Computer Assisted Decision Support System for High Level Infrastructure Master Planning:

Case of the City of Portland Supply and Transmission Model (STM)

Richard N. Palmer’, Azad Mohammadi2, Margaret A. Hahn3, Dennis Kessleg, Joseph V. Dvorak’ and David Parkinson’

‘Professor of Civil and Environmental Engineering, University of Washington, Seattle, WA 981952700, (206) 685-2658, palmer@u.washington.edu.

2Portland Water Bureau, 1120 SW 5th Avenue, Portland, Oregon 97204-1926, (503) 823- 60 15, azadm@water.ci.portland.or.us

3Research Assistant, Department of Civil and Environmental Engineering, University of Washington, Seattle, WA 98 195-2700, (206) 616-1775, x@u.washington.edu.

4Portland Water Bureau, 1120 SW 5th Avenue, Portland, Oregon 97204-1926, (503) 823- 7473, dkessler@water.ci.portland.or.us

‘Portland Water Bureau, 1120 SW 5th Avenue, Portland, Oregon 97204-1926, (503) 823- 4889, jdvorak@ci.portland.or.us

6CH2M-Hill, 777-l 08th Avenue N.E., P.O. Box 91500, Bellevue, Washington, 98009, (425) 453-5005, dparkins@ch2m.com

Abstract

Quantitative techniques and computer-based models were first introduced into the water resources discipline in the early 1960s. Since then models developed for and applied to various aspects of managing water resources have evolved from “black box” paradigms to complex and comprehensive user-friendly interactive computer models. These efforts have been motivated by the apparent and increasing demand for more effective planning and policy making, and to aid in operating complex and often multi-objective and multi- purpose water resources systems. The success and effectiveness of water resources computer models in public agencies is impacted by the institutional framework in which the model was developed, the existence of an effective interface mechanism that can link modeler-model-user interactions, and the model’s contribution to facilitate and support the decision making process. The “Supply and Transmission Model” (STM) was developed to support a large infrastructure master plan designed to explore a wide variety of water supply, water management and water transmission options for the Portland Water Bureau (PWB). The model is a highly interactive evaluation tool, allowing PWB staff to create a shared understanding of the model’s assumptions and encourage the Water Bureau staff to explore how various future expansion and conservation alternatives will allow them to meet their water supply and natural resources management goals. The model has been created in the STELLA Research 5.1.1 simulation environment. One of the purposes of this paper is to discuss the critical factors that have contributed to the acceptance of STM as a planning decision support tool for the PWB.

Introduction

Since their introduction in the early 1960’s, computer programs have been used to provide information for civil engineers in their efforts to solve the challenge of providing

1

for the nation’s infrastructure. Undoubtedly, our ability to make rapid calculations has allowed engineers to explore potential designs more completely, and perhaps has enhanced our abilities to generate alternatives. What is less clear is the degree to which computer tools have been used directly in the decision making process to interactively help decision-makers choose between large, infrastructure alternatives. The term “decision support system” has been used widely in recent years. Loucks (1995) has defined it as “computer-based models together with their interactive interfaces” . . . . . ..“that provide timely information that support human decision makers.”

This paper describes the development of a decision support system that was created to aid the Portland Water Bureau in decisions concerning its Infrastructure Master Plan (IMP). Infrastructure master plans attempt to translate the long-term goals and objectives of an agency into specific actions and policies. In the case of water suppliers, this includes defining the region’s service area, the level of reliability that will be provided and the infrastructure that can best provide this service. The goal of Portland’s plan was to define a strategic direction for the bureau and to determine the type of infrastructure and management policy needed to implement the bureau’s chosen direction. Although these are broad policy questions, this paper suggests that computerized decision support systems can help frame these problems in a fashion that facilitates understanding of the underlying trade-offs that are implicit in the definitions of these strategic goals.

