-cabled FAST platform). FAST-3 is the best immediate option for deploying the hydrophones. The platform is available, can be deployed using a small, local vessel (e.g. Tidal Runner), which allows for a faster deployment time (FAST-1 requires Dominion Victory, which is based in Halifax). FAST-1 would also require more changes to layout in order to direct battery power to acoustic recorders instead of the Vectron which uses a lot of power). FAST-3 is also autonomous but if required can be cabled. Use of the FAST-3 platform would be coordinated with FORCE activities.

Time to acquire Tritech unit: 11 weeks (rent or purchase). Time to acquire icListen unit: 9 weeks (rent or purchase)

If the need to have a FAST platform becomes a long-term option CSTV will need to return FAST-3 to FORCE and consider building a platform.

Design time: 4 weeks to consider design options with OpenHydro engineers and other experts. Time to build: 4 weeks. *If deemed necessary (i.e., battery power insufficient for useful lengths of data collection) cabling (fibre optic) for this platform can be considered: cable acquisition requires about 4-6 weeks.

Page | 26

Contingency for damage or loss of environmental devices Loss or damage to any of the monitoring devices is possible due to the extreme conditions of the Minas Passage. A number of contingency measures have been incorporated into the monitoring program to address this such as redundancy (more than one device), use of specific materials during manufacturing of the devices, and protective structures to surround devices, where possible. Table 3 provides details around potential contingency options related to damage/loss of a device. Contingency during retrieval/deployment operations Preparatory activities related to turbine retrievals or deployments involve times when there is a disconnection from the electrical supply to the Turbine Control Center. These activities are planned to occur within a single tidal window however; in the event that the turbine cannot be retrieved or deployed within the planning window (i.e., the neap tide), contingency monitoring must be implemented to avoid large gaps in data collection. Contingency will be considered for both short-term and long-term periods. Short-term contingency is considered for time periods of 2-4 weeks. The turbine will usually be disconnected from the subsea cable 2-3 days ahead of retrieval. This is likely to be shorter during deployment operations. A gap in monitoring data collection will occur during these times. During this time, FORCE’s monitoring of the CLA will continue and CSTV will maintain communications with DFO and NSE to ensure that they are aware of the situation. The data gap will be acknowledged in the reporting. Since operations can only occur during neap tides, windows for retrievals and deployments are very short in duration. If turbine retrieval is delayed to the next neap tidal window (i.e., two weeks later), the result is a short-term gap which will require implementation of short-term contingency monitoring options and/or collaboration with FORCE monitoring components that may already be on-going. All potential short-term contingency monitoring options will be developed and discussed with NSE,DFO and FORCE. Once all parties are satisfied with an approach, a formal letter outlining those options and when they will be implemented will be submitted by CSTV to NSE. Copies will be provided to DFO and FORCE. Long-term contingency is considered for time periods greater than 4 weeks. Turbine retrieval and deployment operations may experience delays due to unknown situations (e.g., debris around the turbine or subsea base that must be removed, weather conditions etc.) and also are dependent upon tidal and weather windows. Due to varying circumstances, there may be instances where retrieval and deployment operations are delayed to future tidal windows. In the event that this occurs, long-term contingency measures would be implemented to cover monitoring requirements. Where this can be done safely, CSTV could consider deployment of instrumentation similar to those devices already fixed to the turbine (i.e., imaging sonars). All potential long-term contingency monitoring options will be discussed with NSE, DFO and FORCE. Once all parties are satisfied with an approach, a formal letter outlining those options and when they will be implemented will be submitted by CSTV to NSE. Copies will be provided to DFO and FORCE.

Page | 27

5.0 Additional Items As noted in the Q1 report, CSTV is looking into additional monitoring components to complement the present monitoring program and to continue to meet the commitments made to stakeholders prior to deployment. The newest component will follow-up from an on-going study by OpenHydro, FORCE and Open Seas Instrumentation Inc., that involves the integration, testing, qualification and deployment of a second Gemini imaging sonar on one of the FORCE Fundy Advanced Sensor Technology (FAST) platforms. Wet tests were completed at the FORCE site while the turbine was in Saint John over the summer. The sonar will be deployed with the FAST platform and positioned near a turbine, to provide a view-plane of the side of the turbine structure. The purpose of this initiative is to explore the potential for monitoring the close vicinity of one side of the turbine rotor. Details on this new component of the EEMP are under final development and will be provided in the next report. Additional components include:

• Participating in FORCE’s beach walk monitoring program; and • Continuing to engage with fishery stakeholders to explore consultation processes.

Page | 23

Appendix 1

Gemini Imaging Software Target Identification:

Preliminary Analysis Report

Introduction The following document is a summary of a preliminary comparison of automated target tracking and human visual identification of possible targets using video recorded by the Gemini Multi-beam Imaging Sonar installed on the Cape Sharp Tidal Turbine gravity base and deployed at the Fundy Ocean Research Center for Energy in November 2016. For this report, the Gemini Imaging Sonar Software was used to detect targets in the available Cape Sharp video files recorded on November 17th, 2016. Automated target detection was conducted by Tritech International Ltd. The following manual target detection and comparison with automated detection was carried out by Acadia University. This analysis was conducted in order to examine the performance characteristics of the automated target detection software, with the assumption being that the manual target detection would be an accurate and reliable reference point. The eventual goal of this work is to use automated and manual target detection comparisons to improve the automated target tracking algorithm.

Gemini Data The Gemini Multi-beam Imaging Sonar was mounted on the Cape Sharp turbine gravity base at approximately 4 meters above the seafloor, and facing east towards Minas Basin. The Gemini acoustic camera images a swath of water using an array of 256 acoustic beams at 720 kHz. With the Gemini oriented horizontally, it has a 120° spread in the horizontal dimension and a 20° spread in the vertical direction. In the data files examined, the first few meters of data show the water column, after which the sampled volume extends to the seafloor (Figure 1).

Page | 24

Figure 1. Gemini Imaging Software screenshot, with the first five meters shown in white brackets and the view that includes the seafloor in blue brackets. The first 20 meters are shown in the magnification window overlaid on the right.

The video files were recorded at approximately 11 frames per second, which allowed the identification of potential targets (moving objects that could possibly be marine life). Figure 2 shows a typical example of an identified target moving toward the turbine (the target appears as a line, which traces its movement across the screen), although targets may vary in color, shape, and brightness depending on such variables as target size and distance to the Gemini Sonar. The individual video files reviewed were on average 307,400 KB in size and approximately five minutes long.

Figure 2. Example of an identified target in the Gemini video, labelled with a white arrow in the magnification window.

The semidiurnal tidal pattern on 17 November 2016, with two high and two low tides within 24 hours, is shown in Table 1. This day was chosen for an example because it was the day with the overall highest rate of detections, mostly false positives. Conversely, other days (e.g., November 21) were chosen because of the lowest rate of detections, which meant it was easier to examine each target individually. These comparisons allow researchers to compare different days and understand the scope of the analysis and how it can vary.

Page | 25

Table 1. Tidal cycle at the FORCE tidal test site for November 17, 2016. (Fisheries and Oceans Canada, 2016)

Atlantic Standard Time (hh:mm)

Coordinate Universal Time (hh:mm)

Water level (m) Tidal Stage

02:05 06:05 12.9 high 08:21 12:21 0.3 low 14:26 18:26 13.1 high

The tidal cycle and direction of current help to explain specific behaviours exhibited by targets, such as moving away from the turbine (ebb) or towards the turbine (flood).

Software Settings The following settings were used when manually reviewing the video footage:

• Image Orientation: The image was inverted upwards, oriented to the left, and rotated

upward (Figure 3). The most prominent feature on screen was a v-shaped structure, which

was located at a range of (-7.50 m, 8.48 m).

Figure 3. Image orientation of Gemini video during analysis showing vertex of prominent v-shaped structure (arrow).

• Gain: The gain, or brightness of the video on screen, was usually kept between 78% and

100% (Figure 4) depending on the background noise present, which showed up as ‘speckles’

on the screen. However, for some video files lower gain was necessary as the video image

itself already appeared very bright, making it difficult to identify brighter targets. Higher gain

proved to be more effective in the identification of smaller targets.

Page | 26

Figure 4. Comparison of Gemini video frame viewed at 78% gain (left) and 100% gain (right).

• Speed: The speed of the video replay was adjusted according to ease of viewing and the

number of identifiable targets. When there was an increase in background noise or number

of targets present, the speed was reduced to 90%.

• Averaging filter: Averaging (in the ‘Advanced’ tab, under filter settings) takes a weighted

average over current and previous frames to smooth frame-to-frame variation, which can

make moving targets easier to see against the stable background. Averaging was left at 50%

because no improvement in the video quality could be observed when changing this setting.

• Persistence filter: The persistence can be used to highlight movement across frames. As

the frames progress, the persistence filter retains a decreasing number of earlier frames as

well as the current frame, creating a line that tracks a target’s movement over time. The

persistence level was changed depending on the targets being identified. Extreme persist

(99% and 99.5%) was the most effective when trying to track smaller, dimmer targets, as

their track would persist longer, making it easier to pinpoint target position over time. Long

persist (96% or 98%) was more suitable for larger and brighter targets, as they were easily

identifiable without extreme persistence (Figure 5). The shorter persistence caused the

tracks to disappear faster, which made it easier to move forward and backward frame by

frame without the track remaining in view.

Page | 27

Figure 5. Comparison of the same target identified from the Gemini video with 96% (left) and 99.5% (right) persistence, with magnification windows focused on target.

• Movement filter: The movement filter, which reduces signal from any objects that remain

stationary from frame to frame, was kept at 60% to allow better identification of targets, but

this filter level could be increased if background noise is low (Figure 6).

Figure 6. Comparison of video viewed with 100% movement filter (left) versus 0% movement filter (right).

• Target tracking: Target tracking was not enabled as it was not accurate enough for smaller

targets (less than 0.5 meters). For this study, only manual target detection was used for

tracking movements.

Target Identification The following procedure was used to identify possible targets (moving objects that may be marine animals). The automated tracking program was used by Pauline Jepp at Tritech to identify files with many potential targets, and these files were then reviewed to get an idea of what targets look like. Once accustomed to target identification, entire five-minute files were viewed. Each video file was watched twice, focusing on different ranges each time. First, the 0-15 m range was watched using 1.5x magnification to identify smaller targets (Figure 7), which were generally only visible at close range due to decreasing resolution with range in addition to bottom interference and ‘specks’ decreasing visibility on the screen.

Page | 28

Figure 7. Example of the larger view of the Gemini video, with the 1.5x magnification window (right) used for viewing the first 15 m.

When watching a second time, the focus was on the 15-60 m range. The use of magnification was discontinued after examining three videos because often what could be seen on the magnification screen could not be identified on the larger view.

During viewing of the Gemini video, potential targets were first identified, then the video was re-wound and watched frame by frame (by clicking the frame button). The target position and time was recorded for every second, or more if the target was either moving larger distances over the one second time frame or the track showed up within one second (Table 2).

Page | 29

Table 2. Example of Excel sheet created for targets manually detected in Gemini video. Columns are file number, date, time, position, notes, tidal stage, target size, matched automated detection, and the numerical IDs of automated targets not detected manually.

Later, only the positions at the start and end of the target track were recorded. If a target could not be clearly identified (e.g., if there was increased background noise, or a broken up path), the uncertain targets were classified as either possible or probable targets.

Possible fish aggregations (Figure 8) were identified as a single unit, with a singular time stamp and estimate of the general range of the fish aggregation. Any unusual properties of a target (e.g. very bright and/or wide track) or any particular movements (e.g. moving away from the turbine) were also noted.

Comparison of automated and manual target detections Once all videos in which the automated tracking had identified targets were reviewed, automated detections were compared to those detected manually and matches were identified. A match was determined using the time of automatic detection and the range. Only 2 of the 135 targets identified through the automated tracking were matched to the manually detected targets in the videos reviewed to date. It should also be noted that the automated program identified fewer targets than detected by the human observer.

Page | 30

Figure 8. Example of a possible fish aggregation (white box).

Conclusion Overall, the quality of the video was sufficient to become comfortable with target identification. With experience, targets can be distinguished easily and confidently. One issue that was encountered was that the video did not run smoothly. Jumps of a second or more were common, making target identification and tracking more difficult. This was due to communications issues between the on-shore computer and the Gemini, which has been addressed for the future deployment (refer to Section 4.1.2 of the main Q3 Report). The automated tracking performance needs improvement, as substantially fewer targets were identified by the automated software than by the human observer, and even fewer could be matched with the manually detected targets.

There are a couple of changes that could be made to the software to make it more “user” friendly. It would be beneficial if the position of a target could be pinpointed more accurately. This could be done, for example, if one were able to pinpoint a target’s position directly on the magnification window instead of only in the larger view.

Next steps will include further examination of more of the collected Gemini data. This will provide more comparison material for the development of the automated target detection algorithms. Attempts will also be made to better identify the targets (e.g. as fish).

Page | 31

References

Fisheries and Oceans Canada. (2016). 7 day tidal predictions: Cape Sharp (#250). Retrieved from http://www.tides.gc.ca/eng/station?type=0&date=2016%2F11%2F17&sid=250&tz=AST&pres=.

Tritech International Ltd. (n.d.). Gemini Imaging Sonar Product Manual: Document: 0685SOM-00001, Issue: 10. Retrieved from http://www.tritech.co.uk/media/support/manuals/geminiimaging-sonar-product-manual.pdf.

Appendix B Ocean Sonics – icListen Final Data Analysis Report

Ocean Sonics Ltd.

JANUARY 2018

DATA ANALYSIS REPORT

Tidal Turbine Data Analysis January 2018 P.1

Table of Contents Executive Summary ....................................................................................................................................... 2

Introduction .................................................................................................................................................. 3

Overview ....................................................................................................................................................... 3

Data Analysis ................................................................................................................................................. 5

Acoustic Patterns .................................................................................................................................... 12

Planning for Next Deployment.................................................................................................................... 14

Lessons Learned ...................................................................................................................................... 14

Recommendations for Future Monitoring Considerations ..................................................................... 14

References .................................................................................................................................................. 16

Appendix A .................................................................................................................................................. 17

Appendix B .................................................................................................................................................. 18

Tidal Turbine Data Analysis January 2018 P.2

Executive Summary Ocean Sonics analyzed the acoustic data collected by icListen Smart Hydrophones from the Cape Sharp Tidal Turbine during the 2016/2017 deployment. Four icListen Smart Hydrophones were set up to sample both waveform data (WAV) and processed spectral data (FFT). The hydrophones would then stream the data sets (WAV and FFT) from the turbine to an onshore FORCE substation via a subsea cable. Due to logistical issues, only a subset of this data was recovered. The limited data prevented a full analysis of the deployment.

The data recovered was processed using harbour porpoise click detectors on Lucy and Coda to determine when harbour porpoises were detected in the Minas Passage. Visual inspection, screenshots of spectral data and third octave processing was also performed.

Results show porpoise click detections on multiple days per month during the deployment period from November 2016 to April 2017. Lucy detected the most harbour porpoise clicks during March and April, which would relate to seasonal variability of the area. March had the most porpoise detections, with detections present on 15 days. April had porpoise detections 8 of the 13 days data was collected, matching the trend found in March. The Lucy click detector had difficulty detecting clicks in high tidal flow noise. Coda detected harbour porpoise click trains throughout the tidal cycle in current speeds between -70% to +90%. Coda detected more harbour porpoise click trains during the night.

Ocean Sonics recommendations for future monitoring • Collection of acoustic emissions for other instruments involved,

to model the soundscape near the turbine • Increased protection of hydrophones • Improved on-site data management plan • Scheduled real-time remote access to hydrophones for diagnostics • Further investigation using porpoise click detectors with future tidal turbine data • Further comparison analysis is suggested with published reports

Tidal Turbine Data Analysis January 2018 P.3

Introduction This report describes the data analysis performed by Ocean Sonics on the acoustic data received from the Cape Sharp Tidal (CST) turbine deployment from November 2016 to April 2017. The report addresses the marine mammal component of the Environmental Effects Monitoring Program, EEMP; specifically, for the harbour porpoise (Phocoena phocoena).

Overview During the 2016/2017 deployment, acoustic data from the CST turbine site was collected using four icListen Smart Hydrophones (Figure 1). The hydrophones were mounted in four different locations on the turbine and subsea base prior to deployment. One hydrophone was located at the top of the rotor assembly and the other three were located inside the three corners of the subsea base (Appendix A).

Figure 1. icListen Smart Hydrophone

Each of the four icListen Smart hydrophones were set up to sample both waveform datasets (WAV) and processed spectral Fast Fourier Transform datasets (FFT). The WAV data is the raw data sampled by the hydrophone in the time series domain. This data is needed to replay the audio part of the recording. FFT is an algorithm that samples a signal over a period of time and divides it into frequency components. This method is used by the hydrophone to convert waveform data (time domain) to frequency data (spectrum). The FFT data is a more compact representation of the acoustic soundscape. Much of the data is not audible to humans but it can be visualized by analysts with real-time processed FFT data.

Both data sets (WAV and FFT) were set-up to stream from each icListen stationed on the turbine to an onshore FORCE substation via a subsea cable. The substation stored both WAV and FFT datasets with the use of Lucy, a PC program that allows researchers to view and interact with acoustic data collected by the hydrophones. Lucy is a powerful program capable of streaming and recording accurate real-time acoustic measurements. The hydrophones themselves also stored the FFT data in their internal memory. The hydrophones were set-up by OpenHydro at the beginning of the deployment with sampling rates logged at the shore station of 512 kilo-samples per second (kS/s) for FFT data and 32 kS/s WAV data.

The FFT data was first reviewed in Lucy. The program includes a porpoise click detector that uses intensity to indicate a porpoise click in the data, amongst other user configurable event triggering. An overview of each day was collected, and events in the spectral data were noted. The porpoise click detector was used and visual confirmation was made for each potential click found in the program. After a click was confirmed it was recorded as a porpoise click. The data from each hydrophone was then

Tidal Turbine Data Analysis January 2018 P.4

reviewed and compared for discrepancies in the quantity and quality of data logged. The results were also compared to operational data of the turbine as seen in Appendix B.

