Multivariate approach for chlorophyll-a and suspended matter retrievals in Case II waters using hyperspectral data

Remote sensing has become very promising in providing temporal and spatial information regarding biogeodynamics in large and open freshwater bodies. However, in optically complex environments, such as in the Western Basin of Lake Erie (WBLE), the water contains multiple biogeochemical constituents or color producing agents (CPAs) such as, phytoplankton, suspend matter, and dissolved organic carbon. Identifying and analyzing in-water constituents in these waters is crucial for understanding and assessing many biogeochemical processes. For example, concentrations of chlorophyll-a and total suspended matter can be used as proxies to assess phytoplankton dynamics and particulate loading. However, quantitatively estimating concentrations of in-water constituents from satellite observations is complicated when working with mixed spectral signatures. With the advent of hyperspectral sensors, hyperspectral remote sensing is fast emerging as a key technology for advanced and improved understanding of optically complex waters. This study focuses on estimating concentration of chlorophyll-a and Total Suspended Matter (TSM) in the WBLE by applying partial least square (PLS) method to a full range (400 - 900 nm) of continuous narrow spectral bands. PLS models the covariance between hyperspectral bands and the CPAs and identifies the optimal bands that characterize most of the variance that exists in the CPAs. This method avoids the curse of dimensionality and the effects of multi-collinearity, a challenge that is associated with new generation hyperspectral satellite sensors. Validation parameters for the PLS - based models produced R2 of 0.84 for Chlorophyll-a with RMSE of 1.18μg/l, and R2 of 0.90 for TSM with RMSE = 1.26mg/l. This illustrates the potential of the PLS in isolating and extracting absorption features characterizing the various CPAs in optically complex Case II type waters.