Johanna Press1, Jennifer Broughton2, Raphael Kudela2 · 2015. 6. 3. · High concentrations of...
1
High concentrations of trace element pollutants in the San Francisco Bay estuary necessitate consistent and thorough monitoring to mitigate adverse effects on biological systems and the contamination of water and food resources. Although the San Francisco Estuary Institute’s Regional Monitoring Program collects annual in situ samples from fixed locations, models proposed by Benoit, Kudela, & Flegal (2010) enable calculation of the water column total concentration and the water column dissolved concentration for 12 trace elements in the San Francisco Bay from a more frequently sampled metric—suspended solids concentration (SSC). This study tests the application of these models using SSC calculated from remote sensing data, with the aim of validating a tool for continuous synoptic monitoring of trace elements in the San Francisco Bay. The preliminary success of this method demonstrates that trace element models can significantly predict total and dissolved trace element concentrations in the San Francisco Bay estuary using suspended solids concentration values calculated from remote sensing reflectance. These results suggest that water quality monitoring programs in San Francisco and other regions may be able to use satellite imagery for consistent, thorough, and cost-effective monitoring. Benoit, Michelle D., Raphael M. Kudela, and A. Russell Flegal. 2010. “Modeled Trace Element Concentrations and Partitioning in the San Francisco Estuary, Based on Suspended Solids Concentration.” Environmental Science & Technology 44(15): 5956–63. Binding, C. E., J. H. Jerome, R. P. Bukata, and W. G. Booty. 2010. “Suspended Particulate Matter in Lake Erie Derived from MODIS Aquatic Colour Imagery.” International Journal of Remote Sensing 31(19): 5239–55. Nechad, B., K.G. Ruddick, and Y. Park. 2010. “Calibration and Validation of a Generic Multisensor Algorithm for Mapping of Total Suspended Matter in Turbid Waters.” Remote Sensing of Environment 114(4): 854–66. Miller, Richard L., and Brent a. McKee. 2004. “Using MODIS Terra 250 M Imagery to Map Concentrations of Total Suspended Matter in Coastal Waters.” Remote Sensing of Environment 93(1-2): 259–66. San Francisco Estuary Institute. Contaminant Data Display and Download. http://www.sfei.org/rmp/wqt. U.S. Environmental Protection Agency. 2014. “National Recommended Water Quality Criteria.” http://water.epa.gov/scitech/swguidance/standards/criteria/current/index.cfm. U.S. Geological Survey. Water Quality of San Francisco Bay. http://sfbay.wr.usgs.gov/access/wqdata. Johanna Press 1 , Jennifer Broughton 2 , Raphael Kudela 2 1 Department of Geology & Geophysics, Yale University, New Haven, CT; [email protected] 2 Ocean Sciences Department, UC Santa Cruz, Santa Cruz, CA Emily Schaller and Rick Shetter, National Suborbital Education and Research Center Ron Smith, Yale Geology & Geophysics; Larry Bonneau, Yale Center for Earth Observation Funding provided by NASA Student Airborne Research Program OS23D-1248 Figure 1: 42-27-11-RGB display of a HICO image containing much of the San Francisco Bay taken on 2011-09-17, during the September 2011 RMP sampling period. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 405 416 428 439 451 462 474 485 497 508 520 531 542 554 565 577 588 600 611 623 634 646 657 668 680 691 703 714 726 737 749 760 772 783 794 806 817 829 840 852 863 875 886 R-Squared Band Wavelength (nm) Linear Fit of Single-Band Proxies for SSC Calculating Suspended Solids Concentration Semi-analytical and empirical algorithms for SSC were tested against USGS in situ measurements. Methods tested included a semi-empirical algorithm using water-leaving reflectance to calculate SSC (Nechad, Ruddick, & Park, 2010) as well as empirical location-specific models for SSC developed from ratios of R rs in various bands (Doxaran, Froidenfod, & Castaing, 2002) and R rs values of individual bands (Binding et al., 2010; Miller & McKee, 2004). Models using single band proxies exhibited better