Application of Aqua MODIS sensor data for estimating chlorophyll a in the turbid Case 2 waters of Lake Erie using bio-optical models

Figures & data

Figure 1. Bathymetric map (in meter depth) of the Western Basin of Lake Erie and the locations of 18 sampling stations (in red) visited during multiple cruises on Research Vessels Gibraltar. The lake interacts with terrestrial environment through fluxes of the Sandusky, Portage, Toussaint and other rivers such as the Maumee and Detroit, which drain into the far western side of the basin. For full color versions of the figures in this paper, please see the online version.

Table 1. Dates of each cruise and the date of the corresponding MODIS overpass of WBLE.

	Synchronous MODIS data
Field Cruise	04 June 12	11 June 12	18 June 12	26 June 12	02 July 12
MODIS overpass	06 June 12	10 June 12	19 June 12	26 June 12	04 July 12

Table 2. Algorithms used to estimate chlorophyll a concentration (C) in μg/l from MODIS observations. Note that where the maximum operator (max) appears, the largest of the quantities in the square brackets is used.

Algorithm	Equations	References
		O’Reilly et al. (1998)
		O’Reilly et al. (1998)
		O’Reilly et al. (1998)
		O’Reilly et al. (1998)
		O’Reilly et al. (2000)
		O’Reilly et al. (2000)
		O’Reilly et al. (2000)
		O’Reilly et al. (2000)
		O’Reilly et al. (2000)
OC3		O’Reilly et al. (2000)
		Darecki and Stramski (2004)
		Witter et al. (2009)
		Witter et al. (2009)
NIR/red
Ali _WBLE

Table 3. Descriptive statistics of chlorophyll a and TSM measured during five field campaigns.

Cruise date 12	Variable	N	Mean	S.D.	Variance	Min	Max
04 June.	Chlorophyll a	18	2.67	3.68	13.52	0.42	12.07
11 June.	Chlorophyll a	18	6.96	2.37	5.62	3.71	11.65
18 June.	Chlorophyll a	18	4.15	2.65	7.01	1.21	10.41
24 June.	Chlorophyll a	18	4.84	3.91	15.29	0.88	14.94
02 July.	Chlorophyll a	18	4.85	5.74	32.98	1.15	21.19
04 June.	TSM	18	5.03	2.81	7.90	1.60	13.30
11 June.	TSM	18	2.73	4.26	18.18	0.40	16.00
18 June.	TSM	18	4.10	4.96	24.62	1.30	19.50
24 June.	TSM	18	6.39	4.43	19.64	1.90	17.80
02 July.	TSM	18	3.44	3.46	11.99	1.50	15.00

Figure 2. Spatial variability of (a) chlorophyll a and (b) TSM, in the WBLE during five cruise periods.

Figure 3. (a) Remote sensing reflectance (Rrs) spectra of the cruise stations and (b) the derivative of the Rrs spectrum for the MODIS observations.

Table 4. Performance and comparison of the fifteen chlorophyll a algorithms applied to MODIS data, grouped by error response and sorted by bias.

Algorithm	Statistics of the one-to-one line			Statistics relative to observations
	Slope	Intercept	R^2	Bias	RMSE
Group 1 Algorithms
Morel 2	0.59	2.65	0.59	1.19	2.13
CalCOFI two-band linear	0.50	2.57	0.59	0.78	1.94
OC3	0.52	1.48	0.58	−0.23	1.80
OC2v4	0.40	1.77	0.59	−0.38	1.89
Baltic	0.43	1.44	0.6	−0.61	1.91
Morel 1	0.30	1.93	0.46	−0.61	2.13
Morel 3	0.41	1.49	0.47	−0.62	2.04
LE Witter	0.47	0.34	0.56	−1.55	2.39
WBLE Witter	0.48	−0.18	0.55	−2.03	2.73
Ali WBLE	0.56	1.63	0.62	0.68	1.36
Group 2 Algorithm
CalCOFI three-band	0.67	2.38	0.59	1.24	2.18
Group 3 Algorithms
Morel 4	2.00	9.73	0.59	13.63	14.52
CalCOFI two-band cubic	0.58	2.17	0.58	1.47	2.42
OC4v4	0.81	1.48	0.57	0.85	2.14
Group 4 Algorithm
NIR/red	0.09	2.64	0.19	−0.64	2.45

Figure 4. Satellite chlorophyll a estimates comparisons against in situ concentration using fifteen ocean color models.

