[Abstract]

We propose an innovative multi-pixel atmospheric correction approach (MPACA) to process high-spatial-resolution satellite measurements over coastal waters based on a revised POLYMER model. MPACA assumes the aerosol type to be uniform within a relatively small region, while the aerosol load and water properties are allowed to vary. Landsat-8 OLI images over six coastal locations with various turbidities were utilized to evaluate the performance of MPACA. The retrieved remote sensing reflectance (R_rs(λ)) by MPACA is validated with in situ matchups obtained from two sources: ship-based field campaigns and the AERONET-OC networks. It is found that, at each of OLI’s four visible bands, MPACA provided accurate R_rs(λ) products over such coastal environments, with the Root Mean Square Difference (RMSD) and Mean Absolute Percentage Difference (MAPD) less than 0.0006 sr^-1 and 16.2%, respectively. In contrast, the R_rs(λ) values retrieved with NASA’s SeaDAS (v7.5), where each pixel was treated independently, showed RMSD and MAPD as ~0.0018 sr^-1 and ~38.8%, respectively. Acolite-DSF, which assumed some spatial dependency, obtained MAPD almost two times that of SeaDAS for each visible band. Further, it appears that Acolite-EXP did not perform well for this evaluation dataset, where RMSD is ~0.0062 sr^-1 and MAPD is ~228.2%. These results suggest that MPACA is a promising scheme for atmospheric correction in coastal waters, especially for measurements from multi-band satellites that have a high spatial resolution along with at least two bands in the NIR or SWIR domain.

Fig.1. Scatterplot comparison between Landsat-8 OLI-derived R_rs and in situ R_rs at Landsat-8 each visible band from SeaDAS, Acolite-DSF, and Acolite-EXP. The black dash line represents the 1:1 line. The different color solid lines represent the linear regression for different AC algorithms corresponding to all matchups, respectively (blue line for SeaDAS, yellow line for Acolite-DSF, and red line for Acolite-EXP). “Field” and “OC” represent the in situ R_rs from the field campaigns (open symbols) and AERONET-OC sites (solid symbols), respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2. Same as Fig. 1, but for results of MPACA.

Fig. 3. (a–e) The comparisons between R_rs at 443, 482, 561, 655 and 865 nm derived from SeaDAS (x-axis) and MPACA (yaxis), respectively, for 2014-04-01 Landsat-8 OLI image over Belgian coastal waters. The dashed black line is 1:1 line, and the solid red line is the linear regression. The red dots represent the R_rs values retrieved by MPACA and SeaDAS for AERONETOC site-Zeebrugge_MOW1 on this day. (f) R_rs spectra of in situ measurement (red line), Landsat8 OLI-derived by MPACA (green line), SeaDAS (blue line) and Acolite-DSF (yellow line) method for AERONET-OC site-Zeebrugge_MOW1 over Belgian coastal waters on April 1st, 2014. The NIR spectral region is magnified in a small box. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Wang, J., Lee, Z., Wang, D., Shang, S., Wei, J., & Gilerson, A. (2021). Atmospheric correction over coastal waters with aerosol properties constrained by multi-pixel observations. Remote Sensing of Environment, 265, 112633