基于卫星遥感数据的任意地区双变量降雨事件概率模型构建与应用

Establishment and Application of Bivariate Probability Model for Rainfall Events in Arbitrary Regions Based on Satellite Remote Sensing Data

摘要:

滑坡是常见的地质灾害，降雨是主要诱发因素。构建精确的降雨事件概率模型对于准确识别滑坡灾害的高风险区以及制定科学合理的防灾减灾策略至关重要。然而，由于缺少降雨数据，当前仅少数地区建立了双变量降雨事件概率模型。本文提出了一种基于卫星遥感降雨数据的任意地区双变量降雨事件概率模型构建方法。首先， 通过网络爬虫技术批量自动获取全球任意地区的卫星遥感降雨时间序列。以临海市邵家渡街道为例，依据降雨事件间隔时间，将连续降雨时间序列划分为独立降雨事件，提取降雨量和降雨历时等变量。在此基础上，基于Copula 函数建立双变量降雨事件概率模型，并利用K-S检验评估模型拟合优度。进一步地，阐述了双变量降雨事件概率模型中重现期的定义，并与单变量降雨事件概率模型分析结果进行了对比。最后，构建了浙江省各地区的双变量降雨事件概率模型，揭示了省内降雨变量的空间分布。建立的双变量降雨事件概率模型为降雨诱发滑坡风险评估提供了精确的降雨数据支撑。

Abstract:

Landslide is a common geological hazard, with rainfall being a major inducing factor. Estab lishing an accurate probability model for rainfall events is crucial for accurately identifying high-risk ar eas of landslides and developing scientific and reasonable disaster prevention and mitigation strategies. However, due to the limited availability of rainfall data, only a few regions have established bivariate probability models for rainfall events. This paper proposes a method for establishing bivariate probabil ity model for rainfall events in any region based on satellite remote sensing data. First, satellite remote sensing rainfall time series from arbitrary regions in the world were automatically obtained using web crawler technology. Taking Shaojiadu Subdistrict in Linhai City as an example, the continuous rainfall time series were divided into independent rainfall events based on inter-event time definition (IETD), and variables such as rainfall amount and duration were extracted. On this basis, a bivariate probability model for rainfall events was established using Copula function, and K-S test was used to evaluate the goodness of fit of the model. Furthermore, the definition of the return period in the bivariate probabili ty model for rainfall events was discussed and compared with the analysis results of the univariate prob ability model. Finally, the bivariate probability model for rainfall events was established for each re gion in Zhejiang Province, revealing the spatial distribution of rainfall variables across the province. The established bivariate probability model for rainfall events provides accurate rainfall data support for the risk assessment of rainfall-induced landslides.