Assessment of ocean color data records from MODIS-Aqua in the western Arctic Ocean

A broad suite of bio-optical and biogeochemical observations collected during the NASA-funded ICESCAPE expeditions to the western Arctic Ocean in 2010 and 2011 was used to validate ocean color satellite data products in this region, which is undergoing fast ecological changes in the context of a changing climate. Satellite-to-in situ match-ups for the MODIS instrument on board Aqua (MODISA) were evaluated using standard NASA empirical and semi-analytical algorithms to estimate chlorophyll-a (C-a), spectral marine inherent optical properties, and particulate organic carbon (POC). Results for the empirical algorithms were compared with those from the semi-analytical Generalized Inherent Optical Property (GIOP) algorithm. The findings presented here showed that MODISA Ca estimates were positively biased relative to in situ measurements, in agreement with previous studies that have evaluated ocean color retrievals in the Arctic Ocean. These biases were reproduced using both satellite and in situ measured remote sensing reflectances, R-rs(lambda), indicating that estimation errors are derived from the application of the empirical algorithm and not by the observed radiometry. This disparity appears to be caused by contributions of high spectral absorption from chromophoric dissolved organic matter (CDOM), which is a well-documented feature of Arctic Ocean waters. The current MODISA empirical algorithm (OC3M) appears to attribute CDOM absorption in the blue region of the spectrum to phytoplankton absorption. In contrast, GIOP showed significant improvement over OC3M C-a estimates by effectively discriminating between phytoplankton and CDOM absorption. Additionally, executing GIOP with an expanded set of spectral bands derived from in situ radiometry, instead of just six MODISA bands, further improved the performance of absorption estimates. These findings reinforce previous suggestions that semi-analytical approaches will provide more reliable data records for Arctic studies than existing empirical methods. POC estimates showed no clear bias relative to in situ measurements, suggesting that the empirical algorithm better represents high latitude oceans with regards to the bio-optical signature of suspended particulate stocks. (C) 2015 Elsevier Ltd. All rights reserved.