High performance of chlorophyll-a prediction algorithms based on simulated OLCI Sentinel-3A bands in cyanobacteria-dominated inland waters

In this research, we have investigated whether the chlorophyll-a (chl a) retrieval algorithms based on OLCI Sentinel-3A bands are suitable for cyanobacteria-dominated waters. Phytoplankton assemblages model optical properties of the water, influencing the performance of bio-optical algorithms. Understanding these processes is important to improve the prediction of photoactive pigments in order to use them as a proxy for trophic state and harmful algal bloom. So that, both empirical and semi-analytical approaches designed for different inland waters were tested. In addition, empirical models were tuned based on dataset collected in situ. The study was conducted in the Funil hydroelectric reservoir, where chl a ranged from 2.33 to 208.68 mg m−3 in May 2012 (austral fall) and 4.37 to 306.03 mg m−3 in October 2012 (austral spring). OLCI Sentinel-3A bands were tested in existing algorithms developed for other sensors and new band combinations were compared to analyze the errors produced. Normalized Difference Chlorophyll Index (NDCI) exhibited the best performance, with a Normalized Root Mean Square Error (NRMSE) of 9.30%. Result showed that wavelength at 665 nm is adequate to estimate chl a, although the maximum pigment absorption band is shifted due to phycocyanin fluorescence at approximately 650 nm.