Spatial and temporal analysis of chlorophyll-a using sentinel-2 data at Gajah Mungkur Reservoir, Indonesia

A water quality monitoring program typically employs a direct measurement method, which requires substantial efforts and resources. One of the option to minimised these efforts is utillize satelite imagery for estimating chlorophyll-a concentration. Sentinel-2 with its vast multi-spectral imagery is capable and broadly used for retrieving chlorophyll-a concentration. This study provides validation of five chlorophyll-a retrieval algorithm to identify the best-fitting model for Gajah Mungkur Reservoir in Indonesia. The chlorophyll-a model predictor was developed using four pre-processed Sentinel-2 images taken over Gajah Mungkur Reservoir calibrated with 7 in-situ datasets where chlorophyll-a ranged from 0.17 mg/m3 to 33.96 mg/m3 and mean of 10.16 mg/m3. Five algorithms of different chlorophyll-a Band Ratio Predictor was evaluated with multiple validation parameters. One of the models built incorporating algorithm which use Band B4, B5, and B6 for chlorophyll-a retrieval achieved Pearson R coefficient of 0.93, coefficient of determination (R2) = 0.87, Normalized Root Mean Square Error (NRMSE) = 0.14, and Nash Sutcliffe Efficiency (NSE) = 0.87. This model evaluation has shown better accuracy compared to the other four algorithms. The results indicate promising potential of developing a model for accurate chlorophyll-a retrievals utilizing Sentinel-2 imagery, providing valuable tool for water quality monitoring system in Gajah Mungkur Reservoir, thus minimizing the needs of extensive onsite sampling which then can be used only for local calibration.