Deep learning-based super-resolution for harmful algal bloom monitoring of inland water

Inland water frequently occurs during harmful algal blooms (HABs), rendering it challenging to comprehend the spatiotemporal features of dynamics. Recently, remote sensing has been applied effectively detect behaviors in expensive bodies. However, image sensor resolution limitation can render understanding relatively small bodies challenging. In addition, few studies have improved images investigate inland quality, owing limitations. Therefore, this study deep learning-based Super-resolution for transforming satellite imagery 20 m airborne 5 m. After performing atmospheric correction acquired images, we adopted super-resolution (SR) methodologies using a convolutional neural network (SRCNN) and generative adversarial networks (SRGAN) estimate Chlorophyll-a (Chl-a) concentration Geum River South Korea. Both methods generated SR with reflectance at 665, 705, 740 nm. Then, two band-ratio algorithms 665 nm wavelengths were Chl-a maps. The SRCNN model outperformed SRGAN bicubic interpolation peak signal-to-noise ratios (PSNR), mean square errors (MSE), structural similarity index measures (SSIM) validation dataset 24.47 (dB), 0.0074, 0.74, respectively. maps from provided more detailed spatial information on compared obtained images. these findings showed potential by providing further according dynamics management