fit than other types of algorithms. Models using each of the 86 HICO bands were tested, and the 634nm-band model exhibiting the best linear fit (R 2 = 0.794) was applied to the image to create a map of SSC. Figure 2: R-squared values of linear models created from R rs in individual bands. HICO bands include visible wavelengths (405-703nm, shown in the color bar) and near infrared wavelengths (707-892nm). Addressing Sunglint When the sun reflects off the ocean surface at the satellite’s viewing angle, it can produce sunglint that erroneously raises remote sensing reflectance values and leads to false-positive SSC data. Areas affected by sunglint were excluded from further analysis. Figure 3: Region A displays high modeled SSC values corresponding with elevated SSC measured in situ. Region B is affected by sunglint and displays similar modeled SSC but lower measured SSC. Trace Element Models Models were developed by Benoit, Kudela, & Flegal (2010) to predict concentrations from a 10-year dataset of in situ trace element and SSC samples. The elements modeled are Ag, As, Cd, Co, Cu, Fe, Hg, Mn, Ni, Pb, Se, and Zn. Each model calculates three parameters: Water Column Total Concentration (WCT), Distribution Ratio (D), and Water Column Dissolved Concentration (WCD). Coefficients m and b are specified by element. Constants α, ε, and δ account for geographic region in the Bay, season of sampling, and an interaction term, respectively. These models were applied to the map of SSC to create a map of WCT and WCD for each element. Eqn. 1: Formulae for (a) Water Column Total Concentration, (b) Distribution Ratio, and (c) Water Column Dissolved Concentration. Introduction Methods Results Model Results Statistically significant (p < 0.01) results for WCT were found for ten elements—As, Cd, Co, Cu, Fe, Hg, Mn, Ni, Pb, and Zn—with R 2 values between 0.50 and 0.80. R 2 was greater than or equal to that of the original models of Benoit, Kudela, & Flegal (2010) for seven of these elements. WCD was successfully modeled for six elements—As, Cd, Co, Cu, Ni, and Zn—with R 2 values between 0.40 and 0.85, exceeding that of the original models for each element. Water Quality Implications WCT values for As, Fe, and Pb were found to exceed EPA water quality criteria; Cd, Cu, Mn, and Ni were present at more than 50% of the recommended concentration (U.S. EPA 2014). All elements were present at highest concentrations in the southern region of the Bay. Figure 4: WCT maps for the 10 successfully modeled elements. Conclusions Continuing Work LANDSAT 8 images LANDSAT 8 captures an image of the full San Francisco Bay estuary every 16 days at a spatial resolution of 30m. Although it is not a hyperspectral sensor, a similar extent of visible and NIR wavelengths are captured in 5 bands. The method is being adapted to exploit the higher temporal and spatial resolution and larger spatial extent of LANDSAT 8 data, with the aim of enabling regular water quality monitoring. Assessing Environmental Conditions As demonstrated by sunglint in the HICO imagery, environmental conditions can necessitate adjustment or removal of data. The effects of water depth, as well as wind, cloud, and tidal conditions at the time of imaging and in situ sampling are being investigated. Figure 5: 432-RGB display of a LANDSAT 8 image containing the complete San Francisco Bay estuary taken on 2013-09-23, during the USGS SSC sampling period. Figure 6: Bathymetry of the San Francisco Bay estuary. Literature Cited Acknowledgements Remote Sensing Data Satellite imagery of the San Francisco Bay was chosen to align with the dates of annual Regional Monitoring Project (RMP) trace metal sampling. The Hyperspectral Imager for the Coastal Ocean (HICO) provides 86 bands of data with a spatial resolution of 90m and a bandwith of 5.7nm. HICO imagery was radiometrically corrected to remote sensing reflectance values (R rs ), georectified, and masked to contain only water pixels.
Transcript of Johanna Press1, Jennifer Broughton2, Raphael Kudela2 · 2015. 6. 3. · High concentrations of...
