Time-series surface water enhancement using water occurrence in high-elevation areas

Optical satellite remote sensing imagery is frequently utilized for monitoring surface water, owing to its high resolution, continuous observational capabilities, and easily accessibility. However, these images are easily contaminated by clouds, cloud shadows, and terrain shadows, leading to missing data in the images and hindering the generation of long-term surface water time series data. Here, we have developed an Elevation and Water Occurrence-based Area Enhancement (EWOAE) algorithm, which applies water occurrence in high-elevation areas identified through cumulative distribution function, to reconstruct the contaminated pixels in images. The performance of the EWOAE algorithm was evaluated in six study areas worldwide using the Global Surface Water dataset from the Joint Research Centre. The visualization comparison results showed that the EWOAE algorithm can effectively fill the data gaps formed by clouds and other adverse conditions in the images. After applying the algorithm, the correlation between the surface water area time series and the water level time series in all study areas has been improved (e.g., the highest coefficient of determination R2 = 0.98 in the Mead Lake). Compared with two representative algorithms based on water occurrence threshold, the performance of the EWOAE algorithm achieved significant improvement, with a lower mean relative deviation (-0.43%) and mean absolute relative error (2.54%). Moreover, we reconstructed the global reservoir monthly surface area (GRMSA) dataset based on this algorithm and revealed a significant increasing trend in global reservoir surface area between 1984 and 2021. The proposed algorithm can advance surface water dynamics research, supporting freshwater management aligned with Sustainable Development Goal 6.