基于多源卫星数据的洪涝灾害监测研究

【作者】 张磊；

【导师】 丘仲锋；

【作者基本信息】 南京信息工程大学 ， 海洋气象学， 2022， 硕士

【摘要】 随着气候增暖,极端降水事件导致的洪涝灾害日益频繁。我国是世界上洪涝灾害发生最频繁的地区之一,洪涝灾害严重威胁着我国人民的生命及财产安全。针对洪涝灾害开展的监测评估工作以提高防灾减灾效率具有重要意义。利用传统人工调查方法监测洪涝灾害存在周期长、具有危险性、耗费大量资源等缺陷,而卫星遥感技术具有覆盖范围广、时效性高、信息量大等优点,能够弥补传统技术手段的不足,已在洪涝灾害监测评估、水资源调查等领域得到广泛应用。如今基于极轨卫星资料的洪涝灾害监测研究已经比较成熟,借助高空间分辨率的影像能够实现洪水淹没范围的精确提取,然而极轨卫星的重访周期动辄数天,目前还无法完全满足洪涝灾害监测工作对于时效性的要求。我国新一代静止气象卫星FY-4A搭载的AGRI(Advanced Geosynchronous Radiation Imager)传感器覆盖了可见光、近红外、中红外以及长波红外波段,其时空分辨率较高,在洪涝灾害监测工作中能够有效弥补极轨卫星重访周期较长的缺陷。本文阐述了基于卫星遥感进行水体监测的理论和方法,分别建立了极轨卫星和静止气象卫星的洪涝灾害监测模型,主要研究内容及结果如下:(1)基于极轨卫星数据的多种水体指数模型适用性评估及洪灾监测应用。基于Sentinel-2影像资料,对比分析了10种常见波段指数模型的水体提取效果,构建了一种新的水体指数模型WSSI(Water Spectral Similarity Index)。以结合目视解译的高分辨率影像水体识别结果作为真值,运用F1-score、虚警率和漏检率三个指标,从水体识别准确率、水陆混合像元检测能力和阈值稳定性三个方面对所有模型进行了评价。采用WSSI算法对长江中下游洪涝灾害情况进行研究,分析了洪涝灾害面积的变化趋势。结果表明WSSI稳定的最佳分割阈值保证了其在大范围水体信息识别中的准确性;近三年中,2020年洞庭湖、鄱阳湖和太湖的洪涝灾害情况最严重,水体范围出现明显扩张。(2)静止气象卫星数据的去云处理及洪涝灾害监测应用。使用多时相快速合成算法完成气象卫星数据的去云处理,进而得到研究区发生洪涝灾害时的晴空影像,以便进行后续水体提取以及灾情评估工作。针对FY4A AGRI的波段特点改进了前人在多时相快速合成算法中使用的云检测模型,实现了去云影像质量的提升。基于去云处理后的研究区域晴空影像,采用算法评估结果中适用于气象卫星数据的算法完成了洪水淹没范围的提取,结果表明基于气象卫星数据提取的洪水信息与高分辨率资料的水体信息一致性显著。本文的工作一方面解决了水体识别研究中面对复杂场景时最佳分割阈值会发生波动的问题,另一方面在一定程度上克服了云层对基于光学传感器影像进行地表水体提取工作的干扰问题,能够实现针对洪涝灾害的实时广域监测,为洪涝灾害的防治工作提供基础支撑,为相关部门了解具体受灾情况、制定宏观决策提供基础数据支撑。更多还原

【Abstract】 As global warming is intensified,extreme precipitation events become more frequent and result in flood disasters throughout the world.Since China is one of the areas with the most frequent flood disasters in the world,which seriously threatens the life and property safety of Chinese people.Thus,monitoring and extraction of surface water areas in real-time is of paramount importance.Investigation method using the traditional artificial monitoring flooding exist cycle is long,dangerous,spend a lot of defects such as resources,and satellite remote sensing technology has a wide range,high efficiency and large amount of information,can make up for the inadequacy of traditional technology,the monitoring and evaluation has been flooding,water resources research in areas such as widely used.Today,the algorithms for flood disaster monitoring from polar-orbiting satellites has been quite mature.With the help of high spatial resolution images,accurate extraction of flood submerges can be realized.However,the revisit cycle of polar-orbiting satellites often varies from several days,and they are unable to deal with the flood disasters that occur in the data vacancy period.Feng Yun-4A(FY-4A)is China’s new generation of geostationary meteorological satellite,it carries onboard Advanced Geosynchronous Radiation Imager(AGRI)and allows for observing the earth atmosphere at visible,near infrared,mid-infrared and longwave infrared wavelengths.Its high temporal and spatial resolution can effectively compensate for the long revisit period of polar-orbiting satellites in flood disaster monitoring.This thesis develops the theory and method of water monitoring using multi-source satellite data and establishes a flood disaster monitoring through both polar-orbiting and geostationary meteorological satellites.The main research contents and results of this thesis include:(1)Evaluation of multiple water index models based on polar-orbiting satellite data.Based on Sentinel-2 image data,a new Water Spectral Similarity Index(WSSI)model is constructed by comparing and analyzing the extraction effects of 10 common band index models.The water recognition results of high-resolution images combined with visual interpretation are used as the truth value,and f1-Score,false alarm rate and missed detection rate are used to evaluate all models from three aspects: water recognition accuracy,water-land mixed pixel detection ability and threshold stability.The flood disaster situation in the middle and lower reaches of Yangtze River is studied by WSSI algorithm,and the change trend of flood disaster area was analyzed.The results show that the optimal segmentation threshold of WSSI ensures the accuracy of WSSI in large range water body information recognition.In the past three years,dongting Lake,Poyang Lake and Taihu Lake experienced the most severe flood disaster in 2020,and the water area expanded significantly.(2)Cloud removal processing of meteorological satellite data and flood disaster monitoring application.Multi-temporal rapid synthesis algorithm is used to complete cloud removal of meteorological satellite data,and then clear sky images of flood disasters in the study area are obtained,so as to facilitate subsequent water extraction and disaster assessment work smoothly.According to the band characteristics of FY4 A AGRI,the cloud detection model used by predecessors in multi-temporal rapid synthesis algorithm is improved to realize the improvement of cloud removal image quality.Based on the clear sky images of the study area obtained by cloud removal,the algorithm applicable to meteorological satellite data in the algorithm evaluation results is used to complete the extraction of flood submergence area.The results show that the flood information extracted from meteorological satellite data is significantly consistent with the water information extracted from high-resolution data.The work of this paper on the one hand,solve in the research of water body recognition in the face of complex scene will happen when the best segmentation threshold fluctuation problem;on the other hand,to a certain extent,overcome the cloud based on optical image sensor interference problems in the extraction,the surface water can be achieved against floods of wide-area real-time monitoring,provide the foundation for flood prevention and control work,It provides basic data support for relevant departments to understand the specific disaster situation and make macro decisions.更多还原