Estimation of cyanobacteria pigments in the main rivers of South Korea using spatial attention convolutional neural network with hyperspectral imagery

Although remote sensing techniques have been used to monitor toxic cyanobacteria with hyperspectral data in inland water, it is difficult optimize conventional bio-optical algorithms for individual water bodies because of the complex optical properties various components. Therefore, this study adopted a spatial attention convolutional neural network (spatial CNN) estimate chlorophyll-a (Chl-a) and phycocyanin (PC) concentrations Geum, Nakdong, Yeongsan rivers South Korea order evaluate using reflectance data. The CNN model utilized module analyze importance bands Then, was compared different each area. generalized pigment target rivers, performance evaluated by correlation coefficient (R) root mean squared error (RMSE) between observed estimated algal pigments. model, which had R values above 0.87 0.88 Chl-a PC, respectively. However, optimized band ratio PC 0.83 0.70, Hence, showed better than algorithms. provided weights visualizing important features Specifically, 600 nm, 650 near-infrared regions high estimating PC. Based on these findings, demonstrated that has potential good application bodies.