The data was then run using two types of porpoise click detector software; PAMGuard and Coda. These software programs were first used to locate porpoise clicks in the data and then the matches were used to find porpoise click trains. A porpoise click train is series of clicks, described by the time between clicks, the inter-click interval (ICI). The click detectors used a minimum of 3 clicks and an ICI of 0.2 seconds to define a click train. The click train is used to minimize false positive detections by eliminating single clicks from other sources. Each click detector has different algorithms, and thus behaves differently under the full range of conditions. PamGuard settings were adjusted for the data but the high tidal flow noise and sonar signals caused many false positive and false negatives, with few true positive detections. Coda performed better in the high noise environment, so it was chosen as the preferred program for assessing porpoise presence.

The final data processing step involved a third octave analysis which was used to compare acoustic data during different tidal flows: flood (loud); and slack (quiet) tides. The third octave analysis is performed as a sound power distribution which splits the power spectrum into adjacent one-third octave frequency bands. This presents the acoustic data in a logarithmic frequency scale on the x-axis and sound pressure level on the y-axis of a graph. The third octave analysis is useful because it can be used to understand the broadband sound pressure level and demonstrate frequency dependent propagation characteristics of an environment, over time. The acoustics community has adopted standard 1/3-octave frequencies to facilitate comparisons between studies which are used in this report (MacGillivray & Chapman, 2005).

The high tidal flow noise environment of the FORCE site is important to note because of the interference, especially in the upper frequencies where porpoise clicks are located. If the sound of the tidal noise is greater than the porpoise clicks (signal), the detectors would not be able to differentiate a click from the surrounding sound. This would be too low signal to noise ratio and shows the importance of signal to noise ratio in detecting porpoise clicks. The tidal flow noise would be considered acoustic masking when it is masking or concealing the signals of interest. The turbine also has active sonar equipment that create additional sound in the upper frequencies where harbour porpoise clicks occur further increasing the difficulty of detection by the hydrophones.

Some issues were experienced during the deployment (Table 1). The hydrophone (Hydrophone 3 – Appendix A) located on the top of the rotor was damaged early in deployment, likely due to debris travelling in the water column. Although the unit still collected data, the damage compromised that data past November 11, 2016 and it could not be used for the analysis. Hydrophones 1406 and 1407 had communication issues related to cabling during the deployment and collected data periodically. Hydrophone 1404 collected data throughout the deployment.

On March 8, 2017, the hydrophone sampling rate for the WAV data from one hydrophone (1404) was remotely increased to 64 kS/s and then increased further on March 24, 2017 to 512 kS/s (200 kHz). OpenHydro increased the sampling rate by remote access to the FORCE substation while the turbine was still deployed. The sampling increase was made for hydrophone 1404 as it had the most reliable communication. This increased sampling rate for the waveform was needed because early analyses of data indicated that the original set-up was a suboptimal sampling rate for use in the click detector programs. The optimal sampling rate needed for Coda and PAMGuard is 512 kS/s to sample up to 256 kHz.

Tidal Turbine Data Analysis January 2018 P.5

Data Analysis Table 1 provides a summary of the FFT and WAV data recorded for each hydrophone during specific periods.

Table 1. Data recorded on each hydrophone from November 8, 2016 to April 13, 2017

Hydrophone Serial Number Dates FFT WAV 1404 11/08/16 – 03/08/17 512 32

03/08/17 – 03/24/17 512 64 03/25/17 – 04/13/17 512 512

1405 11/08/16 – 03/08/17 512 32

03/08/17 – 04/13/17 512 64

1406 Invalid data (periodic recording)

1407 11/09/16 – 11/14/16 512 32 Tidal Current Percentages Percentages were used as a measurement of current speed by OpenHydro with -100% and +100% being the greatest current speeds and 0 being no current speed known as slack tide. The negative and positive current speeds are used to show the ebb (falling) and flood (rising) tides respectively.

Lucy The Lucy click detector software was used with the FFT data from November 8, 2016 to April 13, 2017. A detected click was reviewed visually by a data analyst to determine viability and if a positive visual identification of a click was made, the day was recorded. Preliminary results from Lucy showed porpoise clicks on multiple days each month of the deployment. March had the most visits with 15 days over the month (Table 2); however, it should be noted that data was only collected until April 13 (at which time hydrophones were disconnected for retrieval). The number of porpoise click detections in the beginning of April is at least proportional to March. This demonstrates the seasonal frequency of harbour porpoise detection in the FFT data set.

Table 2. Days with Detected Porpoise Clicks with the Lucy Click Detector

Month Days of Porpoise Detections November 10, 11 December 23, 24, 25, 29 January 1, 4, 8, 29 February 2, 5, 23, 24, 26 March 1, 2, 3, 5, 10, 12, 13, 20, 21, 23, 25, 26, 28, 30, 31 April 1, 2, 6, 7, 9, 10, 11, 12

The results from the days of porpoise detection (Table 2) were also compared to operational data of the turbine as seen in Appendix B. The graph shows time during turbine power generation and porpoise

Tidal Turbine Data Analysis January 2018 P.6

click detections. Initial results could suggest an inverse relation when power is being generated, however, further research over a longer time period and with fully functioning devices is required in order to draw any conclusions.

Lucy porpoise click detections were made by visual inspection so individual clicks could be identified as porpoise activity, as well as, click trains. Porpoise clicks were detected with high frequency sonar signals in the data and identified during slack tide and during the beginning of tidal flow. Verification of porpoise clicks during tidal flow was difficult because the low signal to noise ratio which caused acoustic masking. Lucy had difficulty detecting clicks during high tidal flow noise. Examples of porpoise detection in the Lucy program are shown below in Figures 2 and 3.

Figure 2. Screenshot from Lucy. Dec. 23, 2016. Reference 60 dB re 1µPa with a 10 dB step. Porpoise clicks are shown in the middle of the spectrogram with the detection intensity spiking in the upper graph. Horizontal banding from 120 to 204 kHz is found throughout the spectrogram, the sound was created by the sonar impulse signals. This shows the Lucy Click Detection software can identify porpoise clicks while there are sonar signals present.

Tidal Turbine Data Analysis January 2018 P.7

Figure 3. Screenshot from Lucy. Dec. 23, 2016. Reference 60 dB re 1µPa with a 10 dB step. Porpoise clicks are shown in the middle of the spectrogram, at the beginning of a tidal flow. The graph indicates increasing noise on the right-hand side of the spectrum as tidal flow increases. The upper graph shows the rise in intensity where the clicks detected by the Lucy Click Detector.

Coda Coda was used to detect porpoise clicks for the tidal turbine data from March 25 to April 13, 2017. The porpoise clicks detections were noted for four days in March and 11 days in April (Table 3).

Table 3. Days with Porpoise Click Trains Detected by Coda

Month Days with Porpoise Click Train Detection March 25, 27, 28, 31 April 1, 2, 5, 6, 7, 8, 9, 10, 11, 12, 13

For the data collected from March 25 to April 13, 2017; Coda detected porpoise clicks all days except for March 26, 29, and 30 and April 3 and 4, 2017. High intensity harbour porpoise detections were found by the Lucy Click Detector as well as Coda on April 11 and 12, 2017. The mean current when there were click train detections was 44.25 %, of peak current.

Coda detected, and recorded porpoise clicks in the data and then processed the detection results further to find and record click trains. Click trains were used as a detection method and further verification of the clicks to minimize false positive and maximize true positive detections. The click train parameters used were 3 or more detected clicks with a maximum 0.2 second inter-click interval. The click train data were then recorded, reviewed and then analyzed to provide more information on the porpoise clicks detected, as described in the table below (Table 4). The click trains were then compared to the current speed to determine the rate of detections regarding current speed.

Tidal Turbine Data Analysis January 2018 P.8

The following graphs and tables use DP5M (detection positive five-minutes) as the base measurement for detection. This was used because the acoustic waveform data files were five minutes in length. DP5M refers to the number of detection-positive 5-minute intervals.

Table 4. Coda Statistics on Click Trains Detected.

Day #DP5M #Trains #5minute p se 25-Mar-17 1 1 148 0.007 0.0067 26-Mar-17 0 0 288 0 0 27-Mar-17 4 20 288 0.014 0.0069 28-Mar-17 6 17 288 0.021 0.0084 29-Mar-17 0 0 288 0 0 30-Mar-17 0 0 288 0 0 31-Mar-17 2 3 288 0.007 0.0049 01-Apr-17 3 10 288 0.01 0.006 02-Apr-17 4 7 288 0.014 0.0069 03-Apr-17 4 5 288 0.014 0.0069 04-Apr-17 0 0 288 0 0 05-Apr-17 2 3 288 0.007 0.0049 06-Apr-17 6 10 288 0.021 0.0084 07-Apr-17 8 18 288 0.028 0.0097 08-Apr-17 6 15 288 0.021 0.0084 09-Apr-17 5 10 288 0.017 0.0077 10-Apr-17 5 9 288 0.017 0.0077 11-Apr-17 40 1424 288 0.139 0.0204 12-Apr-17 10 136 288 0.035 0.0108 13-Apr-17 2 3 54 0.037 0.0257

Notes: #5minute refers to the number of 5-minute intervals of acoustic data that were collected each day; #Trains refers to the number of harbour porpoise click trains; p refers to the proportion of detection-positive 5-minute intervals in a day; (p = #DP5M /#5minute); se refers to the standard error of p; [se = sqrt{p*(1-p) / #5minute}]

Tidal Turbine Data Analysis January 2018 P.9

The graphs below show the proportion of detection positive 5-minutes distributed as a function of hour of the day, in 2-hour increments (Figure 4) and current speed (Figure 5). The vertical lines indicate standard error.

As shown in Figure 4, the hours with the highest proportion of porpoise clicks were found between 22:00 to 06:00 (UTC); at the turbine site this would have been between 19:00 to 03:00 ADT, Atlantic Daylight Time, suggesting greater use of the site at night.

Figure 4. Proportion of Detection Positive 5 minutes, +/- standard error comparing hours of the day.

Figure 5 indicates that the current speed with the highest proportion of DP5M is -30%.

Tidal Turbine Data Analysis January 2018 P.10

Figure 5. Proportion of Detection Positive 5 minutes, +/- standard error comparing current speed based on OpenHydro measurements using percentages.

Tidal Turbine Data Analysis January 2018 P.11

Overview of Click Detection Methods Lucy with Visual Inspection

• Can use FFT or WAV data for processing • Loses ability to detect porpoises in high flow noise environments • Can process real-time or stored data • Based on spectral cross-correlator • Click model can be fine-tuned • Labour intensive due to visual inspection • Detections based on single porpoise clicks

Coda

• Only uses WAV data • Created to identify clicks in noisy data • Can process real-time or stored data • Data output of detections can be reviewed, and results further processed in the program • Can be programmed to selectively record data based on porpoise click detections • Detections can be shown by each click or by click trains

Lucy was used to identify clicks in the FFT data, while Coda was used for the last month of data where there was full bandwidth WAV data collected. Coda was developed based on the experience of the past five years using the Lucy Click Detector.

Differences found between results in days with porpoise detection is based on the different algorithms used in the two methods. Lucy used single clicks or click trains but all clicks were individually confirmed by a data analyst, increasing labour intensity. Coda used a further processing method of detecting only porpoise clicks that were in click trains, this allowed for less false and more true positive detections.

Third Octave Analysis Sound levels were found using third octave processing for a tidal cycle on April 3, 2017. April 3, 2017 was chosen for third octave analysis because it had a representative tidal cycle from the waveform data after the sampling rate was increased to full bandwidth. The third octave graph shows the difference in sound levels during the tidal cycle. The increased ambient noise levels created by the flow of tides is around 130 kHz which is the level at which a porpoise clicks (Table 5 and Figure 6). This can create difficulty in detecting clicks because of the low signal to noise ratio. Table 5 shows the SPL at 1 kHz, chosen as a base measurement for tidal noise and 130 kHz the frequency where harbour porpoise clicks are located. Table 5. Sound pressure level (SPL) at two chosen frequencies 1 kHz and 130 kHz

State of Tide SPL at 1 kHz SPL at 130 Hz Slack Tide 79 dB re μPa 94 dB re μPa Flood Tide 135 dB re μPa 144 dB re μPa

Tidal Turbine Data Analysis January 2018 P.12

Figure 6. Third Octave Graph showing the difference in sound pressure level (dB re μPa) between slack tide and flood tide.

Localization Using four hydrophones allows for the option of localizing sound sources that were detected (i.e., determine the position of a vocalizing animal in relation to the turbine). During the recent deployment however, only one usable channel was acquired so localization could not be performed.

Marine Mammal Vocalizations Marine mammal vocalization processing was not performed due to the quality and quantity of recorded data.

Acoustic Patterns Impulsive Signals from Active Sonar The majority of the data contained many broadband impulsive signals, believed to be from the active sonar mounted on the CST turbine. The sonar created the impulsive signals at various intervals and the sound intensity was greatest between 150 to 200 kHz. The sonar emitted sound between 720 kHz +/- 50 kHz. The impulsive signals found in the data between 150-200 kHz were sideband effects from the higher frequency emissions of the sonar. The signals increased the noise levels in the upper frequencies where harbour porpoise clicks are located. The impulse signals created false positive porpoise clicks in PAMGuard without revised settings.

Figure 7. Screenshot from Audacity showing the sonar signals over 5 minutes. The spectrum display has frequency on the left side between 0-256 kHz. The top of the display on the figure is the time axis, with 0-5 minutes.

20

40

60

80

100

120

140

160

180

20 200 2000 20000 200000

Soun

d Pr

essu

re L

evel

(dB

re μ

Pa)

1/3 Octave Band (Hz)

Slack Tide

Flood Tide

Tidal Turbine Data Analysis January 2018 P.13

Tidal Flow Pattern During both flood and ebb tides there is a substantial amount of noise being recorded by the hydrophone due to the tidal flow noise. A distinct pattern of high-frequency ambient sound increases as current picks up as can be seen in Figures 8 and 9.

Figure 8. Lucy screenshot from hydrophone 1404 on March 28, 2017. A reference of 70 dB re µPa, 10 dB steps. The waterfall display shows the flood and ebb tides, from a 24-hour period. The spectrogram begins with a flood tide then the ebb, and repeats. There is a porpoise click detection between the first flood and ebb tide detected by the spike in the Lucy Click Detector in the upper intensity graph shown by a red box.

Figure 9.Lucy screenshot from hydrophone 1404. A reference of 70 dB re µPa, 10 dB steps. Three days of data showing tidal cycle pattern from April 4th, 5th and 6th, 2017. Tones from the sonar are clearly visible in the upper frequencies.

Tidal Turbine Data Analysis January 2018 P.14

Planning for Next Deployment Lessons Learned Over the course of the deployment there were learning opportunities that can be used to improve the data quality and quantity for the next deployment. Issues with data loss (due to a failure of the communication network), synchronization set-up issues and loss of the sensor in the hydrophone mounted on the top of the rotor have all been discussed with OpenHydro and integrated into a plan to prepare for future monitoring during the next deployment.

Mitigation measures and improvements have also been developed for implementation in the next deployment to progress the passive acoustic monitoring of the near-field area of the turbine. This includes detailed set-up for the hydrophones to sample at full bandwidth, new cabling, more robust guards for increased protection of the sensor tips and specialized guard bars on either side of the hydrophone on the top of the rotor. All hydrophones will be able to record data that can be used in the porpoise click detectors. There is also ongoing work to understand the synchronization of the units with the active acoustic system. The result will be a fully synchronized system that is expected to use passive acoustic data to locate noise based on arrival times, such as the detection of porpoise clicks used for harbour porpoise localization. Although there was success in determining porpoise clicks, no localization was performed because data was recorded on just one channel, preventing the possibility of localization. Specific speed categories were not defined because the turbine was in a commissioning phase over the deployment period. This will be further investigated during the next deployment.

Recommendations for Future Monitoring Considerations The following are recommendations to consider for future monitoring

1. A better understanding of other equipment on the tidal turbine and the acoustic emissions with respect to passive acoustic monitoring. This will be useful for a more in-depth acoustic analysis in the future.

2. Scheduled real-time access to hydrophones, as required, for ongoing diagnostics will improve identification of potential concerns and timely implementation of solutions.

3. Implementation of protective devices for the hydrophones to protect from debris carried by the current.

4. Improvements to data management prior to the planned increase in data collection will facilitate the access and analyses of data. It is expected that the hydrophones will be recording and sending 200 kHz processed spectral data and up to 200 kHz waveform data to the shore station resulting in up to 512 GB of sound data per day or approximately 15 TB per month. This requires a formal data handling protocol.

5. Further investigation of the PAMGuard and Coda porpoise click detector programs to increase understanding of when porpoises are present in comparison to turbine operations. Further investigation of PAMGuard with additional full bandwidth WAV data could also provide more

Tidal Turbine Data Analysis January 2018 P.15

favourable results. Building on datasets such as this will result in a greater understanding of the use of the Minas Passage by harbour porpoise, allowing for the development of a future strike risk model.

6. Further data analysis is suggested for the following a. Comparison of turbine operations and acoustic data; b. Comparison of published studies with full bandwidth WAV data; and c. Marine mammal detection with PAMGuard whistle and moan detectors.

Tidal Turbine Data Analysis January 2018 P.16

References MacGillivray, A. O., & Chapman, N. R. 2005. Results from an acoustic modelling study of seismic airgun survey noise in Queen Charlotte Basin. University of Victoria report.

Tidal Turbine Data Analysis January 2018 P.17

Appendix A

Figure 10. Tidal Turbine diagram showing hydrophone placement

Caption in Figure Hydrophone Serial Number Hydrophone 1 1407 Hydrophone 2 1404 Hydrophone 3 1405 Hydrophone 4 1406

Tidal Turbine Data Analysis January 2018 P.18

Appendix B

Appendix C Acadia University and Tritech Ltd. – Gemini Sonar Final Data Analysis Report

Cape Sharp Tidal Gemini Multibeam Imaging Sonar: Monitoring Report

(November 2016 - April 2017)

Report to Cape Sharp Tidal

Prepared and submitted by

Haley Viehman, Franziska Gnann, and Anna M. Redden

Acadia Centre for Estuarine Research

Acadia University Wolfville, NS

ACER Technical Report No. 123

Citation:

Viehman, H., F. Gnann and A.M. Redden. 2017. Cape Sharp Tidal Gemini Multibeam Imaging Sonar: Monitoring Report (November 2016 - April 2017). Report to Cape Sharp Tidal. ACER Technical Report No. 123, 36 pp, Acadia University, Wolfville, NS, Canada.