In the past decision support systems have been shown to facilitate short-term decision making (Palmer and Tull 1987). This is particularly true for repetitive decisions, such as the routing of water through a transmission system or hydropower releases from a reservoir (Palmer 1999). In these settings decisions occur often, and one incorrect decision can be easily repaired later. However, planning for a small number of very large infrastructure investments is an extremely different setting, and errors once made often can not be recouped (Keyes and Palmer 1993). Successful model application in these settings depends largely on the institutional setting in which model development and application is taking place. A framework for supporting modeling activity is essential for any organization in dealing with use of models in managing their water resources system (Mohammadi, et. al. 1991). The topic of this paper is a decision support system that was created for large infrastructure investments in a supportive setting.

Portland Water Supply System

The Portland Water Bureau (PWB) has served the city of Portland and its outlying areas since 1895. Today, the system serves some 840,000 people and delivers an average of 110 million gallons per day (mgd). PWB obtains its water from two primary sources, the Bull Run watershed, and the Columbia South Shore Wellfield (CSSW). Two major dams are located on the Bull Run River, known as Dam 1 and Dam 2. Dam 1 was constructed between 1927 and 1929 and currently impounds approximately 9.9 billion gallons (30,384 acre-feet). Dam 2 was constructed in the early 1960s and holds approximately 6.8 billion gallons (20,87Oacre-feet). Both reservoirs have dead storage components, and the total available water supply from the reservoirs is 10.2 billion gallons (3 1,305 acre- feet). There is also a natural lake (Bull Run Lake) in the watershed from which 1 billion gallons of water can be obtained on an emergency basis.

2

The CSSW was constructed in the early 1980s as an alternative, emergency source of water. The wellfield can produce up to 90 mgd and has been used for an extended period only once, during the 1987 drought event. This drought is the most severe drought to impact Portland in its recent hydrologic history. PWB has three major intown storage reservoirs, Powell Butte, Mt. Tabor, Washington Park. Total intown storage is 177 million gallons. The PWB has contracts with 19 water purveyors in the Portland, Oregon metropolitan area, serving some 300,000 customers. These customers account for 20% of the annual water sales and 40% of the annual water demand. The vast majority of Portland’s wholesale water contracts are up for renewal in 2004. The five largest wholesale customers, comprising 90 percent of its wholesale sales, are the City of Gresham, Rockwood Public Utility District, Powell Valley Road Water District, Tualatin Valley Water District (TVWD), and City of Tualatin.

PWB currently provides pH adjustment to minimize metal corrosion in plumbing and chloramines to disinfect the water. It is likely that filtration will be required for Bull Run water in the future. In addition to these drinking water concerns, the National Marine Fisheries Service (NMFS) listed Columbia River steelhead trout as a threatened species under the Endangered Species Act in 1997 and Chinook salmon in 1999. Because of these listings, there is considerable interest in the fisheries resources in the Bull Run River. Studies are underway to determine the impacts of increased fish flows in the river. In addition, dams in the lower portion of the river, the Little Sandy and Marmot dams, are being considered for removal to reopen prime salmon spawning habitat.

Water Supply Planning Studies

In 1989 Portland regional water providers (a group that came to be known as the Consortium) began discussing future regional water supply alternatives, regional transmission options, and governance issues. In 1996 the group completed its Regional Water Supply Plan (RWSP). This study investigated a number of specific issues including: 1) Filtering Bull Run water, which would provide an additional 2.7 billion gallons of water for water supply, 2) Increasing the elevation of Dam 2, which would increase the water supply storage by 2 billion gallons, 3) Constructing Dam 3 on Bull Run, providing 19 billion gallons of water, 4) Modifying Dam 1 gates, providing 200 million gallons, 5) Utilizing groundwater in the Bull Run basin, and 6) Increasing the groundwater potential of the CSSW to 100 mgd.

In 1998 Portland initiated its own regional water supply study. Its primary goals were to define specific alternatives for the Portland system that fit into the broad planning alternatives defined by the RWSP and to improve the Bureau’s ability to evaluate quickly and accurately system improvement alternatives. Other goals of the study included: 1) Provide water system operating and cost data for upcoming wholesale contract renegotiations, 2) Examine emergency supply issues, including interconnections between regional suppliers, 3) Define cost-effective procedures to deal with the Bureau’s aging infrastructure, 4) Develop evaluation procedures for alternatives, and 5) Identify preferred alternatives and other options that meet a variety of requirements in the areas of regional growth, environmental concerns, regulatory requirements and political realities.

