remote sensing and water modelling · 2020. 3. 2. · remote sensing models higher spatial and...

26
Australia’s National Science Agency Remote sensing and water modelling A 5 year outlook Tim Malthus | CSIRO Oceans and Atmosphere | 2020-02-27

Upload: others

Post on 31-Dec-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Australia’s National Science Agency

Remote sensing and water modellingA 5 year outlook

Tim Malthus | CSIRO Oceans and Atmosphere | 2020-02-27

Page 2: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

• Perspectives

• Trends in remote sensing

• Remote sensing and modelling relationship

• Issues and opportunities

• Conclusion

Structure

Page 3: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Observations(Data)

ObservatoriesProcessing /

Correction Modelling

My pipeline - observations and models critical to the success of decision making

Value extraction / application

Decisions Impacts

Value accumulates along the chain

Scientific understanding

Page 4: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Remote Sensing Models

Higher spatial and temporal coverage than in situ measurements

High spatial and temporal resolution

40+ year time series (hindcasting) Tool for forecasting, scenario assessment

Surface view only 3D structure

Cloud interference Cloud free, continuous

Large uncertainties Uncertainties?

Satellite drifts, calibration for time series Process understanding

Regional and temporal biases

Remote sensing and modelling

Page 5: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

National perspectives

Space Future Science Platform

Page 6: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Trend - Pervasiveness • Constellations

• Commercialisation

9/3/19, 9 ' 37 pmA sky full of (commercial) eyes – Up to 1,300 Earth observat ion satellites in next 5 years – Space IT Bridge

Page 1 of 6ht tps://www.spaceitbridge.com/a- sky- full- of- commercial- eyes- up- to- 1300- ear th- observat ion- satellites- in- next- 5- years.htm

Home > Hot Topics > A sky full of (commercial) eyes – Up t o 1,300 Earth observat ion satellit es in next 5

years

! LEAVE A COMMENT

" HOT TOPICS, IMAGING

A sky full of (commercial) eyes – Up to

1,300 Earth observation satellites in next 5

years

BY DOUG MOHNEY JULY 24, 2018

Today, there are around 300 or so Earth obser vat ion (EO)

LATEST NEWS

NanoAvionics

announces

2

satellite

launch

for 3

customers

UbiquitiLink

demos

“cell

tower

in

space”

OneWeb

launches

first six

broadband

satellites

OneWeb

#

Page 7: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Trend – Data continuity

Data supply guaranteed until 2030s

Supported by Landsat:Landsat 9 – Dec 2020Landsat 10 – in planning

Lymburner et al. (2016) Remote Sensing of Environment

Page 8: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

• Platforms:• UAVs

• Planes, balloons

• Cube and micro-sats

• Satellites

• Sensors:• Hyperspectral

• Lidar

• Radar

• Geostationery

Trend – Diversity

3890 M. F. McCabe et al.: The future of Earth observation in hydrology

Figure 2. An Earth-observing “system of systems” for revolutionizing our understanding of the hydrological cycle. This multi-scale, multi-

resolution observation strategy isnot really aconcept, asthetechnology existsand is largely in placenow. Supporting traditional space-based

satellites, therearenow arange of orbital options from commercial CubeSats to demonstration sensors on-board the International Space Sta-

tion. Beyond orbiting EO systems, technological advances in hardware design and communications are opening the skies to stratospheric

balloons and solar planes, as well as an explosion of UAV-type platforms for enhanced sensing. At the ground level, the ubiquity of mo-

bile devices are expanding traditional in situ network capacity, while proximal sensing and signals of opportunity are opening up novel

measurement strategies.

Hydrol. Ear th Syst. Sci., 21, 3879–3914, 2017 www.hydrol-ear th-syst-sci.net/21/3879/2017/

Page 9: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Trend - Hyperspectral satellites

Prisma (2019) Gaofeng-5 (2018) EnMap (2020)

DESIS (2018) HISUI (2019)

High spectral resolution promises improved parameters and reduced uncertainties

Page 10: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Lidar

Terrestrial Laser scanning - Mangroves

Lidar offers 3D structure, fine scale

Page 11: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Radar

NovaSAR (2018)

Sentinel-1 (2015)

SAR offers weather independence, structure, water delineation

Page 12: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Geostationery

Himawari-8 (2014)

Geostationery offers high temporal resolution, dynamic processses

Page 13: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Challenge – Data volumes

Page 14: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

• Analysis Ready Data

• Data cubes / Hubs (the pixel is key)

• Open data / Open source

• Off-the-shelf products

• Lowering the barrier to entry

• Operational services

• Processing in the cloud

Trend – Discoverability & processing

Dice…

…and Stack

Page 15: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Algal bloom detection and visualisation

Lake Hume, time series, January to July 2016

WorkflowProcessing Visualization

Malthus et al. (2019) Remote Sensing, 11:2954

Page 16: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

• Access is mixed• Continuity of the time series and products• Duplication of effort• Continuing calibration and validation

• From single-sensor products to the development of temporal merging, multi-sensor products, data assimilation

• Hierarchical approaches - Small scale high resolution to wider scale low resolution

• Automated recognition – structures, habitats, species

Issues and opportunities

Page 17: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

• Exploits high temporal, spatial and spectral resolution

• No single sensor achieves it, but we have sensors that satisfy each requirement

• Can we combine these to satisfy what we need?