High concentrations of trace element pollutants in the San Francisco Bay estuary necessitate consistent and thorough monitoring to mitigate adverse effects on biological systems and the contamination of water and food resources. Although the San Francisco Estuary Institute’s Regional Monitoring Program collects annual in situ samples from fixed locations, models proposed by Benoit, Kudela, & Flegal (2010) enable calculation of the water column total concentration and the water column dissolved concentration for 12 trace elements in the San Francisco Bay from a more frequently sampled metric—suspended solids concentration (SSC). This study tests the application of these models using SSC calculated from remote sensing data, with the aim of validating a tool for continuous synoptic monitoring of trace elements in the San Francisco Bay.
The preliminary success of this method demonstrates that trace element models can significantly predict total and dissolved trace element concentrations in the San Francisco Bay estuary using suspended solids concentration values calculated from remote sensing reflectance. These results suggest that water quality monitoring programs in San Francisco and other regions may be able to use satellite imagery for consistent, thorough, and cost-effective monitoring.
Benoit, Michelle D., Raphael M. Kudela, and A. Russell Flegal. 2010. “Modeled Trace Element Concentrations and Partitioning in the San Francisco Estuary, Based on Suspended Solids Concentration.” Environmental Science & Technology 44(15): 5956–63.
Binding, C. E., J. H. Jerome, R. P. Bukata, and W. G. Booty. 2010. “Suspended Particulate Matter in Lake Erie Derived from MODIS Aquatic Colour Imagery.” International Journal of Remote Sensing 31(19): 5239–55.
Nechad, B., K.G. Ruddick, and Y. Park. 2010. “Calibration and Validation of a Generic Multisensor Algorithm for Mapping of Total Suspended Matter in Turbid Waters.” Remote Sensing of Environment 114(4): 854–66.
Miller, Richard L., and Brent a. McKee. 2004. “Using MODIS Terra 250 M Imagery to Map Concentrations of Total Suspended Matter in Coastal Waters.” Remote Sensing of Environment 93(1-2): 259–66.
San Francisco Estuary Institute. Contaminant Data Display and Download. http://www.sfei.org/rmp/wqt.
U.S. Environmental Protection Agency. 2014. “National Recommended Water Quality Criteria.” http://water.epa.gov/scitech/swguidance/standards/criteria/current/index.cfm.
U.S. Geological Survey. Water Quality of San Francisco Bay. http://sfbay.wr.usgs.gov/access/wqdata.
Johanna Press1, Jennifer Broughton2, Raphael Kudela2 1 Department of Geology & Geophysics, Yale University, New Haven, CT; [email protected] 2 Ocean Sciences Department, UC Santa Cruz, Santa Cruz, CA
Emily Schaller and Rick Shetter, National Suborbital Education and Research Center
Ron Smith, Yale Geology & Geophysics; Larry Bonneau, Yale Center for Earth Observation
Funding provided by NASA Student Airborne Research Program
1.41”
OS23D-1248
Figure 1: 42-27-11-RGB display of a HICO image containing much of the San Francisco Bay taken on 2011-09-17, during the September 2011 RMP sampling period.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
40
5
41
6
42
8
43
9
45
1
46
2
47
4
48
5
49
7
50
8
52
0
53
1
54
2
55
4
56
5
57
7
58
8
60
0
61
1
62
3
63
4
64
6
65
7
66
8
68
0
69
1
70
3
71
4
72
6
73
7
74
9
76
0
77
2
78
3
79
4
80
6
81
7
82
9
84
0
85
2
86
3
87
5
88
6
R-S
qu
are
d
Band Wavelength (nm)
Linear Fit of Single-Band Proxies for SSC
Calculating Suspended Solids Concentration Semi-analytical and empirical algorithms for SSC were tested against USGS in situ measurements. Methods tested included a semi-empirical algorithm using water-leaving reflectance to calculate SSC (Nechad, Ruddick, & Park, 2010) as well as empirical location-specific models for SSC developed from ratios of Rrs in various bands (Doxaran, Froidenfod, & Castaing, 2002) and Rrs values of individual bands (Binding et al., 2010; Miller & McKee, 2004). Models using single band proxies exhibited better fit than other types of algorithms. Models using each of the 86 HICO bands were tested, and the 634nm-band model exhibiting the best linear fit (R2 = 0.794) was applied to the image to create a map of SSC.