Figure 5. Correlation between the derivative spectrum and model errors for the chlorophyll a algorithms grouped by response (see Table 5 for group membership list).

Table 5. Distribution of algorithms by potential bias.

Source of Error	Algorithm Group 1	Algorithm Group 2	Algorithm Group 3	Algorithm Group 4
CDOM	Moderate positive	Moderate positive	Weak positive	Moderate Positive
Suspended Sediment	Moderate negative	Moderate negative	Moderate negative	Strong negative
Phycocyanin to chorlophyll a ratio	n/a	Weak negative	Weak negative	Weak positive
Atmospheric Error	Weak negative	Weak negative	Weak negative	Strong positive
Algorithms in class	Morel-2 CalCOFI two-band linear OC3 OC2v4 Baltic Morel-1 Morel-3 Witter LE Witter WBLE Ali WBLE	CalCOFI three-band	Morel-4 CalCOFI two-band cubic OC4v4	NIR/red

Figure 6. The observed RMSE for the various algorithms presented as a function of samples size. The dotted line represents the expected RMSE for the OC4v4 vicarious calibration (n = 2803) extrapolated to smaller sample size based on the central limits theorem. Algorithms with RMSE values of <10% are performing as well as can be expected for the size of the vicarious calibration sample based on comparison with the central limits theorem, while those with RMSE of >10% exhibit bias that is greater than predicted by the central limits theorem.

Table 6. Sensitivity of models performance on the mismatches of MODIS overpass dates relative to dates of in situ observations.

	RMSE	RMSE	RMSE
Algorithm	Same day data	One day later data	Two days later data
Morel 1	0.83	2.20	2.82
Morel 2	0.96	2.11	2.95
Morel 3	0.96	2.22	2.40
CalCOFI two-band linear	0.77	1.87	2.82
CalCOFI three-band	0.87	2.23	2.93
CalCOFI two-band cubic	0.92	2.52	3.16
Morel 4	9.88	15.25	16.80
OC2v4	0.53	1.90	2.66
OC4v4	0.59	2.29	2.76
Baltic	0.84	1.96	2.60
Witter LE	1.51	2.67	2.47
Witter WBLE	1.97	3.05	2.61
OC3	0.66	1.87	2.36
NIR/red	2.07	4.43	5.07
Ali WBLE	1.18	1.18	2.54

O’Reilly, J. E., S. Maritorena, B. G. Mitchell, D. A. Siegel, K. L. Carder, S. A. Garver, M. Kahru, and C. McClain. 1998. “Ocean Color Chlorophyll Algorithms for SeaWIFS.” Journal of Geophysical Research: Oceans (1978–2012) 103 (C11): 24937–24953. doi:10.1029/98JC02160.

 Web of Science ®Google Scholar

O’Reilly, J. E., S. Maritorena, D. A. Siegel, M. C. O’Brien, D. Toole, B. G. Mitchell, M. Kahru, F. P. Chavez, P. Strutton, and G. F. Cota. 2000. Ocean Color Chlorophyll a Algorithms for SeaWIFS, OC2, and OC4: Version 4: SeaWIFS Postlaunch Calibration and Validation Analyses, Part, Vol. 3, 9–23. Washinton, DC: American Geophysical Union.

 Google Scholar

Darecki, M., and D. Stramski. 2004. “An Evaluation of MODIS and SeaWIFS Bio-Optical Algorithms in the Baltic Sea.” Remote Sensing of Environment 89 (3): 326–350. doi:10.1016/j.rse.2003.10.012.

 Web of Science ®Google Scholar

Witter, D. L., J. D. Ortiz, S. Palm, R. T. Heath, and J. W. Budd. 2009. “Assessing the Application of SeaWIFS Ocean Color Algorithms to Lake Erie.” Journal of Great Lakes Research 35 (3): 361–370. doi:10.1016/j.jglr.2009.03.002.

 Web of Science ®Google Scholar