Abstract:

Under the influence of climate change, the water area of Poyang Lake exhibits seasonal fluctuations. To elucidate these changes comprehensively, we propose a water information extraction approach that integrates multi-temporal radar images and optical data. Leveraging Sentinel-1A radar images and Sentinel-2 optical images as our research datasets, we initiated a sequence of data preprocessing steps on the remote sensing image set. Employing the Sentinel-1 dual polarized water index (SDWI) and the improved normalized differential water index (MNDWI), we delineated the lake’s boundary and computed its water area using radar and optical data. The precision, timing, and change detection of our water extraction results were evaluated meticulously. Analyzing the lake area’s change trends aims to furnish scientific insights for disaster management and Poyang Lake’s conservation. Our findings revealed that: 1)The water bodies of Poyang Lake extracted from radar and optical remote sensing data were mostly congruent. Radar image extraction outperformed optical image extraction when delineating farmlands, small water bodies, and cloud-covered areas, suggesting radar images’ superiority in capturing comprehensive water information. 2)The average water area of Poyang Lake during normal, wet, and dry seasons is 3 686.49, 4 077.73, and 2 612.81 km², respectively, with the wet season’s water volume being 1.56 times that of the dry season. 3)Time-series water extraction results from radar images exhibited strong consistency with water level variation data from Xingzi Station, Duchang Station, Hukou Station, and Kangshan Station, yielding Pearson correlation coefficients of 0.89, 0.87, 0.90, and 0.81, respectively.

Key words: Poyang Lake, water extraction, spatiotemporal variation, MNDWI, SDWI, change detection