To begin their study process, the Bureau reviewed its “Strategic Direction and Mission” statement. This policy statement provides PWB a range of potential roles in the region. These include: 1) Being the regional water provider, 2) Being the provider for Multnomah County, and to others as water is available, 3) Serving primarily Portland and providing water to others as available, and 4) Serving the historic service base, including its growth needs. These roles create very different opportunities and requirements for the agency. Implicit in the IMP process was the goal of determining which of these goals is the most appropriate and which best serves the citizens of public and the outlying areas.

To facilitate this process, a number of studies were conducted. Among these studies was the development of the Storage & Transmission Model (STM). The goal of this computer model is to allow Bureau engineers and planners to simulate the water supply demands and the major transmission linkages, terminal storage, and water supplies. By modeling the future water supplies and demands, Bureau staff could assess how various assumptions about the water system actually perform during a 50-year period.

Construction of the STM

The STM model was developed by a team of researchers from the University of Washington, water resource engineers from CH2M- Hill, and PWB staff. It was created using an iterative process requiring three phases of construction and critique. A number of fundamental issues had to be answered prior to actual model construction, including: 1) Selecting the modeling environment, 2) Choosing an appropriate time step, 3) Defining the appropriate level of detail, and 4) Designing the user interface. Initial efforts, before addressing these four issues, included extensive interviews with the staff of the PWB and review of past models and reports describing the water supply system components and operations. These interviews were extremely important in ensuring that the model developers understood the range of needs of the PWB staff and that the model, when completed, could address these needs. As in most large water supply agencies, various perspectives existed concerning how a model might be best implemented. Throughout the model construction process, the model developers continued to recognize who would use the model and how it would be used in practice.

One approach that was used to ensure that the construction process would result in a successful tool was to identify essential questions that had to be answered by the PWB through the IMP process. Among these questions were:

1. What is the safe yield of the current Bull Run River water supply? 2. How much does the safe yield increase if Dam 3 was constructed? 3. How much does the safe yield decrease if there was less reliance on groundwater? 4. How much does the safe yield increase if the available storage is increased in

Dam 2 by either increasing the operating height of the dam or providing increased treatment of the water?

5. Does the transmission system limit PWB’s ability to provide reliable service? 6. Should intown storage be increased? 7. In what future years is increased supply required and where?

4

Model development was guided by determining how best these questions could be answered and how best the results of the model could be communicated to PWB management staff.

An early, key question that required resolution was the choice of an appropriate modeling environment. The RWSP study resulted in a model that was written in Microsoft Visual Basic. Some PWB staff voiced strong support in developing the next model in a similar fashion; however, it was determined early in the modeling process that the STELLA modeling environment would be used. Selection of an appropriate time step was also resolved early in the process. A daily time step was chosen for the STM, as it provided the ability to look at daily maximum flow rates in the transmission system. A weekly time step would likely have been chosen if the only purpose of the model had been to evaluate the water supply yield of the Bull Run River. However, because of the importance of daily peak demands, the computational disadvantages of a daily time step outweighed the advantages of a weekly model.

Defining the appropriate level of modeling detail represented, perhaps, the most critical decision made by the modelers. A common goal of a modeler is to following O&ham’s razor. According to this fourteenth century philosopher, the best explanation is the one that uses the fewest variables to answer a question correctly. This helps ensure that the model is appropriately simple, without sacrificing the model’s accuracy. In the development of larger scale models, there is a constant and natural tension between making the model more detailed for the sake of completeness, versus adding only those elements that actually impact the model’s final answer. As noted previously, the final model had approximately 1,200 variables for each daily time step. Although this appears to be extensive, it should be noted that the model does simulate a relatively complex supply and transmission system. The model was developed in fifty conceptual building blocks, and this approach helped provide users with an organization structure that provided an increased degree of clarity.