CST Gemini Monitoring Report

ii

Executive Summary

Cape Sharp Tidal (CST) installed an OpenHydro Open-Centre tidal in-stream energy conversion (TISEC) device on 7 Nov 2016, in Berth D at the Crown Lease Area at the Fundy Ocean Research Center for Energy (FORCE). The turbine was retrieved in June 2017. As part of this demonstration project, CST implemented an environmental effects monitoring program (EEMP). The EEMP was initiated upon deployment (November 2016) until disconnection with the subsea cable in April 2017, as part of the preparation operations for retrieval.

The overall purpose of the CST EEMP is to better understand potential effects and interactions of specific environmental components (i.e., fish, marine mammals, operational sound) in the near-field environment with the Open-Center in-stream tidal device. This understanding will be useful for verifying the accuracy of the environmental effect predictions made in the environmental assessment and will inform future monitoring plans.

This report addresses the fish component of the EEMP. Active acoustic monitoring was used to gather information on the occurrence of fish within the near-field (i.e., < 100 m) area of the turbine. To achieve this, a Gemini multibeam imaging sonar was mounted on the turbine structure and used to monitor marine life in the near-field area during the 2016/2017 deployment. The Acadia Centre for Estuarine Research at Acadia University was contracted to analyze the data collected by the sonar and address the specific objectives of the active acoustic Gemini sonar study under the CST EEMP.

The goals of the Gemini sonar study were to increase understanding of potential interactions of marine life with in-stream tidal turbines, including the use of the site by wildlife, as determined by target detection and tracking, and to further develop monitoring methodology as it relates to Gemini data collection, processing, analysis, and presentation. The work described in this report used manually-processed Gemini data to (1) assess trends in target abundance within the sampled volume, over short and long-time scales and with respect to tidal stage and current speed; (2) characterize target movement with respect to current direction; and (3) identify targets that may be fish schools.

Data collected were from the near-field area immediately in front of the turbine (facing into the current during ebb tide). Given the sonar’s downward-angled orientation, the volume sampled was below the depth of the turbine rotor. A subset of the sonar dataset was manually processed to identify and track targets in the volume of water column sampled. This involved a human observer reviewing five-minute long video clips at two-hour intervals for 1 full day per week, for the full five-month period of data collection. The observer searched for ‘targets,’ which are defined as objects moving independently of the seafloor background that could be marine life. The time of detection and the net movement direction was recorded for each target. Target abundance was found to decrease with falling winter temperatures, which is consistent with other biological surveys of this area. Target abundance did not differ significantly between day and night for the duration of the dataset, but abundance of targets in the volume sampled was consistently lower during the flood tide than the ebb tide, possibly due to effects on the flow field in the area sampled by the sonar as flood tide waters moved through and around the TISEC

CST Gemini Monitoring Report

iii

device. Target movement direction exhibited patterns that reflected the flow environment, with most targets moving in the same general direction as the current. However, variation in movement direction of targets within the sampled volume was greater during the flood tide, when targets were downstream of the turbine, than during the ebb tide, when targets were upstream (approaching the device). Again, this difference could be related to the physical effect of the TISEC device on the flow regime in the near-field; examination of fine-scale hydrodynamics upstream and downstream of the device would be needed to determine wake effects.

The results of target detection and tracking presented here are encouraging for the future use of the Gemini sonar to monitor marine life presence and behavior at turbine rotor height in the near-field of the CST TISEC device. Efficiency and extent of sonar data processing will increase greatly with further development and validation of automated Gemini data processing techniques. Assessment of the near-field hydrodynamics, and examination of the data provided by FORCE’s mobile and stationary active acoustic surveys of fish, will be important for the interpretation of Gemini sonar data collected from turbine rotor height in future studies.

The potential for an overall improved dataset from a re-oriented Gemini sonar and a longer (planned) deployment of a turbine will provide an opportunity to obtain data with increased spatial and temporal coverage. This will help to clarify the results presented and discussed in this report, improve understanding of year-round presence and spatial distributions of marine animals, and therefore help to meet the overall objective of understanding how fish and marine mammals might interact with the CST in-stream turbine.

CST Gemini Monitoring Report

iv

Contents

1 Introduction ................................................................................................................................. 1

1.1 Overview and EEMP context ................................................................................................ 1

1.3 Gemini multibeam sonar ....................................................................................................... 2

1.4 Project objectives .................................................................................................................. 3

2 Methods....................................................................................................................................... 3

2.1 Data collection ...................................................................................................................... 3

2.2 Data processing ..................................................................................................................... 5

2.2.1 Target identification ....................................................................................................... 6

2.2.2 Data subsampling ........................................................................................................... 7

2.2.3 Auxiliary data................................................................................................................. 8

2.3 Data analysis ......................................................................................................................... 8

3 Results and Discussion ............................................................................................................... 9

3.1 Trends in target abundance and size ..................................................................................... 9

3.1.1 Seasonal trend in target size ......................................................................................... 12

3.1.2 Seasonal trend in target abundance .............................................................................. 14

3.1.3 Diel and tidal trends in target abundance ..................................................................... 15

3.1.4 Target abundance in relation to current speed ............................................................. 16

3.1.5 Seasonal trend in size of target aggregates (potential schools) .................................... 17

3.2 Target movement and direction .......................................................................................... 18

3.2.1 Individual targets ........................................................................................................ 18

3.2.2 School-like targets ...................................................................................................... 21

4 Conclusions ............................................................................................................................... 21

5 Recommendations ..................................................................................................................... 22

5.1.1 Data collection ............................................................................................................. 22

5.1.2 Data processing ............................................................................................................ 23

5.1.3 Data analysis and interpretation ................................................................................... 23

5 References ................................................................................................................................. 25 6 Appendix 1 ……………………………………………………………………………………28

CST Gemini Monitoring Report

1

1 Introduction

1.1 Overview and EEMP context The Cape Sharp Tidal (CST) tidal in-stream energy conversion (TISEC) device was deployed at the Fundy Ocean Research Center for Energy (FORCE) test site from November 2016 to June 2017 (Figure 1). This test site is located in the 5.5-km wide Minas Passage of the upper Bay of Fundy, where the tidal range reaches 13 m and currents can exceed 5 m∙s-1 (Karsten et al. 2013). The CST device is an OpenHydro design, which consists of an open-center turbine mounted on a stationary bottom support frame, the subsea base, resting on the sea floor (Figure 1b). The device is 20 m high, and the turbine is 16 m in diameter.

Figure 1. Study location and diagram of in-stream tidal energy device. (a) the Fundy Ocean Research Center for Energy (FORCE) tidal energy demonstration site in Minas Passage, Nova Scotia, showing the location of the device

within Berth D. (b) The Cape Sharp Tidal TISEC device, with arrow indicating Gemini imaging sonar.

The upper Bay of Fundy is home to a diverse seasonal assemblage of fish, marine mammals, seabirds, and other marine fauna, some of which are commercially important (e.g. river herring, Alosa sp.), threatened (e.g., Atlantic sturgeon, Acipenser oxyrinchus; American eel, Anguilla rostrata) or endangered (e.g., striped bass, Morone saxatilis) (Baker et al. 2014, Keyser et al. 2016, Redden et al. 2014, Stokesbury et al. 2017). The potential effects of in-stream tidal devices on marine life in the Minas Passage are currently unknown. The uncertainty is mainly due to the low number of TISEC devices that have been deployed worldwide to date. However, recent years have seen a growth in the extent of research on marine life and potential interactions with in-stream turbines in areas with fast tidal currents (Copping et al. 2016), including:

1. behavior of animals near TISEC devices (Bevelhimer et al. 2017, Hammar et al. 2013, Viehman and Zydlewski 2015);

2. spatial and temporal distribution of animals at tidal energy sites (Benjamins et al. 2016a&b, Daroux and Zydlewski 2017, FORCE 2017, Keyser et al. 2016, Melvin and Cochrane 2014, Redden et al. 2014, Stokesbury et al. 2017, Viehman and Zydlewski

CST Gemini Monitoring Report

2

2017, Viehman et al. 2015, Viehman et al. 2017, Waggitt et al. 2016a&b, Waggitt et al. 2017);

3. likelihood that animals may overlap with TISEC devices (Sanderson et al. 2017, Shen et al. 2016, Viehman et al. 2017); and

4. methods and best practices for gathering and analyzing the necessary data to detect TISEC device effects (Fraser et al. 2017, Jacques and Horne 2014, Jepp 2017, Wiesebron et al. 2016a&b, Williamson et al. 2015, Williamson et al. 2017).

This study aims to reduce uncertainty by building knowledge of the presence and behaviour of marine wildlife in the near-field of a CST TISEC device at FORCE. The research presented here forms part of the CST environmental effects monitoring program (EEMP). The overall purpose of the CST EEMP is to better understand interactions of specific environmental components (i.e., fish, marine mammals, operational sound) with the CST TISEC device and any resulting device effects. This understanding will be useful for verifying the accuracy of the environmental effects predictions made during the environmental assessment and will inform future monitoring plans.

To monitor animal behavior in the near-field area of the CST TISEC device, a Gemini multibeam imaging sonar (manufactured by Tritech Ltd) was mounted on the turbine’s subsea base structure (Figure 1b). The intent was to view the area of the water column directly aligned with the turbine rotor and to determine the seasonal frequency of occurrence, and behavior, of targets that could be marine life within the near-field environment. However, an error in mounting the instrument resulted in a view comprised of near-bottom water and the sea floor rather than the water column at turbine rotor height. This limited the applicability of the data for assessing animal behavior in relation to the turbine. Regardless, the data collected over the course of the deployment period were biologically relevant and useful for assessing general trends in target abundance and movements and also allowed a better understanding of the performance and potential of the Gemini sonar. In addition, this work will provide a foundation for developing and refining the methodologies to be implemented for future Gemini data collection, processing, analysis, presentation and interpretation.

1.3 Gemini multibeam sonar The Tritech Gemini imaging sonar is an active acoustic monitoring device. It is a high frequency multi-beam sonar that uses reflected sound to build up a picture of the underwater environment. The sonar sends out a ‘ping’ (an acoustic pulse), and the intensity of the echoes received from the multiple beams are used to create an image of the sampled volume. The sonar pings multiple times per second, resulting in a video-like record of the sampled volume. The size, shape, and movements of objects (targets) within the sampled volume can be extracted for identification (but may not be at species level) and behavior analyses.

Images created by high frequency sonars like the Gemini are low-resolution when compared with contemporary video technologies. However, unlike video cameras, multibeam sonars function without light and in high turbidity (cloudiness or haziness of water caused by suspended solids). This makes multibeam sonars a highly suitable tool for observing marine life in environments such as Minas Passage, where light penetration is limited and the water is highly turbulent and at times turbid.

CST Gemini Monitoring Report

3

1.4 Project objectives The CST EEMP includes use of a Gemini sonar to investigate the presence and behavior of marine animals in the near-field (images extend to 60 m) of the CST TISEC device. The specific EEMP objective was to determine the seasonal frequency of fish and marine mammals within the near-field environment of the turbine and to track their movements. Under this objective, the tasks included an assessment of the performance of the Gemini imaging sonar, analysis of the dataset collected while the device was deployed from November 2016 to April 2017, assessment of the data for trends in target presence and behavior, and development of a general set of methodologies for processing and analysing Gemini data for future deployments.

The types of analyses presented here can be used with data processed manually or via automation. All data used for this report were manually processed (i.e. human observations of Gemini video files), but future analyses will likely utilize automated data processing techniques as algorithms are further developed and validated (Jepp 2017, Appendix 1). Automated processing is likely to provide a larger suite of metrics than can be realistically extracted manually from the data (e.g., the position and size of a target in every frame in which it is detected, Jepp 2017).

The work described in this report used manually-processed Gemini data to: (1) describe trends in target abundance within the sampled volume, over short and long-time scales; (2) characterize target movement with respect to current speed and direction and stage of tide; (3) identify potential fish schools; and (3) develop recommendations for future data collection, processing, and analysis. Information gained from this study improves understanding of the presence and activities of marine animals at this site, provides methods for future data processing and analyses of Gemini data sets, and will inform future research and monitoring of the potential effects of TISEC devices.

2 Methods

2.1 Data collection A Gemini 720i multibeam imaging sonar was mounted on the CST device’s subsea base structure (Figure 2), facing into the current (upstream) during the ebb tide and facing downstream during the flood tide.

CST Gemini Monitoring Report

4

Figure 2. Volume sampled by the Gemini sonar. View from above includes a still frame from the footage, which shows one of the three feet of the sub-sea base (A), its acoustic shadow (B), and the sea floor (C).

The sonar was located approximately 4 meters above the seafloor and 6.6 m away from the face of the turbine, and it sampled a 20° by 120° swath of water using an array of 256 acoustic beams operating at 720 kHz (Jepp 2017). Due to the mounting orientation, the acoustic beam was angled horizontally and slightly downward (Figure 2b), thereby capturing more of the seafloor and less of the water column than what was intended. The volume sampled was therefore mainly

CST Gemini Monitoring Report

5

below the turbine, and the view was dominated by acoustic backscatter from the sea floor rather than from the water column.

The sampled volume spanned 120° in the horizontal direction and 20° in the vertical direction, and extended to a range of 60 m. The imaging sonar would have encompassed a volume on the order of 5.3·104 m3 if not intercepted by the bottom. The Gemini sampled this volume at a rate of approximately 11 frames per second, resulting in a 2D, video-like representation of the sampled swathe (Figure 2a). The angular resolution of the Gemini was approximately 1°, so resolution was highest near the instrument (approximately 9 cm at 5 m range) and decreased with range as the acoustic beams spread (approximately 1 m at 60 m range).

The sonar data collection period spanned 08 November 2016 to 13 April 2017. During this period, communication issues with the Gemini caused frequent interruptions in data transfer, resulting in very small data files and time gaps between adjacent frames (Figure 3). High-quality data files were available for only 40% of the entire collection period. Although most data gaps were on the order of a few seconds in length, this resulted in jumpy Gemini footage which hindered ease and completeness of target detection and tracking. Manual processing of the data allowed some subsampling flexibility and thus reduced the magnitude of this effect; for example, the 5-minute samples analyzed were sometimes shifted slightly to include better-quality data. For future long-term monitoring of the TISEC device, and for automated target tracking to be successful, communication issues that result in data gaps will need to be addressed. Uninterrupted data collection should achieve closer to 100% coverage of the periods of time when the Gemini is operating, whether that be continuously or on a duty cycle (e.g., several minutes per hour).

Figure 3. Gemini data collection from 08 November 2016 to 13 April 2017. Gray indicates periods of uninterrupted data collection. Days used in analysis are highlighted in black on the horizontal axis.

2.2 Data processing Manual data processing was undertaken using Tritech’s Gemini SeaTec software (version 2.01.04.01), and required a human observer to play the footage and search for moving objects (targets) that could be marine animals. Efforts are ongoing to adapt Tritech’s target tracking algorithm for use on small targets (e.g. fish; Appendix 1, Jepp 2017). The current version of automated target tracking software was not used for this assessment due to gaps in the data files and backscatter interference from the sea floor. Though movement filters can help remove a relatively stationary backdrop, the reflection from the seafloor was bright enough to wash out the weaker signals from targets above it. Validation of the automated tracking algorithms that are under development will proceed during the next turbine deployment, when the orientation of the Gemini is adjusted to view the water column at turbine rotor height.

CST Gemini Monitoring Report

6

2.2.1 Target identification

The Gemini sonar can resolve shapes and track their movement, but cannot determine the identity (marine life or debris) of the acoustic targets. Acoustic targets are defined as objects moving independently of the background that could possibly be marine life. Many targets showed some degree of directed movement, which is expected of marine vertebrates, but it is possible that some targets were debris moving with the flow. Targets were most easily seen while using a high-persistence filter (see Appendix 1 for more information on the data processing software and settings). This filter made the path taken by a target more obvious against the background (Figure 4). Targets were easiest to see when within approximately 10 m of the Gemini. Over this distance, Gemini image resolution was high and the view was relatively uncontaminated by the sea floor. Where backscatter from the bottom was very strong, targets were difficult to see, and small targets (such as fish) were furthermore less likely to be detected at greater ranges due to decreasing resolution with distance from the sonar.

Figure 4. Example of an identified target in Gemini video. Target track indicated by white arrows in the larger Gemini view and in the inset magnification window. Start and end locations of the track indicated in the inset.

Data recorded for each target detected included the time at which it became visible and then no longer visible, the corresponding x and y coordinates, and any observations of unusual appearance or behavior. All targets were also measured (sized) using the click-and-drag measurement tool available in the Gemini SeaTec software. Measurement accuracy was limited by the resolution of the Gemini as well as the resolution of the viewing screen. Because measurements could only be obtained in the main viewing window, not the magnification window, targets within a few meters of the sonar were difficult to measure. For this reason, targets were assigned coarse size categories: < 0.5 m, 0.5 to 1.0 m, and > 1.0 m. If a target appeared to be an aggregation of smaller objects (e.g., a school of fish; Figure 5), it was recorded as such and its longest dimension was measured (Appendix 1).

CST Gemini Monitoring Report

7

Figure 5. Example of a target that may be a fish school. Note that targets are more clearly seen when their movements in the video clip are observed.

The human observer who processed the Gemini data for this study was trained on a separate data subset prior to beginning work on this project (see Appendix 1). This training period was essential for the observer to become comfortable with the data and the software and to be able to consistently judge what constituted a target and what did not. In addition, observer precision was tested at the end of this study by re-processing ten of the samples previously examined over a 2-month processing period. Target detection data from initial and repeated sample processing showed strong matching of results (88% similar on average). There was also no effect of time (since initial processing of the samples) on the percent target match. This indicated no drift in observer bias over time, confirming that sufficient training had taken place before processing of data samples began.

