Estimation of chlorophyll a from time series measurements of high spectral resolution reflectance in an eutrophic lake

We acquired high spectral resolution reflectance data in Carter Lake, a eutrophic oxbow on the Iowa-Nebraska border, from April 1995 to April 1996. Chlorophyll a, total seston, sestonic organic matter, Secchi depth, and nephelometric turbidity were determined for each respective spectral measurement. Changes in algal taxonomic structure and abundance coincided with the development and senescence of a midsummer through autumn bloom of Anabaena. Taxonomic structure was more diverse in late winter and spring when Synedra sp. (diatom) and several chlorophytes and dinoflagellates were present. Overall, chlorophyll a varied from about 20 to 280 μg.L^-1, Secchi transparency from 18 to 74 cm, and seston dry weight from 11 to 48 mg·L^-1 in February and September, respectively. Particulate matter completely dominated lake water light attenuation. Dissolved organic matter had low optical activity. The most sensitive spectral feature to variation in chlorophyll a concentration was the magnitude of the scattering peak near 700 nm. The 700-nm peak correlated to chlorophyll concentration through the relationships between algal pigment absorption near 670 nm and the cell biomass and surface-related scattering signal in the near infrared. An algorithm relating the height of the 700-nm reflectance peak above a reference baseline between 670 and 850 nm to chlorophyll a was accurate and robust despite large variations in optical constituents caused by both strong seasonality in the algal system and short-term variations in seston from wind-induced sediment resuspension. The present algorithms were successfully used in other systems with different seasonality and productivity patterns. The coefficients of the models relating chlorophyll a and spectral reflectance variables appeared to be ecosystem specific: both the intercept and slope for the models in this study were moderately lower than for several other recently published results. We validated our algorithm coefficients with a second, independent dataset. The standard error for chlorophyll a prediction was ±28 μg·L^-1.