基于光学与雷达影像变化检测的2020年鄱阳湖洪灾评估与分析

利用遥感对地观测技术对2020年夏季鄱阳湖地区发生的洪水灾情进行评估和分析，分别选取Landsat 8卫星4月和 Sentinel-1A雷达卫星7月的影像作为洪水发生前、后的影像，利用遥感专题信息提取、随机森林分类和变化检测技术揭示洪水淹没范围，以及被淹没的主要土地覆盖类型面积，同时结合水文气象和地形数据进行灾情和受灾原因分析。研究结果表明，鄱阳湖2020年洪灾的淹没面积为1961.95 km2，共造成区内110.83 km2建筑用地、760.54 km2耕地、71.59 km2林地、992.02 km2草地和26.97 km2裸地被淹，其中尤以鄱阳县受灾最为严重，其总受灾面积达到514 km2，其次为新建区与余干县，分别达到了330 km2和310 km2。持续2个月的强降雨使得鄱阳湖流域的水位超过了其1998年特大洪水的水位，加上长江水倒灌，湖区地势北高南低，积水不能及时排出，圩堤决口等诸多水文气象和地形因素造成了此次鄱阳湖的特大洪灾。

Assessment and Analysis of the 2020 Poyang Lake Flood Hazard based on Optical and Radar Image Assisted Change Detection

In the summer of 2020, heavy rains made the Poyang Lake region witnessed a big flood event. Remote sensing earth observation technology can help to map and assess the flood hazard quickly and efficiently. Therefore, two 2020 satellite images acquired on April 15 (Landsat-8 OLI) and July 14 (Sentinel-1A SAR) were selected in this study to represent the dates before and after the flood to evaluate the disaster. Using remote sensing thematic information extraction, Random Forest classification and change detection technology, the inundation area and the area of major land cover types within the inundation areas were revealed. Associated with the hydrological, meteorological and topographical data, the specific flooded sites and the factors causing the disaster were identified and analyzed. The results show that the flood-inundation extent in the Poyang Lake region in 2020 is 1961.95 km2, including 760.54 km2 of farmland, 71.59 km2 of forest, 992.02 km2 of grassland, 26.97 km2 of soil, and 110.83 km2 of built-up land. Poyang County was most severely affected in this flood event, with a total inundated area of 514.35 km2. The next two are Xinjian County with 330 km2 and Yugan County with 310 km2. The main hydro-meteorological and topographical factors that caused the flood are considered to be: (1) higher water level than that in the 1998 flood; (2) failure of timely discharge of water due to backflow of the Yangtze River; (3) breach of the embankments.