基于自然语言识别的上海市报警灾情数据识别及其气象灾害特征分析研究  

Research on recognition of alarm disaster data and analysis of meteorological disaster features in Shanghai based on natural language recognition algorithm

作　　者：杨辰 潘顺 严岩 YANG Chen;PAN Shun;YAN Yan(Shanghai Center for Meteorological Disaster Prevention Technology,Shanghai 200030,China)

机构地区：[1]上海市气象灾害防御技术中心,上海200030

出　　处：《自然灾害学报》2021年第3期142-150,共9页Journal of Natural Disasters

基　　金：上海市2020年度“科技创新行动计划”社会发展科技攻关定向项目(20DZ1200403);上海市气象局业务类科研项目(MS202008)共同资助。

摘　　要：上海作为特大型城市,其城市运行极易受到气象条件影响。受制于传统的灾害上报和普查渠道,灾情数据量较为有限且时效性不强。研究基于110气象灾情数据,采用自然语言处理算法,对报警灾情描述信息进行分析提取,识别灾种、承灾体和影响程度等信息,并进一步分析了气象灾害的时空特征以及致灾的气象条件。结果表明,该方法可以较好地提取气象灾害的类别、承灾体及影响信息。近10a灾情分布年际差异较大,月分布以8月和10月为最多,小时分布呈现双峰特征,且工作日更为明显。暴雨和大风的灾情空间分布均以中心城区为最高,其次为紧邻中心城区的外环以内区域以及各区的城市副中心。结合气象资料分析,短时强降雨和长时间的较强降雨是引发暴雨灾害的重要因素,瞬时大风往往会引起大风灾害,而台风和持续性强降水是造成灾情大量爆发的主要原因。As a megacity,the urban operations of Shanghai are extremely susceptible to weather conditions.Subject to traditional disaster reporting and census channels,the amount of disaster data is relatively limited and time-consuming.The study is based on 110 meteorological disaster data and uses natural language processing algorithms to analyze and extract alarm disaster description information,identify disaster types,disaster-bearing bodies,and impact degrees,and further analyze the spatial and temporal characteristics of meteorological disasters and the weather conditions that cause disasterc.The results show that the method can extract the meteorological disaster categories,disaster-bearing bodies and impact information.The distribution of disasters in the past 10 years varies greatly from year to year.Monthly distribution has the largest number in August and October.The hourly distribution has a double-peak characteristic,and more obvious in the working day.The spatial distribution of torrential rain and strong wind disasters is the highest in the central city,followed by the area within the outer ring close to the central city and the urban sub-centers in each district.Analysis with meteorological data,short-term heavy rainfall and long-term rainfall are important factors that trigger heavy rain disasters.Instantaneous high winds often cause strong wind disasters,and also typhoon and persistent heavy precipitation are the main reasons for the large number of disasters.

关 键 词：110灾情 LDA TF-IDF 特征分析 

分 类 号：X4[环境科学与工程—灾害防治] X16