2.2.2 Data subsampling

Manual target detection is extremely time consuming, and a single 5-minute span of data can take a human observer up to 40 minutes to process, depending on the number of targets present. For this reason, a subsampling program was designed to examine a subset of the nearly 5 months of data collected. Five minutes of every 2-hour period were processed for one entire day (midnight to midnight) from each week of data collected. This subsampling regime was chosen to allow the characterization of longer-term (e.g. lunar or seasonal) trends in target abundance, as well as the observation of any shorter-term patterns (e.g., diel or tidal) occurring within each sampled day while avoiding signal aliasing (Viehman 2017). Some gaps in data collection resulted in some missing samples (Table 1). A total of 268 5-minute samples was processed for this study.

CST Gemini Monitoring Report

8

Table 1. Hours for which 5-minute samples were analyzed for each day of data collection. Blank spaces indicate data gaps.

Date Samples analyzed (hour of day, UTC) 0000 0200 0400 0600 0800 1000 1200 1400 1600 1800 2000 2200 2400

17 Nov 2016 24 Nov 2016 01 Dec 2016 08 Dec 2016 15 Dec 2016 22 Dec 2016 29 Dec 2016 05 Jan 2017 12 Jan 2017 19 Jan 2017 26 Jan 2017 02 Feb 2017 09 Feb 2017 16 Feb 2017 23 Feb 2017 02 Mar 2017 09 Mar 2017 16 Mar 2017 23 Mar 2017 31 Mar 2017 06 Apr 2017 13 Apr 2017

2.2.3 Auxiliary data

One-minute averages of current speed (normalized to the maximum) and current direction were modeled by CST using data from Acoustic Doppler Current Profilers (ADCPs) deployed on the CST turbine. The normalized current speed and direction data were used to classify each target as occurring during flood, ebb, or slack tide and for assessing the movement direction of targets relative to the modeled flow.

Water temperature data for the collection period were also available from a temperature and depth logger deployed at a Minas Passage site near the FORCE visitor center (www.oceannetworks.ca/observatories/atlantic/bay-fundy-minas-passage). This temperature logger is located close to shore and may therefore record temperatures slightly warmer or colder than water at the CST TISEC device location.

2.3 Data analysis 2.3.1 Temporal trends in target detections

Target abundance was calculated for each 5-minute data sample that was manually processed. Based on start and end times, these 5-minute samples were then each assigned tidal stage (ebb, flood, low slack, or high slack), diel stage (day or night), temperature, normalized current speed, and average current direction.

CST Gemini Monitoring Report

9

Data from the 5-minute samples were grouped by day to assess trends occurring over the length of the data collection period (e.g., seasonal changes). To assess shorter-term changes, such as differences in day and night target abundance and how these changed over the study period, the entire sample set was divided into three groups, each spanning a third of the collection period (i.e., late fall/early winter, mid-winter, late winter/early spring). Statistical tests were used where applicable and included chi squared and ANOVA tests with 5% confidence levels.

2.3.2 Target movement

The movement direction of each target was calculated using the target’s start and end position in the beam. Movement directions were presented graphically to assess differences related to current speed and direction, tidal stage, and diel stage. Directional movements of marine wildlife are important to understand. At other tidal energy sites, fish have been found to generally move with the current, with more random movement occurring at slack tides (Viehman and Zydlewski 2015, Viehman and Zydlewski 2017). Greater variability in target movement may therefore be a useful indicator of unusual behavior, such as responses to a tidal energy device. The spatial dependence of movement variability was explored by splitting the sampled swathe into a grid. Due to the low sample sizes in the subsampled dataset, a coarse grid with 8.3 m x 5 m cells was used. The circular variance of target movement directions was calculated for each cell, with 0 variance indicating uniform movement and 1 indicating completely random movement.

3 Results and Discussion

3.1 Trends in target abundance and size A total of 2,056 targets were detected within the subsampled dataset (268 5-minute samples), with 45 targets identified as potential schools of fish (Table 2). These numbers need to be interpreted keeping in mind that the sampled volume of water, over each 5-minute sample duration, varies with current conditions. For example, many more targets per sample were detected during the flowing tide (ebb or flood) than slack tides, probably due to the greater volume of water that is sampled when the current is moving through the beam than when the water it is relatively still. Similarly, the period of data collection occurred during the late fall to early spring, when nights were longer than daytime periods. Thus, more samples reflect nighttime conditions, somewhat inflating the total number of fish detected at night compared to day (Table 2). When the counts are normalized for the number of samples within each diel and tidal stage category (Table 3), the diel stage difference is not as large.

Table 2. Number of individual targets and (number of schools) detected in all processed data samples.

Diel stage Tidal stage

Total Ebb Flood High slack Low slack Day 476 (14) 253 (6) 3 (0) 1 (0) 733 (20) Night 843 (11) 395 (14) 21 (0) 19 (0) 1,278 (25) Total 1,319 (25) 648 (20) 24 (0) 20 (0) 2,011 (45)

Table 3. Number of individual targets and (number of schools) detected, normalized by number of samples within each category.

CST Gemini Monitoring Report

10

Diel stage Tidal stage

Total Ebb Flood High slack Low slack Day 9.3 (0.3) 4.7 (0.1) 0.8 (0) 0.2 (0) 15.0 (0.4) Night 12.4 (0.2) 5.5 (0.2) 3.0 (0) 2.4 (0) 23.3 (0.4) Total 21.7 (0.5) 10.2 (0.3) 3.8 (0) 2.6 (0) 38.3 (0.8)

Although flood tide currents at the deployment site were approximately 30% faster than ebb tide currents (Figure 7), there were fewer targets detected during flood than ebb (Tables 2 and 3). There are several potential explanations. For example, water velocity downstream of a turbine operating at maximum efficiency could be reduced to as little as one third of the upstream velocity (Figure 8), which would reduce the flood-tide sampled volume to approximately half that of the ebb. Even with no change in target behavior or concentration, this reduction in speed could halve the number of targets observed during the flood tide compared to ebb, which is consistent with the above results (Table 2, Table 3). With future Gemini datasets, the dependency of target numbers on volume sampled over time should be explored in more detail to inform the interpretation of results. To improve estimates of sampled volume, more information is required on the effects of the TISEC device on the local flow field, either through validated models or concurrent measurements of the flow within the volume sampled by the Gemini.

Figure 7. Example of current speed and direction data. Points are one-minute averages, derived by CST from a model informed by ADCP data collected at the device.

CST Gemini Monitoring Report

11

Figure 8. Current speed through a generic in-stream turbine that is operating at the Betz limit (maximum efficiency). The area swept by the blades of the turbine is shaded gray. Blue lines schematically indicate streamlines

that enclose the flow through the turbine. The far upstream current is u. At the turbine, current is two-thirds u and downstream it is one-third u. Source: Brian Sanderson.

The observed tidal stage differences (ebb vs. flood) in target presence in near-bottom waters close to the turbine could also be related to animal distribution patterns and behavior. For example, we might expect fewer targets to be detected on the flood tide than ebb if animals move higher in the water column during the flood. Alternately, asymmetric flow could carry animals through different parts of the passage during ebb and flood tides. While examination of broad-scale distribution patterns are beyond the scope of the CST EEMP, relevant data for interpretation purposes is available via mobile and stationary active acoustic fish surveys, which are conducted in Minas Passage at regular intervals by FORCE.

Few targets were detected beyond the first 10 m of the sampled volume (Figure 9), which corresponds to where the interference from the seafloor became obvious in the view. Most of the detected targets occurred within the first 5 m; a dip in numbers at exactly 5 m is likely related to sound reflected by the foot of the subsea base (Figure 2, A). It should be noted that many more targets, albeit small (e.g., < 20 cm in length), appeared to be present in the first 5 m of the Gemini’s view, but could not be included in the results here because they could not be accurately placed or measured with the manual tools available in the Gemini SeaTec software (Appendix 1). This data processing limitation could be overcome by improvements to the software (e.g., being able to make measurements in the magnification window) and/or automated tracking.

CST Gemini Monitoring Report

12

Figure 9. Number of targets detected vs. range from the Gemini sonar.

The number of targets detected in each grid square (Figure 10) showed the same range dependency, with most targets detected within 10 m of the sonar. The lack of targets at greater ranges is probably related to interference from the seafloor and the decreasing resolution of the Gemini—not necessarily to more targets being present near the turbine. To verify this, when the Gemini is reoriented with a more upward view of the water column, it would be beneficial to calculate the detection probability of different sized targets at various ranges (up to 60 m) from the Gemini, perhaps with field measurements of known objects, and compare this with the spatial distribution of targets detected within the beam.

3.1.1 Seasonal trend in target size

The number of targets in each size category did not vary significantly over time (Figure 11; chi-square p-value > 0.05). Most targets were in the smallest category (97.0% were < 0.5 m), with comparably very few in the larger categories (1.2% were 0.5 – 1.0 m, and 1.8% were > 1.0 m). This suggests that most targets present were fish. Some of the larger targets (>0.5 m) may have been striped bass, which are known to overwinter in Minas Passage and adjacent waters (Keyser et al. 2016). Marine mammals known to occupy the Minas Passage include harbour porpoise, harbour seals, and occasionally white-sided dolphins, most of which are over a meter in length. Small schooling fish like Atlantic herring are likely to occupy the Minas Passage during the winter (Melvin and Cochrane 2014; Viehman et al. 2017). Rainbow smelt and other fishes <0.5 m in length may also be present during the late fall to early spring period (Dadswell 2010).

CST Gemini Monitoring Report

13

Figure 10. Spatial distribution of detected targets. Color indicates number of targets detected per grid cell (8.3 m wide by 5 m high).

Figure 11. Target size distribution over time.

CST Gemini Monitoring Report

14

More accurate target size measurements would be useful in estimating which types of animals (and possibly species in some cases) are present and how their distribution may change over time. This is potentially something that can be improved with automated tracking algorithms, as well as small changes to the software, such as being able to make measurements in the magnification window (see Figure 4).

3.1.2 Seasonal trend in target abundance

The number of targets detected per 5-minute sample decreased over the 5 months of data collection, from 10-15 targets per sample in December to 0-5 targets per sample in April (Figure 12). This trend closely mirrored the decreasing temperature, and could reflect a general decrease in the abundance of fish in the area over the winter, preceding the return of many migratory species in the spring (e.g., river herring). Few fish surveys have occurred in this region in the coldest months of winter, so it is difficult to say with certainty if fish were less abundant then, or if they were simply inhabiting a different part of the water column or Passage. However, this trend in declining fish abundance in winter has been recorded by an ongoing before-after-control-impact acoustic study of fish at the FORCE site (FORCE 2017), which found slightly higher fish densities in December surveys compared to January and March (Daroux and Zydlewski 2017). Another acoustic study at this site found fish density to be higher in early December to early January compared to June to July (Viehman et al. 2017). Acoustically tagged striped bass have been found to be present in Minas Passage through December but scarcer as temperatures drop below 1° C (Keyser et al. 2016). Year-round passive acoustic monitoring of harbour porpoise indicates that their abundance is lowest during winter (Porskamp 2015).

Figure 12. Number of targets detected from November 2016 to April 2017. Points are mean number of targets per 5-minute sample, whiskers represent +/- one standard deviation. Water temperature is shown in blue.

CST Gemini Monitoring Report

15

3.1.3 Diel and tidal trends in target abundance

There was a large, statistically significant difference in the average rate of target detection per 5-minute sample between ebb and flood tide over the duration of the dataset (Figure 13a), but little difference between day and night (Figure 13b). The higher detection rate of targets during ebb tide reflects the absolute target numbers reported in Table 2. As discussed previously, lower target abundance during flood conditions could be due to the effects of the TISEC device structures on the near-field flow conditions (Figure 8), and/or to differences in animal behavior or distribution related to current direction (e.g., asymmetric flow dynamics in the Passage or changing depth preferences of animals).

Figure 13. Number of targets per 5-minute sample during (a) ebb and flood tides and (b) day and night. Points are means, error bars show +/- 1 standard error for each third of the sampling period, as per Section 2.3.1.

The small difference between day and night detection rates compared to the absolute target counts discussed earlier highlights the influence that environmental conditions and data partitioning can have on results. While the total number of targets detected at night was much higher than during the day (Table 2), the number of targets per 5-minute sample (which normalizes for time) indicated a much smaller effect of diel stage on abundance in the portion of water column sampled by the Gemini (Figure 13b).

The tidal and diel differences in target detection rate contrast some results from other acoustic studies conducted in the Minas Passage. However, this study includes only the acoustic detection of targets within approximately 4 m of the sea floor; most other active acoustic surveys of this region omitted the upper- and lower-most layers of the water column due to acoustic interference from the surface and seafloor, respectively. One of those previous studies used an upward-facing echosounder, mounted to a FORCE sensor platform on the seafloor, to examine the vertical distribution of fish during December 2015 to January 2016 (Viehman et al. 2017). That study, which omitted the lowest 3 m of water column, found the vertical distribution of fish in December to January changed noticeably from day to night. However, most of that change

a. b.

CST Gemini Monitoring Report

16

occurred in the upper 10-20 m of the water column; target abundance along the sea floor could have remained relatively unchanged over the course of a day, as was seen here.

Interestingly, Viehman et al (2017) did not observe any noticeable difference in fish vertical distribution or overall density between ebb and flood tides, which contrasts the very obvious difference observed here in waters just above the sea floor. It is possible that animal activity close to the sea floor is not strongly connected to activity in the mid-water-column, which is a very different hydrodynamic environment. The two different studies may also not be comparable due to spatial or temporal variability in the area, neither of which is well understood at this or other tidal energy sites. More datasets collected simultaneously with a variety of acoustic methods would facilitate the merging of results from different portions of water column and different locations within the Passage, leading to a better understanding of fish distribution patterns.

Another difference between this study and previous work in Minas Passage was the presence of the TISEC device. This could certainly contribute to the tidal difference in target detection rates, as the targets detected during ebb tide would have been approaching the device and those detected during flood tide would have been departing from it, potentially at much lower speeds. Additionally, the sampled volume during flood tide is within the device wake, which is likely more turbulent than the water sampled during the ebb. Targets traveling in more turbulent flow would have more variable echo strength as their orientation changes, and they could spend less time travelling laterally within the beam, both of which could lower their probability of detection. Near-field avoidance of the device structure could also result in fewer target detections downstream (flood tide) than upstream (ebb tide), but whether avoidance of the structure would occur along the sea floor cannot yet be determined.

The observed diel and tidal differences in target detection rates were consistent across the sampling period, indicating no major shift in target responses to either. This contrasts previous studies, which found the vertical distribution of fish to change with tidal and diel stages depending on the time of year and species present (Viehman et al. 2017, Melvin and Cochrane 2014, Viehman and Zydlewski 2017, Viehman et al. 2015, Daroux and Zydlewski 2017). More information is needed on the species present over the course of the winter to know if the lack of change observed here is due to a more consistent animal assemblage (and therefore consistent responses to environmental factors) or to the portion of water column observed (which has been omitted from most previous studies). It is possible the animal assemblage did not change much from November to April, as major migrations through Minas Passage occur primarily in the spring (April through May) and fall (September to October) (Baker et al. 2014, Dadswell 2010, Redden et al. 2014). Alternately, if animal movement is primarily governed by the strong currents, shifts in species assemblage structure may not be reflected in observed detection rates.

3.1.4 Target abundance in relation to current speed

Interestingly, target detection rates were higher during mid-range current speeds than at low or high speeds (Figure 14). This was especially evident for the ebb tide. The increase in detection rate from low to mid-range speeds during ebb tide could be due to the increased volume of water

CST Gemini Monitoring Report

17

sampled. However, this relationship does not hold for higher speeds, which could indicate that either the detection probability of targets decreases at the fastest speeds (e.g., they get harder to see), or targets are less abundant at this depth and/or location at higher current speeds (e.g., flow-related changes to animal vertical or horizontal distribution could mean they are outside of the sampled volume at peak flows). Note that flood tide current speeds, as modelled and shown in Figure 14, are representative of the water column at rotor height and not the portion of the water column sampled by the sonar, where flow is likely reduced by the presence of the CST TISEC device.

Figure 14. Targets per 5-minute sample at different current speeds during ebb (blue) and flood (red) tides. Points are mean targets per sample for each speed category, whiskers are 1 standard error.

Normalized current speed data for turbine rotor height, courtesy of OpenHydro.

3.1.5 Seasonal trend in size of target aggregates (potential schools)

Forty-five large targets were detected that appeared to be aggregations of smaller targets, (i.e. possibly schools of fish; Table 2). The sizes of these targets were on the order of 1 to 3 m across (Figure 15), and appeared to increase in size slightly toward the end of the sampling period, though the sample size was not large enough for statistical assessment. The frequency of these school-like targets decreased over the course of the sampling period (Figure 15). Fish in schools have previously been found to react to a test tidal turbine at slightly greater ranges than individual fish (Viehman and Zydlewski 2015). As monitoring continues, the behaviors of individuals and schools should be compared.

CST Gemini Monitoring Report

18

Figure 15. School size over time. Horizontal lines indicate the median, boxes span the 25th to 75th percentile, whiskers extend to 1.5 times the interquartile range, and points are outliers.

3.2 Target movement and direction 3.2.1 Individual targets

The start and end locations of each target were used to calculate each target’s net direction of movement (Figure 16). It should be noted that this method does not take into account changes in direction between start and end points, and may not reflect a target’s final direction if the path was not direct. As automated tracking algorithms are improved, parameters such as track tortuosity (i.e., how much the tracks twist and turn) can be calculated and potentially used as metrics of target behaviour, or as a method to separate passively drifting debris from actively moving organisms. These techniques will be applied to future Gemini datasets that span a large fraction of the water column and a range of 60 m, thus offering better opportunities to determine the nature of the target, and to assess animal avoidance and evasion behaviours.

CST Gemini Monitoring Report

19

Figure 16. Examples of target movement. Shown are 100 randomly sampled targets detected during (a) ebb and (b) flood tide. A random subset was used because plotting of all targets makes the paths impossible to see. Note

that data available for this study is limited to targets detected within 10 m of the sonar.