Figure 2: R-squared values of linear models created from Rrs in individual bands. HICO bands include visible wavelengths (405-703nm, shown in the color bar) and near infrared wavelengths (707-892nm).
Addressing Sunglint When the sun reflects off the ocean surface at the satellite’s viewing angle, it can produce sunglint that erroneously raises remote sensing reflectance values and leads to false-positive SSC data. Areas affected by sunglint were excluded from further analysis.
Figure 3: Region A displays high modeled SSC values corresponding with elevated SSC measured in situ. Region B is affected by sunglint and displays similar modeled SSC but lower measured SSC.
Trace Element Models Models were developed by Benoit, Kudela, & Flegal (2010) to predict concentrations from a 10-year dataset of in situ trace element and SSC samples. The elements modeled are Ag, As, Cd, Co, Cu, Fe, Hg, Mn, Ni, Pb, Se, and Zn. Each model calculates three parameters: Water Column Total Concentration (WCT), Distribution Ratio (D), and Water Column Dissolved Concentration (WCD). Coefficients m and b are specified by element. Constants α, ε, and δ account for geographic region in the Bay, season of sampling, and an interaction term, respectively. These models were applied to the map of SSC to create a map of WCT and WCD for each element.
Eqn. 1: Formulae for (a) Water Column Total Concentration, (b) Distribution Ratio, and (c) Water Column Dissolved Concentration.
Introduction
Methods
Results
Model Results Statistically significant (p < 0.01) results for WCT were found for ten elements—As, Cd, Co, Cu, Fe, Hg, Mn, Ni, Pb, and Zn—with R2 values between 0.50 and 0.80. R2 was greater than or equal to that of the original models of Benoit, Kudela, & Flegal (2010) for seven of these elements. WCD was successfully modeled for six elements—As, Cd, Co, Cu, Ni, and Zn—with R2 values between 0.40 and 0.85, exceeding that of the original models for each element. Water Quality Implications WCT values for As, Fe, and Pb were found to exceed EPA water quality criteria; Cd, Cu, Mn, and Ni were present at more than 50% of the recommended concentration (U.S. EPA 2014). All elements were present at highest concentrations in the southern region of the Bay. Figure 4: WCT maps for the 10 successfully modeled elements.
Conclusions
Continuing Work
LANDSAT 8 images LANDSAT 8 captures an image of the full San Francisco Bay estuary every 16 days at a spatial resolution of 30m. Although it is not a hyperspectral sensor, a similar extent of visible and NIR wavelengths are captured in 5 bands. The method is being adapted to exploit the higher temporal and spatial resolution and larger spatial extent of LANDSAT 8 data, with the aim of enabling regular water quality monitoring. Assessing Environmental Conditions As demonstrated by sunglint in the HICO imagery, environmental conditions can necessitate adjustment or removal of data. The effects of water depth, as well as wind, cloud, and tidal conditions at the time of imaging and in situ sampling are being investigated.
0.9”
Figure 5: 432-RGB display of a LANDSAT 8 image containing the complete San Francisco Bay estuary taken on 2013-09-23, during the USGS SSC sampling period.
Figure 6: Bathymetry of the San Francisco Bay estuary.
Literature Cited Acknowledgements
Remote Sensing Data Satellite imagery of the San Francisco Bay was chosen to align with the dates of annual Regional Monitoring Project (RMP) trace metal sampling. The Hyperspectral Imager for the Coastal Ocean (HICO) provides 86 bands of data with a spatial resolution of 90m and a bandwith of 5.7nm. HICO imagery was radiometrically corrected to remote sensing reflectance values (Rrs), georectified, and masked to contain only water pixels.
