WAMSI KMRP PROJECT 1.4
REMOTE SENSING
Peter Fearns
Acknowledgments • The State Government of Western Australia and WAMSI partners for
funding this research. • Jim Greenwood, Passang Dorji , Helen Chedzey, Mark Broomhall,
Edward King, Nick Hardman-Mountford, Nagur Cherukuru, David Antoine
• WAMSI Dredge Science Node, Themes 2/3 Predicting and measuring the characteristics of sediment plumes due to dredging operations
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 2
Overview
• Phase 1: Where did we start? • Phase 2:
– Remote Sensing – Development of TSS algorithm – Comparison of TSS algorithms – Spatial and temporal analysis – Light at Depth
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 3
Review Process Assets
Finfish
Coral
Seagrass
Invertebrates
Intertidal
Mangroves
Turtles
Cetaceans
Water Quality
Coastal Biological
Wilderness
DPaW-defined assets Define condition/pressure metrics Consider applicability of remote sensing. Data sources, spatial and temporal resolution, accuracy, confidence etc.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 4
Example of feedback for coral asset
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 5
After analysis and review…
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 6
After analysis and review… Turbidity (light)
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 7
KMRP
Goal: Quantify the reliability of remotely sensed turbidity products for use in the Kimberley region Objective 1: Analyse uncertainties of remotely sensed turbidity products by comparison of different algorithms and different resolution products with each other and with archived in situ data Objective 2: Analyse time series of remotely sensed turbidity data to provide first-stage pilot products that may be applicable for future use as marine management tools.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 8
Remote Sensing: A satellite-borne sensor measures radiance (light). We convert radiance to surface reflectance. A true colour Landsat image is shown at left.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 9
We relate reflectance to geophysical products such as TSS. MODIS TSS 2-band NIR AOD with coefficients from Onslow Dredge
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 10
TSS 2-band SWIR AOD with coefficients from Onslow Dredge
Landsat 8 TSS
30 m True colour image
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 11
Develop an algorithm…
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 12
• Compared exponential, linear and semi-analytical models
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 13
With more than 75 published TSS algorithms in the past decade, which one would you use?
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 14
• 5 different TSS types and a range of different Chl and CDOM conditions
• Hydrolight modelling • MODIS and Landsat • 49/27 different algorithms • Empirical and (semi)analytical
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 15
MODIS TSS algorithm comparisons
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 16
Landsat TSS algorithm comparisons
18 WAMSI KMRP Symposium: Perth, 28-29 Nov 2017
19 WAMSI KMRP Symposium: Perth, 28-29 Nov 2017
Paper 3 Impact of the Spatial Resolution of Satellite
Remote Sensing Sensors in the Quantification of Total Suspended Sediment Concentration: A Case
Study in Turbid Waters of Northern Western Australia
Dorji & Fearns
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 20
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 21
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 22
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 23
24
Average TSS over the sampled area
Upper and lower bounds over the sampled area
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 25
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 26
Different sensors: Different results mainly due to atmospheric correction
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 27
Different sensors: Different results mainly due to spatial resolution
Time Series Analysis
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 28
Figure 1. Spatial pattern of the (a) first, (b) second, and (c) third EOF modes of monthly TSS for region 1 (upper panels), and their corresponding time series of expansion coefficients (lower panel, where black=EOF1, blue=EOF2, and red=EOF3).
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 29
Figure 1. Spatial pattern of the (a) first, (b) second, and (c) third EOF modes of monthly TSS for region 2 (upper panels), and their corresponding time series of expansion coefficients (lower panel, where black=EOF1, blue=EOF2, and red=EOF3).
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 30
Figure 1. Spatial pattern of the (a) first, (b) second, and (c) third EOF modes of monthly TSS for region 3 (upper panels), and their corresponding time series of expansion coefficients (lower panel, where black=EOF1, blue=EOF2, and red=EOF3).