User interface design also presented a challenge. A successful model can be characterized by the extent of its use in facilitating and supporting decision making. Such usability could further be enhanced if the end user or the decision-maker can easily interact with the model developers. This interaction requires the understanding of the process involved in designing and developing the model (Mohammadi 1993). It also was recognized that the model was not an independent entity, but rather part of a broader planning process. As the planning process evolved, so did interface requirements of the model. To address these issues the model was demonstrated to PWB staff at regular intervals to receive comments and critiques. Specific screens were designed with the staff to ensure that navigation through the model interface was intuitive and simple. Screens were organized into three primary types: 1) Scenario selection screens that allowed the user to define an alternative, 2) Data output screens that presented raw data to the user, and 3) Data summary screens that provided limited interpretation of the results. In the earliest phases of interface development, many significant suggestions were made and the model’s interface was modified accordingly. As the interations continued, the

extent of the needed modifications decreased dramatically. Figure 1 presents a typical example of the user interface.

Details of the Storage & Transmission Model

The Supply and Transmission Model (STM) . is constructed in the STELLAR (High Performance Systems) modeling environment. In addition, the STM imports and exports data from a series of Microsoft Excel spreadsheets. STELLA is an object-oriented programming environment that allows the development of complex simulation models, but requires a structured modeling approach. This approach facilitates the development of models that are relatively transparent, that is, an educated user can understand the model assumptions. The approach also encourages the development of models with highly interactive user interfaces. The STELLA portion of the model contains the water resource simulation component of the system, and the Excel spreadsheets contain demand -related data and selected output formats. The STM simulates daily system operation over a specified period using historic stream flow data and projected water demands. Typically, a user is interested in how the Bull Run system performs over an annual drawdown refill cycle. Because winters in the Pacific Northwest are typically wet, executing the model from January 1 through December 3 1 of a year normally simulates the system through a complete drawdown refill cycle. The model can also be made to run over many consecutive years.

The model tracks daily streamflows, flows into the Bull Run system, dam releases, dam elevations, flows in conduits, water demands throughout the system, and system shortfall. In total the model has over 1,200 variables, all of which assume daily values. The software is computationally efficient, evaluating the value of approximately 500,000 variables for a yearly run in about 30 seconds in a 550 megahertz Pentium II processor. Currently the model uses daily streamflow data from 1940 through 1998 and demand forecasts for the years 2000 through 2060, in five-year increments. The model can be viewed as a series of mass balance equations that track water as it enters the watershed, determines reservoir operations and transmission decisions, and follows the water movement to the end of the distribution system. The model is relatively large and detailed, but the system it represents is complex. Every effort has been made to make the model as simple as possible, and to provide the user an effective user-interface.

Ranking Alternative

A primary use of the STM model was to provide quantitative information with which alternatives could be ranked. This ranking process was performed in two stages. In the first stage, fourteen major supply improvements were identified, together with four conservation programs and four different service areas. These groupings provided a full range of alternative options. From the very large number of possible combinations of these alternatives, eleven were chosen for initial screening and ranking. Eleven were chosen, in part, to span the range of possible alternatives and to include those determined to be the most promising.

6

In the second stage, the STM model was used to evaluate twenty alternatives and to rank them relative to specific parameters, including 1) Adequacy of supply, 2) Groundwater use, 3) Degree of conservation, 4) Cost, 5) Environmental impacts, 6) Ability to implement, and 7) Flexibility. The STM provided useful evaluations for many of these parameters, although it was not used to evaluate all parameters (such as ability of implement). As the planning process evolved, a number of specific criteria were defined that became part of the alternative ranking process and were incorporated into the model. These included: 1) Minimum storage remaining during drawdown of most sever drought on record, 2) Number of days that groundwater was pumped during two most severe droughts on record, 3) Average cost of alternative, and 4) Average ratio of groundwater to served water used. Finally, based upon the results of the STM and the ranking process, recommendations were made (CH2M-Hill and Montgomery Watson 1999).