The majority of target movement was in the general direction of the modelled flow (Figure 17), as has been observed at other tidal power sites (Hammar et al. 2013, Viehman and Zydlewski 2015, Viehman 2017). However, there were some notable deviations from the modelled current direction. First, during the ebb tide, a large proportion of targets moved at an angle slightly offset from the modelled flow (approximately 5° to 20° counter-clock-wise, Figure 17a). It is possible that this reflects the deflection of flow around the device, or a behavioral response of animals as they approached the structure. During the flood tide, there was noticeably more variation in target movement (Figure 17b). This variation was highest closest to the CST device, directly downstream of the turbine structure (Figure 18b), but this was not seen during the ebb tide, during which variation was relatively constant regardless of location (Figure 18a). This difference could also be explained by changes to the flow field caused by the device structure. During the flood tide, most of the sampled area would have been directly in the wake of the device. The higher turbulence within this wake could cause targets moving with the flow to exhibit greater variation in movement than they would in the relatively uninterrupted flow upstream. Fish have been observed milling in the wake of a test turbine at another location (Viehman and Zydlewski 2015), and it is also possible this was occurring downstream of this device. More fine-scale information on the flow field up- and down-stream of the device, either measured or modelled, would improve our ability to determine whether observed movements were due to the passive movement of targets with the flow or to active behavioral responses of targets to the device or flow field. Automated tracking could also provide further insight on passively vs. actively moving targets by introducing more in-depth target parameters, such as track tortuosity and variation in echo intensity.

CST Gemini Monitoring Report

20

Figure 17. Directions traveled by targets during ebb (a) and flood (b) tides. Results are shown for day (white) and night (gray). Bars indicate the proportion of targets moving in each direction, and the arrow indicates the mean

modelled flow direction for each tide.

Figure 18. Variation in movement direction. The variance of movement direction for each beam grid cell was calculated for (a) ebb tide and (b) flood tide. Variance of 0 indicates unidirectional movement, while variance of 1

indicates random movement.

There was no noticeable difference in target movement direction between day and night (Figure 17). However, when the Gemini is reoriented and viewing the water column at turbine rotor height, comparisons between day and night may detect turbine effects on movement (targets may be more likely to react to the structure when it is visible; Viehman and Zydlewski 2015).

CST Gemini Monitoring Report

21

3.2.2 School-like targets

No general trends in movement could be determined for school-like targets (aggregations of small targets), as there were few detected during each tidal stage (Table 2). There were however visually interesting differences in movement paths (Figure 19). For example, target aggregates approaching on the ebbing tide appeared to move toward the beam center (Figure 19a), while those departing on the flood tide (Figure 19b) displayed generally uniform movement compared to that of other targets (Figure 16b). Fish schools have been found to respond differently to threats and obstacles than individuals—e.g., reacting farther away from a test tidal turbine (Viehman and Zydlewski 2015). The differences in behavior between individual targets and aggregated targets should be assessed in the future, as they may respond differently to tidal energy devices.

Figure 19. Movement paths of detected aggregates (possibly schools) during (a) ebb and (b) flood tides.

4 Conclusions

Biologically relevant trends were apparent during this study. The decrease in target abundance with decreasing winter temperatures is consistent with other biological surveys of the area. The differences observed between day and night and ebb and flood tide were remarkably consistent over time and raise interesting questions about animal activity near the sea floor as opposed to mid-water-column in high-flow tidal channels.

Target movement direction, though calculated using only start and end target locations, also exhibited patterns that aligned well with the physical environment. Differences in movement direction during ebb and flood tides and across a range of current speeds indicated a relationship with water flow, which is to be expected in currents of this magnitude. These data may have also indicated an effect of turbulence downstream of the CST device, during flood tide, on target movement. More information on the fine-scale hydrodynamics around the device could help determine if this is the case.

CST Gemini Monitoring Report

22

While there remain many improvements that should be made to the data processing and analysis methods (e.g., automated processing techniques), the results presented here, albeit for a lower than anticipated portion of the water column, are encouraging for the future use of the Gemini sonar to monitor animal presence and behavior in the near-field of the CST TISEC device.

The potential for an overall improved dataset from a re-oriented sonar device and a longer (planned) deployment of a turbine will provide an opportunity to obtain data with increased spatial and temporal coverage. This will help to clarify the Gemini results discussed in this report, improve understanding of year-round presence and spatial distributions of marine animals, and therefore help to meet the overall objective of understanding how fish and marine mammals might interact with the in-stream turbine.

5 Recommendations

Aspects of data collection, processing, and analysis that can be improved in future applications are described below.

5.1.1 Data collection

To obtain high-quality data amenable to automated processing, care must be taken to orient the Gemini in a way that provides a view of only the water column (i.e., removal of the sea floor from the view) and that reduces any interference from solid structures. For the next deployment, the re-orientation of the sonar will be confirmed through a number of commissioning tests prior to deployment, while the turbine is in port and partially submerged. The influence of turbine structures on the view will also be considered at this time since reflections from objects like the foot of the subsea base can cast acoustic ‘shadows’ or wash out the signal from targets occurring at the same range (Figure 2a).

This study highlights the importance of simultaneous collection of biological and physical data. The modelled current speed and direction data were useful for examining general trends, but high-resolution current speed and direction information, up- and down-stream of the turbine, would help with separating active and passive target behaviors. Communication with the sonar must be high-quality and consistent, which can be difficult when multiple instruments are communicating with shore via the same cable. In this case, communications interruptions resulted in numerous small gaps throughout the dataset that made target detection and tracking more difficult, and these gaps would greatly complicate automated tracking in the future. To address this, all instruments will be run together and tested while the turbine is still in port (i.e. prior to deployment) to ensure that the new cabling and set-up provides improved data acquisition.

If multiple instruments cannot function adequately at the same time, they could be integrated (e.g., alternate pings) or potentially duty-cycled. The results of the present study indicate diel, tidal, and seasonal differences in target abundance and movement, congruent with other studies at tidal power sites (e.g., Viehman and Zydlewski 2017). Any non-continuous data collection (duty cycling) should ensure that sampling occurs often enough to characterize changes related to short-term cycles, such as the tide, but over enough of a time span to capture longer-term

CST Gemini Monitoring Report

23

changes, such as those related to the seasonal cycle. For example, sampling several minutes of every half hour may be sufficient to characterize behavioral responses over the course of a tidal stage, and it is possible that not every day of the year will need to be sampled to capture seasonal differences. However, the most appropriate subsampling routine for the Gemini sonar mounted on the CST device should be determined based on a high-quality, continuous dataset.

5.1.2 Data processing

Validated, automated processing methods are needed for the Gemini sonar if it is to be used for monitoring purposes. It can take a highly-trained observer up to 40 minutes to extract even the basic metrics used above from a 5-minute Gemini data file. While this time may be reduced with a cleaner dataset, manual data processing is simply too labor- and time-intensive for use in long-term monitoring. Moreover, an automated tracking algorithm could export many more potentially useful metrics for target behavior analysis, including frame-by-frame location within the beam, size, and echo strength. Gemini data processing has been automated for large targets such as seals and other marine mammals (Jepp 2017), but the methods for detecting and tracking smaller targets, such as fish, are still in progress (Jepp 2017, Appendix 1). The development of suitable processing algorithms will be facilitated when the Gemini’s view is reoriented to cover only the water column.

A certain amount of manual processing will be necessary to validate the results of an automated system, to quantify its error rate relative to a human observer, and to ensure its continued functionality over time. Manual data processing with the Gemini SeaTec software can be improved substantially by allowing measurements to be taken in the magnification window. This would allow many more of the small targets within the first 10 m of the Gemini to be located and measured for inclusion in the dataset. It is essential that the human observer be trained on a data subset prior to processing Gemini data for use in monitoring or validating automated detections. Afterward, an observer’s precision and consistency should be reassessed periodically with previously examined data subsets.

5.1.3 Data analysis and interpretation

This report explored some of the ways Gemini data can be partitioned and displayed. To detect turbine effects, it will be important to continue to assess temporal and spatial variation in metrics extracted from Gemini data. Combining these data with information on the physical environment (e.g., wake characteristics) will improve identification of active and passive target behaviors and which of those may be responses to the device. Due to the sonar orientation error prior to the November 2016 deployment, almost all targets identified were within 10 m of the device. With a correctly oriented sonar, future assessments will include tracking of target movements over a wider range of distances from the device. Additionally, with more targets detected throughout the sampled volume, a finer grid could be applied to generate summary statistics and images that are more useful for turbine effect assessment.

As automated processing methods are implemented, new metrics will become available for use in behavioural analyses. For example, obtaining target location across multiple frames, rather than just the starting and ending positions, would allow metrics such as tortuosity to be used, and

CST Gemini Monitoring Report

24

any substantial changes in a target’s path could be assessed as potential responses to the device. Frame-by-frame position, echo strength, and size is likely to be useful in separating passively drifting objects (debris) from those with directed movement (animals).

It will also be important to assess the detection probability of targets throughout the Gemini’s field of view under a range of environmental conditions—for example, at low to high current speeds. This is necessary for understanding potential sources of bias in the results.

All results obtained from the near-field of the CST device must be considered in the larger context of Minas Passage and the fish populations that utilize it. Mid- and far-field monitoring are well outside of the CST EEMP requirements, but information from broader-scale studies of the area will help with Gemini data interpretation. Spatial and temporal distribution of animals in the Passage are currently being characterized with various methods, including drifting hydrophones (Sanderson et al., in preparation), mobile active acoustic surveys (Daroux et al. 2017, FORCE 2017, Melvin and Cochrane 2014), and stationary active acoustic fish surveys (Viehman et al. 2017).

Overall, a better general understanding of the fish present in Minas Passage during all times of the year would help interpret any active acoustic data, whether from imaging sonars like the Gemini or from scientific echosounders (e.g. Viehman et al. 2017, Daroux and Zydlewski 2017, Melvin and Cochrane 2014). While not within the scope of the CST EEMP, this information could be acquired by the broader scientific community through physical sampling methods, such as trawling. Given that trawling can be very difficult and dangerous in fast tidal flows, a potential solution, first suggested by Keyser et al. (2016), would be to sample down-stream of the passage, at the start of Minas Channel or Minas Basin, just before slack tide. Then, any fish captured would likely have just been within the passage itself. Acoustically tagging and tracking a variety of fish species and life stages would also be helpful in determining their seasonal to year-round presence and spatial distribution. This has been done for Atlantic sturgeon (Stokesbury et al. 2017), striped bass (Keyser et al. 2016, Broome 2014), and to some extent for American eel (see Redden et al. 2014). Further tagging studies to track fish in Minas Passage are planned for 2018 (pers. comm., Mike Stokesbury, Acadia University), and will contribute to the knowledge base on fish movements through the FORCE test site.

CST Gemini Monitoring Report

25

5 References

Baker, M., M. Reed, and A.M. Redden, “Temporal Patterns in Minas Basin Intertidal Weir Fish Catches and Presence of Harbour Porpoise during April – August 2013.” ACER, Wolfville, NS, Tech. Rep. 120, 2014.

Benjamins, S., A. Dale, G. Hastie, J.J. Waggitt, M. Lea, B. Scott, B. Wilson, “Confusion reigns? A review of marine megafauna interactions with tidal-stream environments,” Oceanography and Marine Biology: An Annual Review, vol. 53, pp. 1-53, 2016a.

Benjamins, S., A. Dale, N. van Geel, B. Wilson, “Riding the tide: use of a moving tidal-stream habitat by harbour porpoises,” Mar. Ecol. Prog. Ser., vol. 549, pp. 275-288, 2016b.

Broome, J.E., “Population Characteristics of Striped Bass (Morone saxatilis, Walbaum, 1792) in Minas Basin and Patterns of Acoustically Detected Movements within Minas Passage”, MSc. thesis, Acadia University, Wolfville, NS, Canada. 2014. Thesis available at: http://openarchive.acadiau.ca/cdm/ref/collection/Theses/i d/1009 thesis.

Dadswell, M.J., “Occurrence and migration of fishes in Minas Passage and their potential for tidal turbine interaction,” BioIdentification Associates, Report prepared for Fundy Ocean Research Centre for Energy, 2010.

Daroux, A., G. Zydlewski, “Fish monitoring to assess effects of a turbine in a tidal energy development site,” paper presented at the 12th European Wave and Tidal Energy Conference, Cork, Ireland, 2017. Fraser, S., V. Nikora, B.J. Williamson, B.E. Scott, “Automatic active acoustic target detection in turbulent aquatic environments,” Limnol. Oceanogr.-Meth., vol. 15, pp. 184-190, 2017.

Fundy Ocean Research Center for Energy, “Environmental effects monitoring program quarterly report: January 1-March 31, 2017,” available at http://fundyforce.ca/environment/monitoring.

Hammar, L., S. Andersson, L. Eggertsen, J. Haglund, M. Gullström, Jimmy Ehnberg, Sverker Molander, “Hydrokinetic turbine effects on fish swimming behavior,” PLoS ONE, vol. 8(12), e84141, 2013.

Jacques, D.A., and J.K. Horne, “Scaling of spatial and temporal biological variability at marine renewable energy sites,” Proc. 2nd Marine Energy Technology Symposium, 2014.

Jepp, P., “Target tracking using sonars for marine life monitoring around tidal turbines,” in Proc. European Wave and Tidal Energy Conference, 2017.

Karsten, R., A. Swan, and J. Culina, “Assessment of arrays of in-stream tidal turbines in the Bay of Fundy,” Phil. Trans. R. Soc. A, vol. 371, 2013.

Keyser, F.M., J.E. Broome, R.G. Bradford, B. Sanderson, and A.M. Redden, “Winter presence and temperature-related diel vertical migration of Striped Bass (Morone saxatilis) in an extreme

CST Gemini Monitoring Report

26

high flow site in Minas Passage, Bay of Fundy,” Can. J. Fish. Aquat. Sci., vol. 73(12), pp. 1777-1786, 2016.

Melvin, G.D. and N.A. Cochrane, “Investigation of the vertical distribution, movement and abundance of fish in the vicinity of proposed tidal power energy conversion devices,” Final report to the Offshore Energy Research Association, OEER/OETR Research Project 300-170-09-12, 2014.

Porskamp, P., A.M. Redden, J. Broome, B. Sanderson and J. Wood. “Assessing marine mammal presence in and near the FORCE Lease Area during winter and early spring – addressing baseline data gaps and sensor performance,” Final report to the Offshore Energy Research Association and FORCE. ACER Technical Report No 121, Acadia University, Wolfville, NS. 35 p, 2015.

Redden, A.M., M.J.W. Stokesbury, J.E. Broome, F.M. Keyser, A.J.F. Gibson, E.A. Halfyard, M.F. McLean, R. Bradford, M.J. Dadswell, B. Sanderson and R. Karsten, “Acoustic tracking of fish movements in the Minas Passage and FORCE Demonstration Area: Pre-turbine Baseline Studies (2011-2013),” Final Report to the Offshore Energy Research Association of Nova Scotia and Fundy Ocean Research Centre for Energy, Acadia Centre for Estuarine Research Technical Report No. 118, Acadia University, Wolfville, NS. 153p, 2014.

Sanderson, B., C. Buhariwalla, M. Adams, J. Broome, M. Stokesbury, A. Redden, “Quantifying detection range of acoustic tags for probability of fish encountering MHK devices,” paper presented at the 12th European Wave and Tidal Energy Conference, Cork, Ireland, 2017.

Shen, H., G.B. Zydlewski, H.A. Viehman, G. Staines, “Estimating the probability of fish encountering a marine hydrokinetic device,” Renew. Energy, vol. 97, pp. 746-756, 2016.

Stokesbury, M.J.W., L.M. Logan-Chesney, M.F. McLean, C.F. Buhariwalla, A.M. Redden, J.W. Beardsall, J.E. Broome, M.J. Dadswell, “Atlantic sturgeon and temporal distribution in Minas Passage, Nova Scotia, Canada, a region of future tidal energy extraction,” PLoS ONE, vol. 11(7), e0158387, 2017.

Viehman, H., “Hydroacoustic analysis of the effects of a tidal power turbine on fishes,” PhD Dissertation, University of Maine, School of Marine Sciences, 2015.

Viehman, H., G.B. Zydlewski, “Fish interaction with a commercial-scale tidal energy device in a field setting,” Estuaries Coast., vol. 38(suppl. 1), pp. S241-S252, 2015.

Viehman, H.A., G.B. Zydlewski, “Multi-scale temporal patterns in fish presence in a high-velocity tidal channel,” PLoS ONE, 2017.

Viehman, H., G.B. Zydlewski, J. McCleave, and G. Staines, “Using acoustics to understand fish presence and vertical distribution in a tidally dynamic region targeted for energy extraction,” Estuar. Coast., vol. 38(suppl. 1), pp. S215-S226, 2015.

CST Gemini Monitoring Report

27

Viehman, H., T. Boucher, A. Redden, “Winter and summer differences in probability of fish encounter (spatial overlap) with MHK devices,” paper presented at the 12th European Wave and Tidal Energy Conference, Cork, Ireland, 2017.

Waggitt, J.J., A.M.C. Robbins, H.M. Wade, E.A. Masden, R.W. Furness, A.C. Jackson, B.E. Scott, “Comparative studies reveal variability in the use of tidal stream environments by seabirds,” Marine Policy, vol. 81, pp. 143-152, 2017.

Waggitt, J.J., P.W. Cazenave, R. Torres, B.J. Williamson, B.E. Scott, “Quantifying pursuit-diving seabirds’ associations with fine-scale physical features in tidal stream environments,” Journal of Applied Ecology, vol. 53(6), pp. 1653-1666, 2016a.

Waggitt, J.J., P.W. Cazenave, R. Torres, B.J. Williamson, B.E. Scott, “Predictable hydrodynamic variations in the density of benthic foraging seabirds in a tidal stream environment,” ICES Journal of Marine Science, vol. 73(10), pp. 2677-2686, 2016b.

Wiesebron, L.E., J.K. Horne, B.E. Scott, B.J. Williamson, “Comparing nekton distributions at two tidal energy sites suggests potential for generic environmental monitoring,” International Journal of Marine Energy, vol. 16, pp. 239-345, 2016a.

