APPLICATION OF FIRST DERIVATIVES AND EMPIRICAL VIS/NIR BIO-OPTICAL MODELS TO ESTIMATE PHYTOPLANKTON CONCENTRATIONS IN CASE II WATERS

Distributions of phytoplankton in space and time have major implications for water quality and ecosystem function. Lake Erie is biological the most active among the Great Lakes and experiences recurring algal bloom associated with harmful algal species such as Microcystis aeruginosa. This is of great concern for human health and is detrimental to the lake's biodiversity. Recently, the Western Basin of Lake Erie (WBLE) exhibited some of the worst water quality conditions, resulting in economic losses from cleanup costs as well as from decreased fishing and recreational activities. Water quality assessments for Lake Erie are largely based on conventional in-situ measurements with limited spatial and temporal resolution. Remote sensing techniques have indicated a more efficient and effective way to monitor water quality at high spatial and temporal scales. The optical property of the WBLE is governed by multiple constituents including phytoplankton, suspended sediments, and colored dissolved organic matter (CDOM), making it optically complex Case II type water. Case II waters present challenge in isolating reflectance trends attributed to a single constituent such as phytoplankton. In this study first derivate hyperspectral data and VIS/NiR bio-optical models are employed to retrieve concentrations of chlorophyll-a (Chl-a), a proxy for phytoplankton density, independently. Five cruises were conducted aboard Research Vessel Gibraltar III to collect samples and measure in-situ optical and biogeochemical data at 18 locations that encompass many of the environments in WBLE ranging from deeper waters, shallower bay waters and riverine discharges. Average concentration of Chl-a in the WBLE was 4.69 µg/l and ranged between 0.42 µg/l and 21.19 µg/l. Blue/Green and Red/NIR bio-optical models and algorithms using first derivatives of hyperspectral data were employed to assess the efficiency of reflectance data in determining Chl-a in the WBLE. The Blue/Green models are able to explain 49% of the Chl-a variability. First derivative data at 710 nm is able to explain 64% of Chl-a variability. A two band and a three band bio-optical models generated using Red/NIR spectral region are able to explain 66% of Chl-a, illustrating the potential of the Red/NIR algorithms in accounting for Chl-a variation in turbid Case II waters.