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 31
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 32
2002 2003 2004 2005
2006 2007 2008 2009
2010 2011 2012 2013
2014 2015
Figure 1. Yearly MODIS Aqua TSS anomaly images between 2002 (half year) and 2015 for the coastal region of Western Australia extending from Eighty Mile Beach towards Broome and along the west side of the Dampier Peninsula. Each yearly average is compared to an averaged TSS background determined from 10 years of monthly data averaged between January 2003 and December 2012.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 33
(1) Broome and surrounds
2002 2003 2004 2005
2006 2007 2008 2009
2010 2011 2012 2013
2014 2015
Figure 1. Yearly MODIS Aqua TSS anomaly images between 2002 (half year) and 2015 for King Sound in northwestern Western Australia. Each yearly average is compared to an averaged TSS background determined from 10 years of monthly data averaged between January 2003 and December 2012.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 34
(2) King Sound
2002 2003 2004 2005
2006 2007 2008 2009
2010 2011 2012 2013
2014 2015
Figure 1. Yearly MODIS Aqua TSS anomaly images between 2002 (half year) and 2015 for the coastal region of Western Australia extending from Collier Bay to ?. Each yearly average is compared to an averaged TSS background determined from 10 years of monthly data averaged between January 2003 and December 2012.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 35
(3) Collier Bay and surrounds
2002 2003 2004 2005
2006 2007 2008 2009
2010 2011 2012 2013
2014 2015
Figure 1. Yearly MODIS Aqua TSS anomaly images between 2002 (half year) and 2015 for the coastal region of Western Australia near Kalumburu. Each yearly average is compared to an averaged TSS background determined from 10 years of monthly data averaged between January 2003 and December 2012.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 36
(4) Kalumburu
2002 2003 2004 2005
2006 2007 2008 2009
2010 2011 2012 2013
2014 2015
Figure 1. Yearly MODIS Aqua TSS anomaly images between 2002 (half year) and 2015 for the coastline along northern Western Australia near Berkeley River. Each yearly average is compared to an averaged TSS background determined from 10 years of monthly data averaged between January 2003 and December 2012.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 37
(5) the region north of Berkeley river
2002 2003 2004 2005
2006 2007 2008 2009
2010 2011 2012 2013
2014 2015
Figure 1. Yearly MODIS Aqua TSS anomaly images between 2002 (half year) and 2015 for the Joseph Bonaparte Gulf. Each yearly average is compared to an averaged TSS background determined from 10 years of monthly data averaged between January 2003 and December 2012.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 38
(6) the region extending beyond the Western Australian border to capture the Joseph Bonaparte Gulf
Figure 49. Approximate boundaries of the 6 regions identified by analysis of spatial patterns in annual anomalies. (1) Broome and surrounds, (2) King Sound, (3) Collier Bay and surrounds, (4) Kalumburu, (5) the region north of Berkeley river, and (6) the region extending beyond the Western Australian border to capture the Joseph Bonaparte Gulf
Approximate boundaries of the 6 regions identified by analysis of spatial patterns in annual anomalies. (1) Broome and surrounds, (2) King Sound, (3) Collier Bay and surrounds, (4) Kalumburu, (5) the region north of Berkeley river, and (6) the region extending beyond the Western Australian border to capture the Joseph Bonaparte Gulf WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 39
Figure 1. Spatial patterns of the (a) first, (b) second, and (c) third EOF modes for region 1 during 2010 (upper panels), and their corresponding time series of expansion coefficients (lower panel, where black=EOF1, blue=EOF2, and red=EOF3).
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 40
Figure 1. Spatial patterns of the (a) first, (b) second, and (c) third EOF modes for region 2A (King Sound) during 2010 (upper panels), and their corresponding time series of expansion coefficients (lower panel, where black=EOF1, blue=EOF2, and red=EOF3).
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 41
Figure 1. Spatial patterns of the (a) first, (b) second, and (c) third EOF modes for region 2B (Collier Bay) during 2010 (upper panels), and their corresponding time series of expansion coefficients (lower panel, where black=EOF1, blue=EOF2, and red=EOF3).
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 42
Figure 1. Spatial patterns of the (a) first, (b) second, and (c) third EOF modes for region 3 during 2010 (upper panels), and their corresponding time series of expansion coefficients (lower panel, where black=EOF1, blue=EOF2, and red=EOF3).
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 43
Figure 1. Power spectrum for the first EOF mode of daily TSS variability during 2010 for (a) region 1, (b) region 2A, (c) region 2B, and (d) region 3.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 44
Light At Depth (LAD)
Sub-surface Photosyntetically Active Radiation (PAR) irradiance from the Hydrorad
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 45
TSS to Attenuation (Kd490)
• TSS and Kd relationships developed from fieldwork measurements taken near Onslow, June 2014
• Kd490 and KdPAR developed. • Natural logarithmic relationship
strong for “high” TSS concentrations but produces negative K at “low” TSS.
• Smith and Baker (1981), Kd490 for clear ocean water = 0.0212 m-1.
