生态调查与评估基于深度学习的白洋淀水体动态变化分析

Mapping long-term water dynamics of Baiyangdian Lake via deep learning method

白洋淀位于雄安新区建设范围的核心区域,对白洋淀水面面积变化趋势进行有效的监测和分析,能够对白洋淀流域的水资源保护提供重要的参考价值。本研究以1990—2020年的Landsat系列遥感影像为数据源,基于深度学习模型WatNet对研究区进行像素级的地表水提取,并将地表水分为永久水体、间歇性水体和非水体,进行基于像元尺度的白洋淀地表水时空变化制图,获得了不同时期的地表水类型动态变化专题图。研究结果表明:1)白洋淀的水面面积经过持续性减小→保持稳定→缓慢增加的3个过程;2)自然或人工补水是维持白洋淀水域的重要影响因素,多次引水济淀措施有效缓解了白洋淀水资源枯竭危机;3)白洋淀水域地表水类型变化趋势先变差后向好发展,部分区域通过补淀措施得到生态修复,但有小部分区域发生了地表水永久流失现象。

Baiyangdian Lake is located in the core area of the Xiong'an New District.Effective monitoring and analysis the changes of Baiyangdian Lake can provide important reference for the protection of water resources.In this study,Landsat series remote sensing images from 1990 to 2020 were used,and a novel deep learning model named WatNet was used to extract water surface at the pixel level.The temporal and spatial changes of water surface in Baiyangdian Lake were obtained,and dynamic changes over different periods were analyzed.The results show that:1)The water surface area of Baiyangdian Lake decreased slowly before 2000,and then kept low from 2000 to 2012,and increased slowly after 2012;2)Natural or artificial water replenishment are important factors in maintaining the water area of Baiyangdian Lake.In the past 30 years,artificial water replenishments had effectively alleviated the crisis of water resources depletion in Baiyangdian Lake;3)The trend of Baiyangdian Lake changes from deteriorated to better.Some areas have been ecologically restored through effective replenishmen,while permanent surface water loss has occurred in a small part of the area.