Abstract
Peridinium is a rare but, toxic bloom-forming dinoflagellate in freshwaters. Its toxic effects were reported from several countries including Sri Lanka although not-much attened. In this study, we developed a remote sensing-based empirical model to quantify Peridinium using Maussakelle Reservoir in Sri Lanka as the model. Since carotenoids are the major light-harvesting accessary pigments of Peridinium and many other dinoflagellates, it serves as a unique biomarker. Thus, spectral signatures of carotenoids allowed us to distinguish Peridinium in the background of chlorophyll-dominated mix population of phytoplankton. Ground data and Sentinel-2 satellite images were collected when a high density of Peridinium and carotenoid pigment levels were present and a set of linear regression models were developed. Among the models, that developed with B2 and B3 bands of Sentinel-2 better regressed with measured carotenoid (R2 = 0.93, p < 0.001). The relationship between measured and model-predicted carotenoid concentrations displayed a correlation (R2) of 0.86 and root mean squared error (RMSE) of 2.82. Further, a second regression model was developed to predict Peridinium cell density using carotenoid as a proxy. The established relationship was strong and significant (R2 = 0.85, p < 0.001). Then a final empirical model was derived by coupling the two regression models to quantify Peridinium cell density (R2 = 0.71, p < 0.001). We highlight that this method would be a novel approach that directs reliable and accurate prediction and quantification of carotenoid pigments and Peridinium cell density in freshwaters.