Wiesebron, L.E., J.K. Horne, A.N. Hendrix, “Characterizing biological impacts at marine renewable energy sites,” International Journal of Marine Energy, vol. 14, pp. 27-40, 2016b.Williamson, B.J., P. Blondel, E. Armstrong, P.S. Bell, C. Hall et al., “A self-contained subsea platform for acoustic monitoring of the environment around marine renewable energy devices–field deployments at wave and tidal energy sites in Orkney, Scotland,” IEEE Journal of Oceanic Engineering, vol. 41(1), pp. 67-81, 2015.

Williamson, B.J., S. Fraser, P. Blondel, P.S. Bell, J.J. Waggitt, B.E. Scott, “Multisensor acoustic tracking of fish and seabird behavior around tidal turbine structures in Scotland,” IEEE Journal of Oceanic Engineering, vol. 42(4), pp. 948-965, 2017.

CST Gemini Monitoring Report

28

Appendix 1

Gemini Imaging Software Target Identification

Preliminary Analysis Report

12 September 2017

Introduction The following document is a summary of a preliminary comparison of automated target tracking and human visual identification of possible targets using video recorded by the Gemini Multibeam Imaging Sonar installed on the Cape Sharp Tidal Turbine gravity base and deployed at the Fundy Ocean Research Center for Energy in November 2016. For this report, the Gemini Imaging Sonar Software was used to detect targets in the available Cape Sharp video files recorded on November 17th, 2016. Automated target detection was conducted by Tritech International Ltd. The following manual target detection and comparison with automated detection was carried out by Acadia University. This analysis was conducted to examine the performance characteristics of the automated target detection software, with the assumption being that the manual target detection would be an accurate and reliable reference point. The eventual goal of this work is to use automated and manual target detection comparisons to improve the automated target tracking algorithm.

Gemini Data The Gemini Multibeam Imaging Sonar was mounted on the Cape Sharp turbine gravity base at approximately 4 meters above the seafloor and facing east towards Minas Basin. The Gemini acoustic camera images a swath of water using an array of 256 acoustic beams at 720 kHz. With the Gemini oriented horizontally, it has a 120° spread in the horizontal dimension and a 20° spread in the vertical direction. In the data files examined, the first few meters of data show the water column, after which the sampled volume extends to the seafloor (Figure 1).

The video files were recorded at approximately 11 frames per second, which allowed the identification of potential targets (moving objects that could possibly be marine life). Figure 2 shows a typical example of an identified target moving toward the turbine (the target appears as a line, which traces its movement across the screen), although targets may vary in color, shape, and brightness depending on such variables as target size and distance to the Gemini Sonar. The individual video files reviewed were on average 307,400 KB in size and approximately five minutes long.

CST Gemini Monitoring Report

29

Figure 1. Gemini Imaging Software screenshot, with the first five meters shown in yellow brackets and the view that includes the seafloor in blue bracket. The first 20 meters are shown in the magnification

window overlaid on the right.

Figure 2. Example of an identified target in the Gemini video, labelled with a white arrow in the magnification window.

The semidiurnal tidal pattern on 17 November 2016, with two high and two low tides within 24 hours, is shown in Table 1. This day was chosen for an example because it was the day with the overall highest rate of detections, mostly false positives. Conversely, other days (e.g., November 21) were chosen because of the lowest rate of detections, which meant it was easier to examine each target individually.

CST Gemini Monitoring Report

30

These comparisons allow researchers to compare different days and understand the scope of the analysis and how it can vary.

Table 1. Tidal cycle at the FORCE tidal test site for November 17, 2016. (Fisheries and Oceans Canada, 2016)

Atlantic Standard Time (hh:mm)

Coordinate Universal Time (hh:mm)

Water level (m) Tidal Stage

02:05 06:05 12.9 high 08:21 12:21 0.3 low 14:26 18:26 13.1 high

The tidal cycle and direction of current help to explain specific behaviours exhibited by targets, such as moving away from the turbine (ebb) or towards the turbine (flood).

Software Settings The following settings were used when manually reviewing the video footage:

• Image Orientation: The image was inverted upwards, oriented to the left, and rotated upward (Figure 3). The most prominent feature on screen was a v-shaped structure, which was located at a range of (-7.50 m, 8.48 m).

Figure 3. Image orientation of Gemini video during analysis showing vertex of prominent

v-shaped structure (arrow).

CST Gemini Monitoring Report

31

• Gain: The gain, or brightness of the video on screen, was usually kept between 78% and 100% (Figure 4) depending on the background noise present, which showed up as ‘speckles’ on the screen. However, for some video files lower gain was necessary as the video image itself already appeared very bright, making it difficult to identify brighter targets. Higher gain proved to be more effective in the identification of smaller targets.

Figure 4. Comparison of Gemini video frame viewed at 78% gain (left) and 100% gain (right).

• Speed: The speed of the video replay was adjusted according to ease of viewing and the number of identifiable targets. When there was an increase in background noise or number of targets present, the speed was reduced to 90%.

• Averaging filter: Averaging (in the ‘Advanced’ tab, under filter settings) takes a weighted average over current and previous frames to smooth frame-to-frame variation, which can make moving targets easier to see against the stable background. Averaging was left at 50% because no improvement in the video quality could be observed when changing this setting.

• Persistence filter: The persistence can be used to highlight movement across frames. As the frames progress, the persistence filter retains a decreasing number of earlier frames as well as the current frame, creating a line that tracks a target’s movement over time. The persistence level was changed depending on the targets being identified. Extreme persist (99% and 99.5%) was the most effective when trying to track smaller, dimmer targets, as their track would persist longer, making it easier to pinpoint target position over time. Long persist (96% or 98%) was more suitable for larger and brighter targets, as they were easily identifiable without extreme persistence (Figure 5). The shorter persistence caused the tracks to disappear faster, which made it easier to move forward and backward frame by frame without the track remaining in view.

CST Gemini Monitoring Report

32

Figure 5. Comparison of the same target identified from the Gemini video with 96% (left) and 99.5%

(right) persistence, with magnification windows focused on target.

• Movement filter: The movement filter, which reduces signal from any objects that remain stationary from frame to frame, was kept at 60% to allow better identification of targets, but this filter level could be increased if background noise is low (Figure 6).

Figure 6. Comparison of video viewed with 100% movement filter (left) versus

0% movement filter (right).

• Target tracking: Target tracking was not enabled as it was not accurate enough for smaller targets (less than 0.5 meters). For this study, only manual target detection was used for tracking movements.

CST Gemini Monitoring Report

33

Target Identification The following procedure was used to identify possible targets (moving objects that may be marine animals). The automated tracking program was used by Pauline Jepp at Tritech to identify files with many potential targets, and these files were then reviewed to get an idea of what targets look like. Once accustomed to target identification, entire five-minute files were viewed. Each video file was watched twice, focusing on different ranges each time. First, the 0-15 m range was watched using 1.5x magnification to identify smaller targets (Figure 7), which were generally only visible at close range due to decreasing resolution with range in addition to bottom interference and ‘specks’ decreasing visibility on the screen.

Figure 7. Example of the larger view of the Gemini video, with the 1.5x magnification window (right)

used for viewing the first 15 m.

When watching a second time, the focus was on the 15-60 m range. The use of magnification was discontinued after examining three videos because often what could be seen on the magnification screen could not be identified on the larger view.

During viewing of the Gemini video, potential targets were first identified, then the video was rewound and watched frame by frame (by clicking the frame button). The target position and time was recorded for every second, or more if the target was either moving larger distances over the one second time frame or the track showed up within one second (Table 2).

CST Gemini Monitoring Report

34

Table 2. Example of Excel sheet created for targets manually detected in Gemini video. Columns are file number, date, time, position, notes, tidal stage, target size, matched automated detection, and the

numerical IDs of automated targets not detected manually.

Later, only the positions at the start and end of the target track were recorded. If a target could not be clearly identified (e.g., if there was increased background noise, or a broken up path), the uncertain targets were classified as either possible or probable targets.

Possible fish aggregations (Figure 8) were identified as a single unit, with a singular time stamp and estimate of the general range of the fish aggregation. Any unusual properties of a target (e.g. very bright and/or wide track) or any particular movements (e.g. moving away from the turbine) were also noted.

Comparison of automated and manual target detections Once all videos in which the automated tracking had identified targets were reviewed, automated detections were compared to those detected manually and matches were identified. A match was determined using the time of automatic detection and the range. Only 2 of the 135 targets identified through the automated tracking were matched to the manually detected targets in the videos reviewed to date. It should also be noted that the automated program identified fewer targets than detected by the human observer.

CST Gemini Monitoring Report

35

Figure 8. Example of a possible fish aggregation (white box).

Conclusion Overall, the quality of the video was sufficient to become comfortable with target identification. With experience, targets can be distinguished easily and confidently. One issue that was encountered was that the video did not run smoothly. Jumps of a second or more were common, making target identification and tracking more difficult. This was due to communications issues between the on-shore computer and the Gemini, which has been addressed for the future deployment (refer to Section 4.1.2 of the main Q3 Report). The automated tracking performance needs improvement, as substantially fewer targets were identified by the automated software than by the human observer, and even fewer could be matched with the manually detected targets.

There are a couple of changes that could be made to the software to make it more “user” friendly. It would be beneficial if the position of a target could be pinpointed more accurately. This could be done, for example, if one were able to pinpoint a target’s position directly on the magnification window instead of only in the larger view.

Next steps will include further examination of more of the collected Gemini data. This will provide more comparison material for the development of the automated target detection algorithms. Attempts will also be made to better identify the targets (e.g. as fish).

CST Gemini Monitoring Report

36

References Fisheries and Oceans Canada. (2016). 7 day tidal predictions: Cape Sharp (#250). Retrieved from http://www.tides.gc.ca/eng/station?type=0&date=2016%2F11%2F17&sid=250&tz=AST&pres=.

Tritech International Ltd. (n.d.). Gemini Imaging Sonar Product Manual: Document: 0685SOM-00001, Issue: 10. Retrieved from http://www.tritech.co.uk/media/support/manuals/geminiimaging-sonarproduct-manual.pdf.

Appendix D JASCO – Acoustic Data Analysis Interim Report

i

Acoustic Data Analysis of the OpenHydro Open Center Turbine at FORCE – Phase 1 Report

Authors: Bruce Martin Loren Horwich

31 January 2018

P001292-002 Document 01516 Version 1.1 DRAFT

JASCO Applied Sciences (Canada) Ltd Suite 202, 32 Troop Ave.

Dartmouth, NS B3B 1Z1 Canada Tel: +1-902-405-3336 Fax: +1-902-405-3337

www.jasco.com

i

Suggested citation:

Martin, B. and L. Horwich. 2017. Acoustic Data Analysis of the OpenHydro Open Center Turbine at FORCE – Phase 1 Report: . Document 01516, Version 1.1 DRAFT. Technical report by JASCO Applied Sciences for CST.

Contents EXECUTIVE SUMMARY ........................................................................................................ 2

1. INTRODUCTION .............................................................................................................. 4

2. CONTEXT ...................................................................................................................... 6

3. METHODOLOGY ............................................................................................................. 8 3.1. Recording Equipment ........................................................................................................................ 8

3.1.1. AMAR Autonomous Recordings ............................................................................................ 8 3.1.2. AMAR Autonomous Drifter ................................................................................................... 12

3.2. Recorder Calibrations ..................................................................................................................... 14 3.3. Acoustic Metrics .............................................................................................................................. 14 3.4. Vessel Detections ........................................................................................................................... 17

4. ACOUSTIC MEASUREMENTS SUMMARY .......................................................................... 18 4.1. Minas Passage Autonomous Recorder (Station 2) ......................................................................... 18 4.2. Outer Bay of Fundy Recorder ......................................................................................................... 19

5. INTERPRETING THE ACOUSTIC MEASUREMENTS USING THE TURBINE STATE AND CURRENT DATA .............................................................................................................................. 23

6. DISCUSSION ................................................................................................................ 27 6.1. Comparison to Effects of Noise Thresholds .................................................................................... 27

7. RECOMMENDATIONS .................................................................................................... 29 7.1. Recorder Methodology .................................................................................................................... 29

LITERATURE CITED .......................................................................................................... 30

Executive Summary Cape Sharp Tidal (CST) installed an OpenHydro Open-Centre instream tidal turbine on 7 Nov 2016, in Berth D at the Crown Lease Area at the Fundy Ocean Research Center for Energy (FORCE). JASCO Applied Sciences was contracted to make acoustic measurements of tidal turbine sound and analyse the data in an effort to characterize the spectral content of turbine operational sound in relation to ambient flow noise present in the Minas Passage and assess the methodology used for this study as a potential option for future studies in high flow environments.

Knowledge of turbine operational sound is important to increase understanding of the potential effects to marine wildlife from instream tidal turbines and guiding the development of mitigation measures. Since acoustic measurements of tidal turbine sound have so far proven problematic due to contamination from flow noise in high-flow environments, information on methodologies for collecting sound data is also important. Flow-noise contamination can be reduced by collecting acoustic data using drifting sound recorders, but this approach provides only brief measurements that preclude a strong statistical sampling of the emitted noise. Due to the exceptionally high current in the Minas Passage, hydrodynamic high-flow moorings (HFM) were used to evaluate the potential for these devices in future studies. Fixed acoustic recorders enable longer recordings, but the instrumentation is more susceptible to flow-noise contamination in comparison to drifting recorders.

JASCO deployed two acoustic recorders in the Minas Passage on 18 Nov 2016. The recorders were deployed on the seabed 100 m and 680 m from the CST Open-Centre turbine. The 680m recorder was intended to serve as a control measurement. The recorder at 100m was retrieved on 19 Jan 2017. The control has not yet been retrieved. To assist with the analysis, information on currents and turbine operation were incorporated. This included the normalized flow speed, current direction, and operating state of the turbine (i.e., not spinning, free spinning, and generating). The combination of information allows for an association between the measured sounds and the turbine operating state.

The primary objective of the study was to perform a Sound Source Characterization (SSC) of a full-size instream turbine operating in different states and at various flows to understand when and at what frequencies the turbine might be distinguishable from the ambient background noise. .

The JASCO analysis of turbine sound is on-going and will be further investigated in 2018. However, Phase 1 results indicate:

- Turbine sound was indistinguishable from flow noise below 60 Hz. The early analysis suggested that there was a band of frequencies in the 60-300 Hz range where the sound characteristics were likely from the turbine.

- The autonomous recorder 100 m from the turbine was able to distinguish turbine sounds from ambient noise for frequencies above 50 Hz, at all flow rates.

- The OpenHydro turbine operational sounds are different when free-spinning compared to generating. In the free-spinning state the turbine is louder than the background at frequencies of 50-500 Hz for flow speeds of 20-60% of the full flow. In the generating state the turbine is louder than the background at frequencies of 50-12500 Hz for flow speeds of 20-60% of full flow. The sound produced by the turbine while generating is dominated by two wide bands, one centered near 1250 Hz and the other centered near 4000 Hz. The 4000 Hz band was higher than background levels at all flow speeds.

- A preliminary comparison to the sound levels near the outer Bay of Fundy shipping lanes was performed. When the turbine was not spinning and currents were below 20% of full flow, the Minas Passage was 15-25 dB quieter in the band of 25-320 Hz, and 5-15 dB quieter for the band of 320-

10000 Hz. As the flow speed increased ambient noise levels in the Minas Passage also increased, at all frequencies. For frequencies below 100 Hz the increases were due to flow-noise over the recorder. Above 100 Hz the source of the noise was primarily sediment movement (i.e. pebbles and gravel striking each other). For current speeds above 80% of full flow, the levels measured in Minas Passage were 10-30 dB higher than the lower Bay of Fundy at all frequencies analyzed (10-12500 Hz).

- The threshold for possible temporary hearing loss, such as humans experience at loud music concerts, is measured for marine life using the daily sound exposure level of the sound weighted by the hearing frequency band of the species group studied (similar to the A-weighting for humans). The temporary hearing loss threshold for porpoise exposed to continuous sounds from human activity (153 dB re 1 µPa²·s weighted SEL) was exceeded on 52 of 63 days monitored by the autonomous recorder. The data analyzed to date show that most noise within the porpoise hearing range is from sediment movement, which agrees with previous results from Admiralty Inlet. Since the Passage is regular habitat for porpoise we speculate that their hearing is not affected, and by extension the temporary hearing loss threshold for continuous human generated sound may not be appropriate for noise from sediment movement.

We caution that the OpenHydro turbine was under engineering evaluation during the sound monitoring period. During this time the turbine control center (TCC), which is an electrical component sub-system attached to the subsea base and connected to the turbine, was being used to test and establish operating parameters and was involved with regulating the turbine operation from not-spinning, to free-spinning and then generating. The results are therefore indicative of a commissioning phase and not an operational phase. Further, we note that the next OpenHydro Open-Center turbine planned for deployment will be an improved version and that some of the sounds noted in this report will no longer be a concern. It is expected that major changes in how the turbine transitions from not-spinning to free-wheeling to generating will occur, and that further sound measurements will be required.

While more work would be required to enhance the acoustic recorder performance, the analysis indicates that this type of moored, long-term, static monitoring allows accurate characterization of the overall soundscape. Seasonal and tidal trends can be determined, and once the moorings are deployed, the devices can record throughout all weather conditions. The method of static moorings is the best method for measurement and analysis, but the other methods, such as drifting hydrophones, do add to the knowledge about the overall soundscape and should not be discounted. The second phase of this project will be conducted in Q1 2018. The focus the analysis will be converting the received sound level models into turbine source level models, then using acoustic propagation modeling to estimate that range around the platform that it exceeds the ambient noise levels.

1. Introduction Cape Sharp Tidal (CST) installed an OpenHydro Open-Centre instream tidal turbine on 7 Nov 2016, in the Crown Lease Area (the study area) (Figure 1) at the Fundy Ocean Research Center for Energy (FORCE) in the Minas Passage, Nova Scotia. As part of CST’s Environmental Effects Monitoring Program (EEMP) JASCO was contracted by CST to deploy two long-term acoustic recorders on 18 Nov 2016 to perform a Sound Source Characterization (SSC). The objective of the SSC was to measure underwater sound levels as a function of distance from the instream tidal turbine. The data would allow JASCO to characterize the sound from the turbine operation and to compare that sound to the ambient (natural) sound created by the environment.

