Auxiliary material for Paper 2010gl045422

An empirical approach to derive MODIS ocean color patterns under severe sun glint

Chuanmin Hu
College of Marine Science, University of South Florida, 
St. Petersburg, Florida, USA

				
Hu, C. (2011), An empirical approach to derive MODIS ocean color patterns under 
severe sun glint, Geophys. Res. Lett., 38, L01603, doi:10.1029/2010GL045422.

Introduction

These figures include three image sets derived from MODIS/Aqua satellite 
measurements over three regions: the Tropical Atlantic, the East China Sea, and 
South Africa. 

The methods on how these images were generated were briefly described in the 
paper, but are detailed here.

MODIS Level-0 (L0) data were obtained from NASA Goddard Space Flight Center 
(oceancolor.gsfc.nasa.gov), processed using the software SeaDAS5.4 to obtain 
top-of-atmosphere (TOA) calibrated radiance (Level 1), which was then corrected 
for Rayleigh scattering and other gaseous absorption (e.g., ozone) using 
software from NASA's MODIS Rapid Response Team (Dr. Jacques Descloitres). The 
resulting reflectance (Rrc) data were mapped to a cylindrical equidistant 
projection using software written in house based on the Perl Data 
Language (PDL) (credit: Judd Taylor). 

The 645-, 555-, and 469-nm Rrc data were used to create RGB (true-color) 
composite images. The same data were also used to derive the MODIS color index 
(CI) images using the methods described in the paper. The chlorophyll-a 
concentration (Chl) images at 4-km resolution were derived from the MODIS 
Standard Mapped Images (SMI) obtained from NASA GSFC.  

1. 2010gl045422-fs01.jpg 
Figure S1. MODIS/Aqua imagery showing examples of the empirical de-glint 
algorithm for the tropical Atlantic (0N to 14N, 54W to 40W). Chl images in (b) 
and (e) were derived using SeaDAS6.1 default algorithms, where a significant 
portion is masked due to sun glint. In contrast, the corresponding CI images in 
(c) and (f) from the empirical algorithm clearly reveal the Amazon River plume 
patterns due to the North Brazil Current as well as the nearby eddies. These 
color patterns appear consistent in time under different glint conditions. 
Significant correlation is found between MODIS Chl (after log transformation) 
and CI (R-square > 0.90, n > 14000) for both cases, with RMS “error” in the CI-
predicted Chl < 32% or 0.11 (in log space) for Chl < 1.0 mg/m-3. Aerosol optical 
thickness at 869 nm derived from SeaDAS ranges between 0.022 and 0.21 (0.093+-
0.035) on 8/26/2010 and between 0.019 and 0.30 (0.088+-0.042) on 8/28/2010.

2. 2010gl045422-fs02.jpg 
Figure S2. MODIS/Aqua imagery showing examples of the empirical de-glint 
algorithm for the East China Sea (25N to 35N, 120E to 130E). Chl images in (b) 
and (e) were derived using SeaDAS6.1 default algorithms, where a significant 
portion is masked due to sun glint. In contrast, the corresponding CI images in 
(c) and (f) from the empirical algorithm clearly reveal the Yangtze (Changjiang) 
river plume other color patterns that are consistent in time under different 
glint conditions. Significant correlation is found between MODIS Chl (after log 
transformation) and CI (R-square > 0.93, n > 8400) for both cases, with RMS 
“error” in the CI-predicted Chl < 25% or 0.09 (in log space) for Chl < 1.0 mg/m-
3. Aerosol optical thickness at 869 nm derived from SeaDAS ranges between 0.04 
and 0.27 (0.11+-0.035) on 8/21/2010 and between 0.029 and 0.31 (0.115+-0.040) on 
8/23/2010. Color legends are shown in Figure S1.

3. 2010gl045422-fs03.jpg 
Figure S3. MODIS/Aqua imagery showing examples of the empirical de-glint 
algorithm for waters off South Africa (41S to 31S, 15E to 25E). Chl images in 
(b) and (e) were derived using SeaDAS6.1 default algorithms, where a significant 
portion is masked due to sun glint. In contrast, the corresponding CI images in 
(c) and (f) from the empirical algorithm clearly reveal the meso-scale eddies 
(rings) and other color patterns that are consistent in time under different 
glint conditions. Significant correlation is found between MODIS Chl (after log 
transformation) and CI (R-square > 0.90, n > 8900) for both cases, with RMS 
“error” in the CI-predicted Chl < 30% or 0.10 (in log space) for Chl < 1.0 mg/m-
3. Aerosol optical thickness at 869 nm derived from SeaDAS ranges between 0.02 
and 0.29 (0.073+-0.038) on 1/21/2007 and between 0.012 and 0.33 (0.097+-0.057) 
on 1/23/2007. Color legends are shown in Figure S1.