• Spectral-temporal / Spatial-temporal fusion

• ML / AI

Data fusion, virtual datasetsHigh temporal frequency

High spectral resolution

High spatial resolution

FreqHSR30 m

Himawari 810 minutes

Low spatial, low spectral

Sentinel 3 OLCI

~15 bandsLow spatial

Sentinel 2 MSILandsat 8 OLI

20 – 30 mLow temporal

Page 18: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

• Greater synergy between RS and models

• RS informs model validation, model skill assessment

• Modelling informs RS, e.g. algorithm improvement

• Challenges:• Mis-match between model outputs and ocean colour products

• Each provides a different “measurement” inhibiting straightforward intercomparisons

• Differences in terminology/similarly names variables, uncertainties, new developments in modelling directly link to RS (example)

Remote sensing and modelling

Page 19: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

True colour from orbiting satellite (MODIS) True colour as a model output

eReefs – Validation using true colour

Page 20: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

20 |eReefs – RS data assimilation

+ =Simulated true coloureReefs 4km model 12:00 12 Sep 2013

Observed true colourMODIS - Aqua 13:20 12 Sep 2013

Assimilated true colourAssimilate RSR 12:00 12 Sep 2013

• Data assimilation system constrained using the mismatch between observed and simulated remote-sensing reflectance.

• Reanalysis (1 June 2013 – 30 October 2016 ) completed on 24 Feb

2017, 4 days ahead of schedule.

Page 21: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

• A RS data explosion, more accessible, higher resolution, diversity

• Modellers need to be better informed of RS use and developments

• Greater dialogue needed between the two communities – how to interpret the comparisons

• Data assimilation needs work – formal mechanisms to synthesize observations and models

• Uncertainty in RS products requires greater understanding

• Both RS and modelling depend on in situ data

Conclusions

Page 22: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Australia’s National Science Agency

Oceans and AtmosphereTim MalthusResearch Group Leader

+61 3 3833 [email protected]

Thank you

Page 23: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Driven by colour - information in spectral reflectance

R2 = 0.91Suspended solids

Cyanobacteria

Chlorophyll

Coloureddissolved organic matter

Suspended solids

Alert level Chlorophyll level

Green <20 ug Chl l-1

Amber>20-50 ug Chl l-1

Red >50 ug Chl l-1

NHMRC/WHO – Guidelines for recreational use

Page 24: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

AquaSAT-1AquaSAT-2

IoT-Sat IoT-SatIoT-Sat IoT-Sat

IoT-Sat

In-Situ Sensor Pack

AquaWatchData Integration Facility

Australia (& Global)

water monitoring system• Ground sensor networks + IoT

• New EO satellites

• Data integration

AquaWatch Australia Mission Concept

GoGC, DNRME, CSIRO Meeting – 3 December 2019

Page 25: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

time

Landsat(t2)

5 Oct 2000 9 Jan 2001 30 Mar 20012tst1t

simulated Landsat(ts)Landsat(t 1)

MODIS(t1) MODIS(ts) MODIS(t2)

Spatial resolution

500 m

25 m

Example - Landsat-MODIS blending

Page 26: Remote sensing and water modelling · 2020. 3. 2. · Remote Sensing Models Higher spatial and temporal coverage than in situ measurements High spatial and temporal resolution 40+

Waterhole persistence and quality

36 | Earth observation application in the Fitzroy, Darwin and Mitchell catchments

Figure 5-2 Waterhole distribution in the Darwin catchments Distribution shown in (a) May and (b) October along with the percentage of time that a pixel is inundated. Results are

based on the number of cloud-free observations.

Figure 5-3 demonstrates how the relatively persistent inchannel waterholes in the lower Mitchell

catchment reduce in size through the dry season, with many drying out completely or only

persisting for 20% of the years from 1988 to 2015.

Figure 5-3 Waterhole persistence in the Mitchell catchment Distribution shown in (a) May and (b) October along with the percentage of time that a pixel is inundated. Results are

based on the number of cloud-free observations.

May October

Chapter 6 Waterhole water quality | 49

‘Darwin 4’ and ‘Mitchell 2’). The data are here displayed as a function of the number of days since

the start of the dry season in each year (1 May).

Figure 6-7 clearly identifies two TSM regimes, namely a period of low and relatively consistent

TSM values during the dry season followed by a series of higher and more dispersed values during

the wet season. Local mean and variance estimates over a shifting window could be used to

provide temporal measures pertaining to seasonal processes affecting the waterhole turbidity.

Figure 6-7 TSM time series (in mg/L) vs time of year (number of days since 1 May) (a) ‘Fitzroy 4’, (b) ‘Darwin 4’, (c) ‘Mitchell 2’.