Remote Sensing of Cyanobacteria in Case II Waters Using Optically Active Pigments, Chlorophyll a and Phycocyanin

Abstract

Nuisance blue-green algal blooms contribute to aesthetic degradation of water resources and produce toxins that can have serious adverse human health effects. Current field-based methods for detecting blooms are costly and time consuming, delaying management decisions. Remote sensing techniques which utilize the optical properties of blue-green algal pigments (chlorophyll a and phycocyanin) can provide rapid detection of blue-green algal distribution. Coupled with physical and chemical data from lakes, remote sensing can provide an efficient method for tracking cyanobacteria bloom occurrence and toxin production potential to inform long-term management strategies. In-situ field reflectance spectra were collected at 54 sampling sites on two turbid, productive Indianapolis reservoirs using ASD Fieldspec (UV/VNIR) spectroradiometers. Groundtruth samples were analyzed for in-vitro pigment concentrations and other physical and chemical water quality parameters. Empirical algorithms by Gitelson et al. (1986, 1994), Mittenzwey et al. (1991), Dekker (1993), and Schalles et al. (1998), were applied using a combined dataset divided into a calibration and validation set. Modified semi-empirical algorithms by Simis et al. (2005) were applied to all field spectra to predict phycocyanin concentrations. Algorithm accuracy was tested through a least-squares regression and residual analysis. Results show that for prediction of chlorophyll a concentrations within the range of 18 to 170 ppb, empirical algorithms yielded coefficients of determination as high as 0.71, RMSE 17.59 ppb, for an aggregated dataset (n=54, p<0.0001). The Schalles et al. (2000) empirical algorithm for estimation of phycocyanin concentrations within the range of 2 to 160 ppb resulted in an r2 value of 0.70, RMSE 23.97 ppb (n=48, p<0.0001). The Simis et al. (2005) semi-empirical algorithm for estimation of chlorophyll a and phycocyanin concentrations yielded coefficients of determination of 0.69, RMSE 20.51 ppb (n=54, p<0.0001) and 0.85, RMSE 24.61 pbb (n=49, p<0.0001), respectively. Results suggest the Simis et al. (2005) algorithm is robust, where error is highest in water with phycocyanin concentrations of less than 10 ppb and in water where chlorophyll a dominates (Chl:PC>2). A strong correlation between measured phycocyanin concentrations and blue-green algal biovolume measurements was also observed (r2=0.95, p<0.0001).