The recorder 100 m from the turbine was recovered on 19 Jan 2017. The ‘control recorder’ at 680 m from the turbine has not yet been retrieved. A basic analysis of the total sound levels from the 100m recording was presented in the Q1 2017 Interim Report. The analysis suggested that there is a band of frequencies in the 60-300 Hz range where the sound characteristics were likely from the turbine. Without the control recorder data, or other information, no further conclusions were made.

Numerous other acoustic and environmental measurements have been made at the Minas Passage site whose data could help us understand the sounds from instream tidal devices. These include:

- icListen Smart Hydrophones on the OpenHydro turbine platform;

- JASCO AMAR drifting hydrophone measurements completed by FORCE at the Crown Lease Area on 18 and 20 October 2016;

- icListen drifting hydrophone measurements made by FORCE on 20 March 2017;

- Current speed and direction measurements made from the OpenHydro turbine platform; and

- Turbine state information logged by CST for the deployment period.

The autonomous recorders used in the CST study allow researchers to avoid the difficulties experienced with drifting hydrophones that, although they display reliability in high flow environments, will only provide a snapshot of the ambient noise and turbine sound as the instrument passes the turbine. Similarly, issues that may be experienced with a suspended hydrophone (e.g., loss of mooring in high flow environments, measurement of strumming noise, inability to measure ambient sound levels in high flow conditions etc.) are also avoided with the use of a bottom-mounted instrument. A bottom-mounted hydrophone is also favourable in high flow environments as the instrument can reduce flow noise due to water pressure since flow speeds at the bottom of water columns are generally lower than elsewhere in the column. This flow induced ‘noise’ can be further minimized by shielding the hydrophone. JASCO incorporates both considerations into their design, resulting in a streamlined design called the High-Flow (HF) Mooring. An important outcome of the current analysis is comparison of the benefits and limitations of each of the measurement methods.

Further contextual results can be obtained by comparing the Minas Passage sound levels to those measured near the Bay of Fundy shipping lanes. JASCO has a suitable five-month data set which is available for such a comparison. The sound levels should also be compared to recommended thresholds for the onset of behavioural disturbance or injury to marine life. For continuous sources such as a turbine, the acoustic metric that is most often recommended is the weighted cumulative sound exposure level (Popper et al. 2014, [NMFS] National Marine Fisheries Service 2016).

This report provides an initial analysis of the recorder 100 m from the turbine using the contextual information from the current meter and turbine state to separate the turbine noise from the background. It also provides a comparison to the outer Bay of Fundy sound levels and the marine mammal temporary hearing threshold shift guidelines. The information provided supersedes the first report on turbine sound that was summarized in the Q1 2017 Interim EEMP Report.

The report has the following sections: Section 2: Contextual information; Section 3: Methods; Section 4: Summary of the sound levels measured on the long-term Minas Passage and outer Bay of Fundy

recordings; Section 5: Interpreting the long-term recording using the environmental data; Section 6: Discussion. This report will be superseded by a more detailed report in 2018.

We caution that the OpenHydro turbine was under engineering evaluation during the sound monitoring period. During this time the turbine control center (TCC), which is an electrical component sub-system attached to the subsea base and connected to the turbine, was being used to test and establish operating parameters and was involved with regulating the turbine operation from not-spinning, to free-spinning and then generating. The results are therefore indicative of this commissioning phase.

Figure 1. Map of the study area, including locations of turbine and recorders. Station 1 was the control site located 680 m from the turbine. Station 2 was located 100 m from the turbine.

2. Context Underwater sound is generated by a variety of natural, or ambient, sources including breaking waves, high flows, and marine life. There is also sound that is generated by a variety of man-made sources, such as ships, and in-water activities like drilling and dredging. Sound can be measured as frequency which is the number of cycles of a sound wave per second. This is expressed as hertz (Hz). The primary sources of ambient noise can be categorized by the frequency of the sound. In the low frequency range of 20-500 Hz, ambient underwater sound in the ocean can be generated by the calls of large baleen whales, whereas in the higher frequency range of 500-100,000 Hz, ambient sound can be created by spray and bubbles associated with breaking waves. This sound increases with increasing wind and flow speeds (NRC 2003).

Increased human induced noise and vibration in the marine environment has the potential to directly or indirectly adversely affect fish and marine mammals since sound is a primary means by which many marine organisms learn about their environment and since sound is also the primary means of communicating, navigating, and foraging for many species of marine mammals and fish (NRC 2003). Marine fish utilize sound for communication, as well as for predator and prey detection, taking advantage of the rapid propagation of sound through water to perceive and discriminate sounds in the marine environment. Loud noises may result in behavioural responses, including avoidance of the source of noise, which could result in avoidance of primary feeding or spawning grounds. In the Minas Passage, marine fish are an important component of the marine ecosystem and play a significant role in the stability of Aboriginal and non-Aboriginal commercial fisheries, as well as recreational fisheries. Adverse effects from noise on the marine fish community may also in turn affect other ecosystem components that rely on marine fish as a food source. Regulatory considerations include the protection of fish and fish habitat under the federal Fisheries Act and additional protection of specific species under the Species at Risk Act.

Sound also plays a significant role for marine mammals as it is used for communication, individual recognition, predator avoidance, prey capture, orientation, navigation, mate selection, and mother-offspring bonding. It is known that high intensity sound can cause injury or behavioral changes in marine mammals but noise effects are, highly dependent on the individual, and the intensity, frequency, and duration of the sound, as well as the context in which the sound is received. This is a concern in the Minas Passage as harbour porpoise frequent the area using the Passage as a feeding area. Similar to fish populations, available information on the near-field effects of tidal energy devices on marine mammals is sparse. Marine mammals play a significant role in the marine ecosystem on both a local and regional scale and there are scientific and public concerns and regulatory protection to consider as well as the potential economic implications associated with the whale-watching industry.

The extent of human induced noise and vibration in the marine environment, in particular the relationship with sound produced by instream tidal turbines, is not well understood and data remains sparse. Due to this lack of published data related to noise monitoring of marine turbines during operation, studies need to be undertaken in order to reliably assess the significance of sound as a risk factor and to understand the environment in which the turbines operate. One objective of the demonstration project is therefore to gather acoustic information that will serve to further evaluate the potential effects of noise in order to determine the nature soundscape, assess acoustic effects, and develop effective mitigation, if required, that will facilitate the future planning.

To gather information on underwater sound, hydrophones or sound recorders are deployed. Hydrophones are passive devices (i.e., do not transmit sound) which are used to listen to sound underwater. These devices use microphones to convert sound into electrical signals which are used to measure the characteristics of the sound, such as amplitude and frequency, including sound created by

in-stream tidal turbines. Acoustic measurements of tidal turbine noise have so far proven problematic due to contamination from flow noise in high-flow environments. Flow-noise contamination can be reduced by collecting acoustic data using drifting hydrophones, but this approach provides only brief measurements that preclude a strong statistical sampling of the emitted noise. Fixed acoustic recorders enable longer recordings, but are more susceptible to flow-noise contamination compared to drifting recorders. This issue can be partly addressed by using equipment designed to minimize flow noise as much as possible.

Characterization of the spectral content of the flow noise and its correlation with current speed would also provide a meaningful understanding on the effects of flow noise on acoustic monitoring of tidal turbines. This knowledge would not only improve the ability to characterize noise from the turbine and increase understanding of potential environmental effects on marine fauna, but also would help to identify ways to mitigate the effects of flow noise on future acoustic measurements.

The overall purpose of the CST EEMP is to better understand potential effects and interactions of specific environmental components (i.e., fish, marine mammals, operational sound) in the near-field environment (i.e., <100 m) of the Open-Center in-stream tidal device. The scope for the turbine sound component of the EEMP is two-fold, involving investigations to determine the best way to record operational and ambient sounds in the Minas Passage and long-term acoustic measurements using fixed sound recorders and subsequent data analysis of the first turbine deployment to characterize the tidal turbine noise relative to the recorder flow noise (i.e., the noise created as the water flows over the recorder).

The tasks of the sound study were to:

1. Characterize the spectral content of flow noise in relation to tidal turbine acoustic measurement and its correlation with current speed;

2. Determine the cut-off frequency below and above which flow noise contaminates the acoustic measurements; and

3. Provide guidance on methodologies for performing acoustic measurements in close proximity to tidal turbines and processing of acoustic data to mitigate effects of flow noise.

3. Methodology

3.1. Recording Equipment

3.1.1. AMAR Autonomous Recordings To measure sound pressure levels (SPL), two bottom-mounted Autonomous Multichannel Acoustic Recorders (AMARs) with surface recovery floats were used (Figure 2). Due to the exceptionally high current in the area, hydrodynamic high-flow moorings (HFM) were used (Figure 3). The recorders were deployed near the OpenHydro turbine on 18 Nov 2016 (Table 1). The AMARs were moored at fixed locations 100 m and 680 m from the turbine on the seabed. Station 2 was retrieved on 19 Jan 2017, but Station 1 has not yet been retrieved due to difficulties in retrieving the unit.

Each AMAR was fitted with two M36-V35-100 hydrophones (GeoSpectrum Technologies Inc.), sampling for 250 seconds at 32,000 samples per second (sps) giving an acoustic bandwidth of 10 to 16 kHz, with a nominal sensitivity of −165 dB re 1 V/μPa. Channel 1 was located near the front, or ‘bow’ of the HFM, and Channel 2 was located near the lifting plate at the centre of the HFM (Figure 3). Channel 1 also sampled for 65 seconds at 375,000 sps, giving an acoustic bandwidth of 10 to 187.5 kHz, also with a nominal sensitivity of −165 dB re 1 V/μPa. There was a 165 second sleep cycle in the recording schedule to preserve battery life and memory. The lower sample rate can capture most mechanical noise from the turbine and vessels, as well as potential vocalizations from most large marine mammals. The high sample rate can capture high-frequency vessel sound sources, such as sonars and acoustic positioning systems. It can also capture high-frequency echolocation clicks from marine mammals. Two hydrophones were used to determine if there were differences in flow-noise reduction inside the HFM.

A similar AMAR recorder to the ones used at Minas Passage was deployed at 143 m water depth underneath the inbound Bay of Fundy shipping lane, adjacent to the North Atlantic Right Whale critical habitat (Table 1). This recorder used the mooring shown in Figure 4.

Figure 2. Configuration for the high-flow mooring.

Figure 3. Inside the high-flow mooring (HFM). Hydrophones are shown with red arrows.

Figure 4. Mooring configuration used in the Outer Bay of Fundy

Table 1. Recorder locations and deployment details from the OpenHydro study.

Device Latitude (N)

Longitude (W) Deployment Retrieval Horizontal range

from source (m) Sensor depth

(m)

AMAR 200, Station 2 45° 21’ 49.51 N 64° 25’ 24.78 W Nov 18 Jan 19 100 43

AMAR 227, Station 1 45° 21’ 45.12 N 64° 25’ 50.88 W Nov 18 Not yet

retrieved 680 46

Outer Bay of Fundy AMAR 44° 33’ 46.50' N 66° 20’ 9.90' W 3 Dec 15 28 Apr 16 -- 143

3.1.2. AMAR Autonomous Drifter An AMAR recorder integrated into a drifting mooring was deployed by FORCE on Oct 18 and 20 2016, before the installation of the OpenHydro turbine. The mobile recorder assembly consisted of a carefully designed catenary mooring in free-drifting arrangement (Figure 5). The mooring was designed was keeping the acoustic recorder from moving in the vertical axis due to wave motions, since 1 cm of vertical motion results in a 120 dB re 1 µPa pressure change, which is 10-times the background sound pressure level in most ocean areas. If the design was successful, then the recorder would effectively be a water ‘particle’ drifting with the water mass and recording the actual sound levels rather than artificial pressure changes caused by the interaction of the recorder and the environment. Temperature depth (TD) loggers were included in the mooring (Figure 5) to verify that different aspects of the mooring were moving as expected. Unfortunately, these devices were not properly activated during deployment, so no TD data was recorded.

The catenary mooring consisted of a buoyant surface unit and the AMAR acoustic recorder (JASCO) attached below it on an alternately weighted and buoyed line 35 m long. The surface unit comprised a large spherical float with an upper mast and a counterweight below at the end of a rigid rod. A satellite beacon and VHF/strobe beacon were mounted on the upper mast to facilitate tracking and retrieval. A pick-up line with floats and a fabric sea anchor (not shown in Figure 5) were also attached to the surface float. The line connecting the AMAR to the surface was fitted with weights and floats so that it formed a catenary shape carefully designed to minimize vertical movement of the recorder.

The AMAR was fitted with an M8E-35dB hydrophone (GeoSpectrum Technologies Inc.), duty cycled between 32,000 samples per second (sps) for 680 sec (acoustic bandwidth: 10 Hz to 16 kHz) and 375,000 sps for 130 sec (acoustic bandwidth: 10 Hz to 187 kHz, nominal sensitivity: −165 dB re 1 V/μPa).

Figure 5. Catenary mooring diagram

3.2. Recorder Calibrations Each AMAR was calibrated using a 42AC pistonphone calibrator (G.R.A.S. Sound & Vibration A/S) to verify the sensitivity of the whole recording system. The pressure response of the recording system was verified by placing the pistonphone and its adapter over the hydrophone while the pistonphone produced a known pressure signal on the hydrophone element (a 250 Hz sinusoid at 152.2 dB re 1 µPa).

3.3. Acoustic Metrics

Sound levels for each AMAR, with individual metrics defined below, are presented as:

• Broadband and approximate-decade-band SPL over time for the 10 Hz to 16 kHz, 10–100 Hz, 100 Hz to 1 kHz, 1–10 kHz, and 10–16 kHz frequency bands.

• Spectrograms of ambient noise were analyzed by Hamming-windowed fast Fourier transforms (FFTs), with 1 Hz resolution and a 50% window overlap. The 120 FFTs performed with these settings are averaged to yield 1 min average spectra.

• Statistical distribution of SPL in each 1/3-octave-band, with boxes of statistical distributions indicating the first (L25), second (L50), and third (L75) quartiles. The whiskers indicate the maximum and minimum data range. The solid line indicates the SPL, also called Leq, in each 1/3-octave.

• Spectral level percentiles in histograms of each frequency bin per 1 min of data. The Leq, L5, L25, L50, L75, and L95 percentiles are plotted. The L5 percentile curve is the frequency-dependent level exceeded by 5% of the 1 min averages. Equivalently, 95% of the 1 min spectral levels are above the 95th percentile curve.

• Daily sound exposure levels (SEL) computed for the total received sound energy and the detected shipping energy. The SEL is the linear sum of the 1 min SEL. For shipping, the 1 min SEL values are the linear 1 min squared SPL values multiplied by the duration, 60 s.

Sound is most commonly described using the sound pressure level (SPL) metric. Underwater sound amplitude levels are commonly measured in decibels (dB) relative to a fixed reference pressure of p0 = 1 μPa. The root-mean-square (rms) SPL is used to quantify the sounds generated by the turbine.

SPL (dB re 1 µPa) is the rms pressure level in a stated frequency band over a time window (T, s) containing the acoustic event:

SPL =

∫ 2

02

10 )(1log10 pdttpT T

.

The SPL is a measure of the effective pressure level over the duration of an acoustic event, such as the emission of one acoustic pulse or sweep. Because the window length, T, is the divisor, events more spread out in time have a lower SPL even though they may have similar total acoustic energy density.

Power spectral density (PSD) level is a description of how the acoustic power is distributed over different frequencies within a spectrum. It is expressed in dB re 1 µPa2/Hz.

The sound exposure level (SEL, dB re 1 µPa2·s) is a measure of the total acoustic energy contained in one or more acoustic events. The SEL for a single event is computed from the time-integral of the squared pressure over the full event duration (T100):

SEL =

∫ 2

002

10

100

)(log10 pTdttpT

where T0 is a reference time interval of 1 s. The SEL represents the total acoustic energy received at a location during an acoustic event; it measures the total sound energy an organism at that location would be exposed to.

Because the SPL and SEL are both computed from the integral of square pressure, these metrics are related by the following expression, which depends only on the duration of the energy time window T:

( )T1010logSEL SPL −= .

Sound level statistics, namely exceedance percentiles, were used to quantify the distribution of recorded sound levels generated by the turbine. Following standard acoustical practice, the nth percentile level (Ln) is the level (i.e., PSD level, SPL, or SEL) exceeded by n% of the data. Lmax is the maximum recorded sound level. Leq is the linear arithmetic mean of the sound power, which can be significantly different from the median sound level, L50. SPL can also be referred to as Leq, which stands for ‘equivalent level’. The two terms are used interchangeably throughout. The median level, rather than the mean, was used to compare the most typical sound levels between AMARs, since the median is less affected by high amplitude outliers (e.g., a crustacean tapping on the hydrophone) than the mean sound level. L5, the level exceeded by only 5% of the data, represents the highest typical sound levels measured. Sound levels between L5 and Lmax are generally from very close passes of vessels, very intense weather events, and other infrequent conditions. L95 represents the quietest typical conditions.

The PSD exceedance percentiles can be directly compared to the Wenz curves (Wenz 1962), which describe the PSD levels of marine ambient sound from weather, geologic activity, and commercial shipping. Figure 6 shows the Wenz curves along with the source levels of various types of anthropogenic sound sources. The “limits of prevailing noise” of the Wenz curves (black lines in Figure 6) represent the typical range of ambient sound PSD levels in the ocean and are plotted as orange dashed lines on the ambient sound PSD results in Section 4 for comparison.

Figure 6. Wenz curves describing pressure spectral density levels of marine ambient noise from weather, geologic activity, and commercial shipping (Wenz 1962). The limits of prevailing noise of the Wenz curves (thick black lines) are plotted as orange dashed lines on the ambient sound PSD results in Section 4. for comparison.

Source levels (SL) of the turbine are calculated by backpropagating the received levels (RL) using two different methods: the practical spreading model and the spherical spreading model. The two methods differ by the spreading term (A), which is determined by the water depth and the range (R) between the source and receiver in the SL equation:

( )RA 10log*RL SL +=

The practical spreading model uses a spreading term of 15 and is typically used for environments where R is a few times the water depth. The spherical spreading model uses a spreading term of 20 and is typically used when R is less than or comparable to the water depth. Source levels are presented for both models.

