Regional algorithms for remote-sensing estimates of total suspended matter in the Beaufort Sea

Shilin Tangab, Pierre Larouchec, Andrea Niemia & Christine Michela* a Fisheries and Oceans Canada, Freshwater Institute , Winnipeg , R3T 2N6 , Canada b State Key Laboratory of Tropical Oceanography, South China Sea Institute of Oceanology, Chinese Academy of Sciences , Guangzhou , 510301 , China c Fisheries and Oceans Canada, Institut Maurice-Lamontagne, Mont-Joli , Québec , Canada , G5H 3Z4 * E-mail: Christine.Michel@dfo-mpo.gc.ca. The large and variable riverine inflow to Arctic continental shelves strongly influences their chemical, biological, and optical properties. The Beaufort Sea receives the largest amount of suspended sediments amongst all Arctic shelves, with sediment-laden Mackenzie river waters strongly influencing bio-optical properties on the shelf. Here, we developed two regional algorithms for the estimation of total suspended matter (TSM) concentration using Medium Resolution Imaging Spectrometer (MERIS) spectral bands, based on in situ optical and suspended particulate data collected in the summer during the Canadian Arctic Shelf Exchange Study (CASES) in 2004 and during the Arctic Coastal Ecosystem Study (ACES) in 2010. The band ratio (where R rs is remote-sensing reflectance) R rs,560/R rs,490 was best correlated with low TSM concentrations (less than 3.0 g m−3), while higher TSM concentrations were well correlated to R rs,681/R rs,560. An empirical piecewise algorithm is thus proposed with the switch between the ratios being triggered by R rs,681/R rs,560 at a threshold value of 0.6. The second algorithm made use of support vector machines (SVMs) as a nonlinear transfer function between TSM concentrations and remote-sensing reflectance ratios R rs,681/R rs,560, R rs,665/R rs,560, and R rs,560/R rs,490. Results show that both algorithms perform better (31% and 25%, respectively) than other published TSM algorithms including the MERIS Case 2 water processor (C2R) neural network algorithm in the study area.