Fine-scale spatiotemporal characterization of chlorophyll-a in Hong Kong’s coastal waters through the fusion of multisource satellite imagery

Coastal water quality in Hong Kong faces challenges from nutrient pollution, contaminant discharge, algal blooms, land-derived sedimentation, and climate change effects. To assess water quality in coastal regions, researchers have utilized varying in-situ monitoring data and remote sensing techniques to quantitatively estimate chlorophyll-a concentrations. Despite these efforts, there remains a lack of fine-scale characterization of the spatial and temporal patterns of chlorophyll-a in Hong Kong’s coastal waters, due to the limited resolutions and cloudy covers. Our study seeks to bridge this gap by fusing multisource satellite observations. Existing spatiotemporal image fusion models are primarily designed for land surface reflectance. In this study, we propose using Generative Adversarial Networks (GANs)-based deep learning techniques to improve the fusion of multisource satellite images specifically for watercolor remote sensing. A spatiotemporal fusion deep learning framework based on GANs has been developed to generate daily surface reflectance and temperature at 300 m spatial resolution using data from Sentinel-3 and Himawari-8 satellites. Furthermore, we have devised a random forest-based chlorophyll-a concentration estimation model  that employs the blended high-resolution data derived from multi-source satellite observations and extensive in-situ monitoring data obtained from 76 marine water quality monitoring stations administered by the Hong Kong Environmental Protection Department (HKEPD). These independent in-situ monitoring datasets also serve as valuable resources for evaluating the performance of satellite-derived chlorophyll-a concentrations. Consequently, we conducted a fine-scale mapping of chlorophyll-a distribution in Hong Kong's coastal waters to analyze spatiotemporal characterization of water quality. This approach holds promise for real-time, fine-scale water quality monitoring using high-frequency satellite observations in the future.