Initial Validation of GOCI Water Products against in situ Data Collected around Korean Peninsula for 2010-2011

Abstract

This paper provides initial validation results for GOCI-derived water products using match-ups between the satellite and ship-borne in situ data for the period of 2010-2011, with a focus on remote-sensing reflectance (R-rs). Match-up data were constructed through systematic quality control of both in situ and GOCI data, and a manual inspection of associated GOCI images to identify pixels contaminated by cloud, land and inter-slot radiometric discrepancy. Efforts were made to process and quality check the in situ R-rs data. This selection process yielded 32 optimal match-ups for the R-rs spectra, chlorophyll a concentration (Chl_a) and colored dissolved organic matter (CDOM), and with 20 match-ups for suspended particulate matter concentration (SPM). Most of the match-ups are located close to shore and thus the validation should be interpreted limiting to near-shore coastal waters. The R-rs match-ups showed the mean relative errors of 18-33% for the visible bands with the lowest 18-19% for the 490 nm and 555 nm bands and 33% for the 412 nm band. Correlation for the R-rs match-ups was high in the 490-865 nm bands (R-2-0.72-0.84) and lower in the 412 nm band (R-2=0.43) and 443 nm band (R-2-0.66). The match-ups for Chl_a showed a low correlation (<0.41) although the mean absolute percentage error was 35% for the GOCI standard Chl_a. The CDOM match-ups showed an even worse comparison with R-2<0.2. These match-up comparison for Chl_a and CDOM would imply the difficulty to estimate Chl_a and CDOM in near-shore waters where the variability in SPM would dominate the variability in R-rs. Clearly, the match-up statistics for SPM was better with R-2=0.73 and 0.87 for two evaluated algorithms, although GOCI-derived SPM overestimated low concentration and underestimated high concentration. Based on this initial match-up analysis, we made several recommendations -1) to collect more offshore under-water measurements of the R-rs data, 2) to include quality flags in level-2 products, 3) to introduce an ISRD correction in the GOCI processing chain, 4) to investigate other types of in-water algorithms such as semi-analytical ones, and 5) to investigate vicarious calibration for GOCI data and to maintain accurate and consistent calibration of field radiometric instruments.