Linear relationship for TSS < 3 mg/L
Kd490 = 0.0765(TSS) + 0.0212
Natural logarithmic relationship for TSS > 3 mg/L
Kd490 = 1.018 (ln(TSS)) - 0.865 WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 46
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 47
Mean Tide Levels
• Broome – 2002, MTL = 5.39 m • Koolan Island – 1984, MTL = 5.69 m • Wyndham – 2015, MTL = 4.57 m • Other reference MTLs Cape Lambert = 3.23 m Onslow = 1.60 m Carnarvon = 1.08 m Jurien Bay = 0.84 m Barrack St, Perth = 0.85 m
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 48
Modelling Tides and LAD in Kimberley region
• Use a MTL of 5.5 m • Tidal ranges: 1 m to 10 m • Bathymetry levels: 1 m to 20 m • Attenuation coefficients: low (0.05), medium
(0.15), high (0.40) • Real Broome tide: November 2002
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 49
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 50
Spring tides
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 51
Neap tides
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 52
Hours of sun
• Broome Airport – approx. 10 hours of direct solar irradiance (not counting diffuse)
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 53
Bathymetry (m) Tidal Range (m)
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 54
TSS (mg/L)
TSS time averaged between 1/2/10 – 28/2/10
Percentage of light at depth time averaged between 1/2/10 – 28/2/10
LADPAR (%)
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 55
LADPAR (%) @
3 locations
Lat/Lon Bathy. (m)
Tidal Range
(m)
TSS (mg/L)
High Tide LAD
Mid Tide LAD
Low Tide LAD
★ 18.06° S, 121.94° E
18.0 8.4 0.80 8% 12% 17%
★ 17.47° S, 121.89° E
19.5 7.6 0.25 19% 23% 28%
★ 16.44° S, 123.48° E
16.3 9.3 2.40 2% 4% 7%
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 56
Implications for management
Remote sensing has been shown to be useful in providing data appropriate for detecting and quantifying turbidity and TSS concentration, as well as for visualizing spatial patterns and identifying change from time series analysis. The key considerations with respect to the efficacy of remote sensing technologies for monitoring and management of the natural environment of the Kimberley include: • Remote sensing technologies can provide data coverage for
the complete Kimberley region on a near-daily basis.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 57
Implications for management
• The long time series of remotely sensed data can provide retrospective views of the environment, and provide an assessment of baseline environmental conditions against which change can be measured
• Remote sensing technologies represent the lowest cost approach for routinely collecting scientific data at a regional scale.
• Remote sensing approaches may potentially be more respectful of Aboriginal culture than on-ground methods, providing greater potential for monitoring of sensitive sites.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 58
Implications for management
• It is important to note that provision of high quality regional remote sensing products relies on the availability of regional in situ measurements for tuning and validating algorithms. These measurements should be taken to constrain various aspects of the uncertainties associated with remote sensing reflectance retrievals (e.g. atmospheric aerosol and seabed reflectance influences) and conversions to derived parameters, such as TSS and chlorophyll.
• The spatio-temporal patterns of remotely sensed TSS were used to identify 6 different coastal regions in the Kimberley.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 59
Key residual knowledge gaps • We see significant value in extending the time series
analysis applied in this study to other variables. Application to satellite-derived sea surface temperature (SST) products would be the next highest priority as it would help identify regions of upwelling and mixing associated with increased nutrients and potential for higher water column productivity, as well as tracking the influence of heat waves and coral bleaching risk. This analysis should also take account of results from WAMSI KSN 2.2.7 on climate variability.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 60
Key residual knowledge gaps • We recommend that some future effort is directed toward
examining the specific processes (e.g. tidal mixing, river discharge) that are determining turbidity in the Kimberley region. One approach to this would be to extend the 3-D coupled physical-biogeochemical model, developed in WAMSI KMRP 2.2.2, to look in more detail at sediment transport dynamics. For this purpose, the archive of satellite TSS maps collated here will provide an invaluable validation data set. Knowledge gained within the WAMSI DSN (where sediment transport modelling is a key component) will also provide support for such a task.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 61
Key residual knowledge gaps • Developing an ongoing record of remotely-sensed TSS will require ongoing
field work to collect in situ radiometry, IOP and water samples for validation of the TSS algorithm.
• IOP-based algorithms that derive multiple variables and spectral-matching approaches that can switch parameterizations depending on optical water type have the greatest potential for being relevant across the range of marine waters in the Kimberley (from TSS to CDOM dominated waters) but these approaches require further development.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 62
Key residual knowledge gaps • Further development of the TSS-based light at depth methodology into
an operational remote sensing product is required. Development of a spectrally-resolved light at depth product would be a useful addition for understanding the availability of different coloured light at the seabed, hence the selective trade-offs between different benthic primary producers.
& Automatic delivery of TSS images/maps/products for any location MODIS (250m – daily) Landsat (30 m – 16 days) Himawari (500 m – 10 minute) Sentinel (10 m ~ 5 day)
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 63
Thank You.
WAMSI KMRP Symposium: Perth, 28-29 Nov 2017 64