AN EVALUATION OF MODIS/AQUA BIO-OPTICAL ALGORITHMS IN ARCTIC WATERS.

Despite the Moderate Resolution Imaging Spectrometer (MODIS/Aqua) presently collects data on operational basis, evaluation of its biogeochemical products from standard bio-optical algorithms is scare especially in Arctic waters. The present study is aimed to evaluate several MODIS/Aqua bio-optical algorithms and MODIS/Aqua remote sensing reflectance (R_rs) products in Kongsfjorden waters on the western side of Svalbard, Arctic. Data sets consisting of in-situ measurements of radiometric quantity (R_rs), Chlorophyll-a (Chl-a) and Suspended Sediment (SS) concentrations, and absorption coefficients of coloured dissolved organic matter (a_CDOM) and phytoplankton (a_ph) collected coincidently with MODIS/Aqua observations during the 2007 and 2008 cruises were used to evaluate performances of the MODIS/Aqua algorithms in a wide range of waters within Kongsfjorden and its offshore regions with Chl-a varying from 0.01 to 5.8 mg m^-3, a_CDOM (400) from 0.01 to 0.4 m^-1, and aph (400) from 0.01 to 0.08 m^-1. The comparison of MODIS/Aqua Rrs from the standard atmospheric correction (SAC) with in-situ measurements at six wavelengths (412, 443, 488, 531, 551 and 667nm) showed that these satellite R_rs values are notably higher at 412nm and 443nm and lower at 551nm and 667nm than in-situ R_rs values at these wavelengths. By contrast, R_rs values from the MUMM atmospheric algorithm are slightly better than those of the SAC algorithm at 412 and 443nm, although being severely underestimated at 551 and 667nm. The over- and underestimated R_rs values in short and long wavelengths may be caused by adjacent effects and sub-pixel contamination of the ice cover and clouds in the region. An evaluation of the bio-optical algorithms revealed systematic and large overestimations with Chl-a and a_CDOM(400) retrievals and underestimations with aph(675) retrievals from the SAC_R_rs data. There was an inverse trend with MUMM that underestimated Chl and a_ph values although producing nearly consistent results for a_CDOM. Among the algorithms tested, MODIS_DC_case-2_Chl (Default case), MODIS_OC3_Chl and MODIS_DAAC-v4_ Chl algorithms had the relatively better statistics than other pigment algorithms. The analyses further demonstrate that algorithms for estimating the absorption coefficients of CDOM and phytoplankton also tended to produce high errors (with overestimation at 400nm for a_CDOM and underestimation at 675nm for a_ph). These analyses suggest that an independent treatment of the atmospheric correction and the development of new bio-optical algorithms are required to obtain reliable estimates of water constituents and inherent optical properties in high latitude waters. [ABSTRACT FROM AUTHOR]