Success of the STM as a Decision Support Tool

As noted in the introduction to this paper, it is difficult to determine the success of a decision support tool that is being used in infrastructure planning. One typically does not have the luxury of performing a blind test in which top level managers in a utility are asked to rank project alternatives with and without the information generated by the computer support system. In addition, much of the value of a decision support system, like the STM, occurs not only during the analysis stage but throughout the planning process. In the development of the model, PWB staff often were challenged to think carefully about their system, how it could best be characterized, and if all members of the staff agreed upon key operations and policy issues. The process of creating a model, of translating the concepts and ideas associated with a system into a codified set of evaluation procedures forces a systematic evaluation of assumptions requiring participants to think about individual components of their system, as well as the system as a whole, in new ways. This can often have positive results.

Anecdotes are of limited value in building a strong case for decision support tools; however, they can help illustrate the use of the STM. During the second stage of the alternative ranking process, PWB engineering, planning and management staff were convened for a day long meeting to review the alternatives. At this meeting, the STM was used to evaluate alternatives in “real time.” Participants suggested system configurations and alternatives, the year in which the alternatives were to be put into place, the year in the future for which the system was to be evaluated, and the hydrologic record over which the system should be tested. The participants then decided whether the alternatives created acceptable results and if the results seemed reasonable. During the day, the results of the STM were often challenged. To some, specific results proved counterintuitive or simply surprising. But in every case, upon further review, the participants were able to explain the results and the source of the surprise. This process proved extremely useful, enhancing the groups understanding of the options available, creating more accurate intuition concerning the system, and identifying alternatives for consideration that might not have been identified otherwise. The opportunity to bring together the management, engineering, and planning staff was invaluable in refining the alternatives, and the STM played a significant role in supporting this process.

Conclusions

Infrastructure planning presents particular opportunities for water supply utilities. For the Portland Water Bureau, it provided the agency an opportunity to evaluate how these planning efforts could be integrated with an existing regional water supply. It also allowed the PWB the opportunity to develop tools to enhance its ability to plan. This paper has described the development of one of these models, the STM. This model allows the Bureau to simulate existing and future water demands, major transmission lines, and reservoirs. With the STM the PWB can evaluate system performance for specified surface water operating rules, use of groundwater, conservation options, flow requirements for fish, and system expansion options. The paper has described the steps involved in model creation and provided a short evaluation of the model’s success in supporting the planning process.

In the Infrastructure Master Plan process, a number of issues were identified that helped guide the creation of the STM. Stated simply, the model had to support the PWB staff in determining if there were advantages in expanding PWB’s service area and how conservation and other alternatives would be used to meet their primary goals. The goals of producing safe, inexpensive, and reliable water remain unchanged. This paper has suggested that decision support tools, such as the STM, can play a significant role in helping throughout the planning and analysis process.

References

CH2M-Hill and Montgomery Watson (1999) “Final Draft Report, Infrastructure Master Plan.” Portland, Oregon.

Keyes, A. M. and R. N. Palmer (1993) “Development of a decision support system for the prioritization of multimedia dischargers.” J. ofEnviron. Management, 601-6 12.

Loucks, D.P. (1995) “Developing and implementing decision support systems: A critique and a challenge.” Water Resources Bulletin, 31(4), 571-582.

Mohammadi, A. (1993) “Framework for Development and Use of Computer-Based Models in Water Resources and Environmental Decision Making.” Proceedings of the International Congress on Modeling and Simulation 1993: Modeling Change in Environmental and Socioeconomic Systems. University of Western Australia, Perth, Australia, December 6- 10, 1993.

Mohammadi, A., D. G. Fontane., and E. Vlachos (1991) “Guidelines for water resources models development and use.” Proceedings of the 18th ASCE National Water Resources Conference, New Orleans, LA, May 20-23, 1991.

Palmer, R.N. (1999) “Modeling water resources opportunities, challenges, and trade-offs: The use of shared vision modeling for negotiation and conflict resolution.” Proceedings of the ASCE ‘s 26’” Annual Conference on Water Resources Planning and Management, Tempe, AZ.

Palmer, R. N., and R. M. Tull(1987) “Expert system for drought management planning.” Journal of Computing in Civil Engineering, ASCE, 1, (4), 284-297.

8

Status Quo

Return to Main Menu To Control Panel

More Portland Metrics

Figure 1 Typical User Interface Screen for the STM

9

0 i