Analyzing the spatio-temporal behavior of Poyang Lake using Google Earth Engine

Abstract

Poyang Lake, China’s largest freshwater lake, undergoes a yearly repeating cycle of drastic inundation and subsequent considerable shrinkage. Essentially, as a result of local precipitation and feedings from its tributaries, as well as a natural water exchange with the Yangtze River, Poyang Lake experiences such dimensional fluctuations on a annual and interannual scale. The ongoing change plays a significant role for the surrounding anthropogenic activity and wildlife. Despite being considered as a hydrological phenomenon, the dynamics of this Chinese water body set up a hurdle for any accurate documentation of its regime and therefore remains insufficiently studied upon to this day. Further impeding the comprehension of Poyang Lake’s behavior is the near inaccessibility and nonexistence of in situ data, such as water level measurements and bathymetric maps. Consequently, this study, driven by its aim to analyze the spatio-temporal behavior of Poyang Lake, focuses solely on satellite observations. Making use of the cloud computing platform, Google Earth Engine, image time series are used from Landsat-8 and Sentinel-1 datasets in order to map Poyang Lake’s spatio-temporal behavior on an annual and interannual scale. Produced from the Landsat-8 dataset, results show that only under circumstances, do techniques, such as the combination of visible and infrared bandsand the calculation of the Normalized Difference Water Index, provide a reasonable approach for the delineation of continental water bodies. For the study on Poyang Lake, a water body subject to humid climate and thus frequent cloud coverage, these techniques do not apply very well. With synthetic aperture radar observations from the Sentinel-1 dataset, dynamic water masks, involving the removal of certain elevated areas and the classification of water from thresholding, could be generated. The resulting binary water masks are then merged with a digital terrain model to create monthly maps of the study area. The results show an evident correlation between this study’s visual and numeric findings. Although the results are nearly impossible to compare with any in situ data, they show a trend that annually occurs in Poyang Lake’s hydrological regime. In particular, they reveal the cycle of drastic inundation in rainy summer months and considerable shrinkage in dry winter months, especially when examining the years prior to 2019. Using SAR imagery for continental water body delineation, particularly in humid climates, proved to be a suitable technique and should be considered for future documentations of the lake.