3.4. Vessel Detections Vessels are detected in two steps:

1. Constant, narrowband tones (tonals) produced by a vessel’s propulsion system and other rotating machinery are detected (Arveson and Vendittis 2000).

2. The rms SPL are assessed for each minute in the 40–315 Hz frequency band, which commonly contains the most sound energy produced by mid-sized to large vessels.

Background estimates of the shipping band SPL and broadband SPL are compared to their median values over the 12 h window, centred on the current time.

Vessel detections are defined by three criteria:

• The SPL in the shipping band is at least 3 dB above the median.

• At least 5 shipping tonals (0.125 Hz bandwidth) are present.

• The SPL in the shipping band is within 8 dB of the broadband SPL (Figure 7).

Figure 7. Example of broadband and in-band SPL and the number of 0.125 Hz wide tonals detected per minute as a ship approached a recorder, stopped, and then departed. The shaded area is the time period of shipping detection. All tonals are from the same vessel. Fewer tonals are detected at the ship’s closest points of approach (CPA) at 22:59 because of the broadband cavitation noise at CPA and the Doppler shift of the tonals .

4. Acoustic Measurements Summary

4.1. Minas Passage Autonomous Recorder (Station 2) Results for Station 2 are presented in Figures 8–10 for the entire deployment period (18 Nov 2016 to 19 Jan 2017). All results presented for Station 2 in this section are from Channel 1. The spectrogram and band level plot (Figure 8) shows that low-frequency noise dominated. The differences over a short time in the band level plot indicate that there were large variations due to tide. Figure 9 (bottom) shows the power spectral density (PSD) compared to the expected limits on prevailing noise. Typically, we compare the median, or L50, to the prevailing noise limits to describe ambient noise conditions. In this area, the L50 exceeded or was very close to the upper limit on prevailing noise for all frequencies. The SEL plot (Figure 10) shows the total and vessel-based SEL for each day of the deployment. The peaks in the SEL plot coincide with full moons on 13 Dec 2016 and 12 Jan 2017 which shows that the tidal flows aredriving the noise levels. At frequencies below 50 Hz flow noise over the hydrophones is the likely noise source. At frequencies above 50 Hz there is a combination of noise sources as discussed in Section 5.

Figure 8. (Top) in-band SPL and (bottom) spectrogram for Station 2.

Figure 9. (Top) Exceedance percentiles and mean of 1/3-octave-band SPL and (bottom) exceedance percentiles and probability density (grayscale) of 1-min PSD levelscompared to the limits of prevailing noise (Wenz 1962) for Station 2.

Figure 10. Total and vessel-associated daily sound exposure levels (SEL) and equivalent continuous noise levels (Leq) at Station 2.

4.2. Outer Bay of Fundy Recorder Broadband levels measured at the Outer Bay of Fundy were largely driven by the contribution of noise in the low-frequency bands. Median rms SPLs in the 10–32 Hz band were only 1–5 dB lower than broadband rms SPLs and at least 17 dB higher than in the 2–8 kHz band (Table 2). Median rms SPL in the 10–32 Hz band were similar at all stations, with Station 1 being the loudest, and the other three stations within 3 dB. The impact of tides and associated currents on the local soundscape, particularly at low frequencies, is again visible on the long-term spectrograms through spikes in noise levels separated by 28 days, i.e., the length of the lunar cycle (Figure 11). The power spectral density plots (Figure 12) do

not reveal any contribution of marine life. The SEL plot (Figure 13) shows the total and vessel-based SEL for each day of the deployment. The peaks in the SEL plot coincide with full moons, similar to the results shown in Figure 10 at Minas Passage.

Table 2. Broadband and in-band noise levels statistics for Stations 1 and 2. Median noise levels are bold.

Sound level statistic (dB re 1 µPa)

10–8000 Hz 10–32 Hz 40–160 Hz 200–1600 Hz 2.0–8.0 kHz

LMin 88.1 74.7 79.6 75.8 70.4

L95 106.6 96.2 97.8 96.5 87.9

L75 114.2 110.2 103.9 101.3 94.7

L50 121.3 120 108.3 103.9 97.4

L25 128.6 128.1 113.4 106.4 99.9

L05 136.8 136.5 121.9 111.5 103.7

LMax 155 152.6 152.4 134.9 130.4

Figure 11. Spectrogram and in-band rms SPL measured at 3 Dec and 28 Apr 2016 in the Outer Bay of Fundy recorder.

Figure 12. Power spectral density levels and 1/3-octave-band rms SPL at the Outer Bay of Fundy recorder between 2 Dec 2015 and 28 Apr 2016.

Figure 13. Total and vessel-associated daily sound exposure levels (SEL) and equivalent continuous noise levels (Leq) at the Outer Bay of Fundy Recorder.

5. Interpreting the Acoustic Measurements using the Turbine State and Current Data To help separate the effects of flow noise and the turbine, CST provided JASCO with the turbine’s operating state information for each minute, which included the normalized flow speed, current direction, and operating state (not spinning, free spinning, and generating). This information allows for a definitive association between the measured sounds and the operating state. Analysis was performed for the full recording band of the 32 kHz sampled data (i.e. 10-16000 Hz), as well as for each of the 1/3-octave-bands (centered at 10-12,500 Hz, see Section 3.3 for a definition of these bands). The analysis found that the minimum sound levels were at or near the 100 Hz 1/3-octave-band. Sound levels in the bands below 100 Hz increased with current speeds due to flow-noise over the sensors, and sound levels above 100 Hz increased with current speeds due to increases in ambient noise associated with the current. Thus we have chosen to use the 100 Hz 1/3-octave bands as a reference for this discussion.

As an example of this analysis, Figure 14 contains typical features found in the autonomous recorder data: the sound levels (vertical axis) increase and decrease throughout each day, with higher sound levels in the higher flows of the flood tide and lower sound levels in the ebb tide. The figures show that the sound levels at 100 Hz are much lower than the total sound levels (a 10 dB increase means that the sound level is 10 times higher).

For both the broadband (10-16000 Hz) and the 100 Hz 1/3-octave-band (Figure 14) the median one-minute sound pressure levels varied by 45 dB on the 17th of January. The broad-band levels were 5-10 dB higher in the flood state than the ebb state, while the 100 Hz levels were 15-20 dB higher in the flood state than ebb state. The 100 Hz SPL increased by 10-20 dB over a five-minute period when the turbine switched from not spinning to the free-spinning or generating states. The effect was less pronounced when more frequency bands were averaged together (Figure 14 left). The generating state produced broadband rasping sounds and the freewheeling operating state produced a knocking and vibrating sound, as well as tones in the 50–200 Hz range. Throughout the recordings there are also occasional impulsive sounds that were observed, possibly produced by unknown items or sediment striking the metal housing of the turbine or recorder. Overall, a wide variety of onsets and transitions were found. These may have depended on how the OpenHydro engineers configured the TCC which was under testing and evaluation during these recordings.

Figure 14. Examples of the measured 1-min sound pressure level (vertical axis) color-coded by the turbine operating state for a 24-hour period (horizontal axis). (Left) the total broadband sound levels measured on 17 Jan 2017. (Right) the sound levels in the only 100 Hz deci-decade band for the same day.

The analysis of the autonomous recorder data showed that the measured sound levels depend on the flow direction, flow speed, turbine state and 1/3-octave-band. In general, the sound levels did not increase linearly with current speed (Figure 15). In the 1250 Hz 1/3-octave-band shown in Figure 15, the sound level increases with flow speed almost linearly when the turbine is not spinning. When free-wheeling, the sound level increases with flow speed up to ~50% full flow, then remains constant until ~70% full flow. The sound level is nearly constant at all flow speeds when the turbine is generating. The background sound levels are similar to the turbine sound levels once the current speed exceeds 70% of full flow.

Figure 155. Example of the distribution of one-minute received sound levels in the 1250 Hz 1/3-octave-band for each of the operating states of the turbine. The red-line through the data is the k = 4 general additive model that best fits the measurements.

General Additive Models (red lines in Figure 15) were fit to the measurements using the software package ‘mgcv’ (Wood 2004) for the ‘R’ programming language. The models were then used to predict the median sound levels for each 1/3-octave-band at normalized flow rates of 20, 40, 60, and 80% of the full flow. The models are only valid above 70% flow for the flood direction since the ebb flow rarely exceeds 70% of the maximum flood flow. Similarly the sound levels at 100% of full flow were not included because we did not have sufficient examples of this operating state to develop a reliable model. These values are plotted for each of the operating states in the top row of Figure 16. These figures show several important features:

1. Below 100 Hz, the sound pressure level (SPL) decreases with increasing frequency at all flow rates, and the levels increase with flow rate. This is the region that is affected by ‘pseudo-noise’ of water flowing around the autonomous recorder.

2. The free spinning and generating turbine emits noise that is measurable 100 m from the turbine for frequencies above 50 Hz at all flow rates.

3. The free spinning turbine emits sounds that are above the background noise for frequencies up to 1000 Hz.

4. The generating turbine emits sounds that have two spectral peaks, one in the 1250 Hz band and the other in the 4000 Hz band. The generating turbine sounds measured 100 m from the turbine were 10–20 dB above the background levels at lower flow rates, and it becomes less significant as the flow rate increased above 50%.

The bottom row of Figure 16 contains the difference between the modeled median sound levels at the autonomous recorder and median sound levels recorded by JASCO at a 143 m water depth in the outer Bay of Fundy over five months in winter 2015-16. The flow speeds at this location were 1–2 m/s, and the recorder was located underneath the in-bound Bay of Fundy shipping lane. These results show:

1. At low flow rates in Minas Passage, the pseudo-noise (below 32 Hz) is similar to the measurements in the outer Bay of Fundy.

2. At frequencies of 32–315 Hz, the Minas Passage is up to 25 dB quieter than the outer Bay of Fundy due to the vessel noise that is always present near the shipping lane.

3. When the turbine is free spinning or generating, the sound levels up to 1000 Hz are similar to the outer Bay of Fundy for Minas Passage flow speeds up to 60% of full speed.

4. Above 315 Hz the free-wheeling and not-spinning sound levels are very similar, indicating that the ambient environment is the source of noise in these states. The sound levels increase with frequency as the flow rate increases due to sediment interaction noise (pebbles striking each other). This appears to be the dominant source of noise above 10 kHz even during the generating state. Further work through comparisons with the icListen hydrophones and drifting hydrophones is needed to verify this assessment.

5. At very high flow speeds, the Minas Passage sound levels are 10–30 dB above those in the outer Bay of Fundy. Up to 100 Hz, this is due to pseudo-noise. Above 100 Hz, the source of sound is a combination of the turbine and local sediment noise.

Figure 16. Effects of turbine operating state, current speed, and direction on measured sound levels. Top row: General Additive Modeled (Figure 15) one-minute 1/3-octave-band SPLs for each flow speed and operating state. Bottom row: Comparison of the modelled sound pressure levels to the median deci-decade SPL measured over four months at a site underneath the shipping lanes in the Bay of Fundy.

6. Discussion

6.1. Comparison to Effects of Noise Thresholds To understand the possible effects of the turbine noise on marine life the measured levels may be compared to published regulatory guidelines and recommendations. Here we consider the Technical Guidance for Assessing the Effects of Anthropogenic Noise on Marine Mammal Hearing issued by the United States National Marine Fisheries Service ([NMFS] National Marine Fisheries Service 2016), and Sound Exposure Guidelines for Fish and Sea Turtles (Popper et al. 2014). For fish with swimbladders involved with hearing, exposure to levels over 158 dB re 1 µPa for 12 hours or more has been shown to cause temporary hearing threshold shifts that last for days to weeks (Amoser and Ladich 2003, Popper et al. 2014). Exposure to a sound pressure level for a period of time is a sound exposure level, equal to the sound pressure level + 10·log10(duration in seconds); this SEL is 204 dB re 1 µPa²·s.

For marine mammals the threshold for exposure to continuous sounds like a tidal turbine is measured using a frequency-weighted SEL, where the frequency weighting approximates the hearing band of the species. The NMFS Technical Guidance contains five weighting curves for marine mammals (shown in Figure 17(A)). The recommended duration for accumulating the SEL is 24-hours. Figure 17(B) shows the daily un-weighted and marine mammal weighted SELs for the 2-month autonomous recording 100 m from the OpenHydro turbine. The sound levels in the low-frequency cetacean (baleen whale) and high frequency cetacean (porpoise) groups are shown in Figure 17(C) and (D). The highest unweighted sound exposure level (SEL) was 191 dB re 1 µPa²·s, which is below the recommended fish threshold of 204 dB re 1 µPa²·s by a factor of 20. For marine mammals the weighted SEL thresholds for the possible onset of temporary hearing threshold shift are given in Table AE-1 of NMFS [NMFS] National Marine Fisheries Service (2016). The thresholds range from 199 dB re 1 µPa²·s for otariid seals (fur seals and sea lions) to 153 dB re 1 µPa²·s for high-frequency cetaceans. The only threshold that was exceeded in the autonomous recording were the TTS thresholds for porpoise on 52 of the 63 monitored days (i.e. when the cyan curve in Figure 18(B) is greater than 153). This implies that porpoise in this environment could experience lower ability to hear because of hearing fatigue. As discussed in Section 5, the high frequency noise source appears to be sediment interactions (pebble collisions (see Bassett et al. 2013)). Thus, if the high frequency noise was dominated by environmental sources and Minas Passage is known porpoise habitat, we believe that porpoise hearing is not affected and the appropriate threshold in an environment like Minas Passage requires further investigation.

Figure 17. Exploring the autonomous recorder data using the marine mammal auditory weighting functions (NMFS 2016). (A) the marine mammal auditory weighting functions that reduce the sound levels for frequencies outside the presumed hearing band of different groups of marine mammals. (B) The 24-hour weighted sound exposure levels for the full autonomous recording period. The colors indicate the weighting function shown in (A), with orange being unweighted. (C) 24-hours of weighted sound pressure level measurements for the low-frequency cetacean group (i.e., baleen whales). (D) 24-hours of weighted sound pressure level measurements for the high-frequency cetacean group (i.e., porpoise). (C) and (D) are the same date (17 Jan 2017) as was shown in Figure 14.

7. Recommendations A full comparison of the data from all hydrophones (i.e., turbine mounted units and drifters used by FORCE) will provide a more detailed analysis of the soundscape in the Crown Lease Area and used towards the development of a model of sound emitted by an operating turbine as a function of frequency, current speed and direction and turbine operating state. The analysis will include:

1. An investigation of the differences in measured high frequency sound levels between the drifters and the autonomous recorders.

2. Further study the daily sound exposure levels at the Minas Passage site in relation to the published temporary hearing threshold shifts for harbour porpoise.

3. If possible, inversion of received levels to obtain a suitable monopole source level of the turbine sound that can be used with acoustic propagation models for future modeling of the cumulative effects of multiple turbines on mid and far field marine environment.

We note that the next OpenHydro Open-Center turbine planned for deployment will be an improved version and that some of the sounds noted in this report will no longer be a concern. It is expected that major changes in how the turbine transitions from not-spinning to free-wheeling to generating will occur, and that further sound measurements will be required.

7.1. Recorder Methodology

To potentially improve the HFM performance, it is possible that increasing the size of the HFM would reduce noise by increasing the distance from the flow shield to the hydrophone. We also recommend moving the Channel 2 hydrophone from its location behind the upright lifting plate to the ‘stern’ of the mooring, opposite Channel 1. Such a mooring is currently being built for the Bedford Institute of Oceanography and will be deployed in the North Atlantic Right Whale critical habitat in 2018.

As part of the analysis planned in 2018 we will provide a comparison of drifters, the turbine mounted hydrophones and the autonomous bottom moored hydrophone data. Although drifting data is useful in measuring turbine sound, the two months of data collected using the HFM proved extremely useful as long-term, static monitoring like this allows accurate characterization of the overall soundscape. Seasonal and tidal trends can be determined, and once the moorings are deployed, the devices are able to record throughout all weather conditions. This method of static moorings is therefore thought to be the best method for measurement and analysis, but the other methods could add greatly to knowledge about the overall soundscape.

Literature Cited [NMFS] National Marine Fisheries Service. 2016. Technical Guidance for Assessing the Effects of

Anthropogenic Sound on Marine Mammal Hearing: Underwater Acoustic Thresholds for Onset of Permanent and Temporary Threshold Shifts. U.S. Department of Commerce, NOAA. NOAA Technical Memorandum NMFS-OPR-55. 178 pp. http://www.nmfs.noaa.gov/pr/acoustics/Acoustic%20Guidance%20Files/opr-55_acoustic_guidance_tech_memo.pdf.

Amoser, S. and F. Ladich. 2003. Diversity in noise-induced temporary hearing loss in otophysine fishes. Journal of the Acoustical Society of America 113(4 Pt 1): 2170-9. NLM.

Arveson, P.T. and D.J. Vendittis. 2000. Radiated noise characteristics of a modern cargo ship. Journal of the Acoustical Society of America 107(1): 118-129.

Bassett, C., J. Thomson, and B. Polagye. 2013. Sediment-generated noise and bed stress in a tidal channel. Journal of Geophysical Research: Oceans 118(4): 2249-2265. http://dx.doi.org/10.1002/jgrc.20169.

Popper, A.N., A.D. Hawkins, R.R. Fay, D.A. Mann, S. Bartol, T.J. Carlson, S. Coombs, W.T. Ellison, R.L. Gentry, et al. 2014. Sound Exposure Guidelines for Fishes and Sea Turtles: A Technical Report prepared by ANSI-Accredited Standards Committee S3/SC1 and registered with ANSI. SpringerBriefs in Oceanography, Volume ASA S3/SC1.4 TR-2014. ASA Press. 87 pp.

Wenz, G.M. 1962. Acoustic ambient noise in the ocean: Spectra and sources. Journal of the Acoustical Society of America 34(12): 1936-1956.

Wood, S.N. 2004. Stable and efficient multiple smoothing parameter estimation for generalized additive models. Journal of the American Statistical Association 99(467): 673-686.