Characterizing water body changes in Poyang lake using multi-source remote sensing data

Abstract

This study analyzed changes in water body characteristics of Poyang Lake using remote sensing data (Sentinel-2, Sentinel-3 and ICESat-2) and machine learning methods (random forest, support vector machine and classification and regression tree). Furthermore, the relationship between the water area and water level was analyzed. The main findings are as follows: (1) the peak water range in 2016 and 2017 was significantly larger than in 2018 and 2019. The water extent is large in the summer (maximum ∼2650 km2 in July 2016) and small in the winter (minimum ∼600 km2 in February 2017); (2) the spatial scale of water level fluctuated widely, while the temporal scale showed seasonal characteristics. Water level showed seasonal characteristics, with the higher water levels in summer (20.5 m in July 2016) and lower water levels in winter (8.8 m in October 2019); and (3) there was a strong agreement between water level and extent from 2016 to 2019, for which two models fitted by the logarithmic method had a small root mean square error (R12 = 0.67, RMSE1 = 9.02; R22 = 0.55; RMSE2 = 4.88).