面向暴雨洪涝灾害的海滨区域敏感性要素分析

Analysis of Sensitivity Factors in Coastal Areas for Heavy Rainfall and Flooding

洪涝灾害指标类型多样且大多基于文献调研与经验得到。为了定量化筛选指标且考虑指标间相互作用机制,首先利用岭回归、Lasso回归、弹性网络回归对响应灾害的关键指标进行筛选,发现建成区供水管道密度是响应雨洪灾害损失的最强指标。其次,构建树状高斯过程模型分析特征变量的相互作用情况及各自变化趋势,发现在交互作用下日生活用水量指标响应灾害效应提高了近4倍。随着日生活用水量和生活垃圾无害化处理率的增加,灾害造成的经济损失增加。最后通过随机森林敏感性验证结果。敏感性要素的小幅度变化能导致灾害经济损失的较大变化,故可用作海滨城市洪涝灾害风险考量的抓手,同时敏感性要素可用于进一步分析城市空间分异情况。

Flood hazard indicators are of various types and are mostly based on literature research and experience.To quantitatively screen the indicators and consider the interaction mechanism between indicators,this paper firstly uses ridge regression,lasso regression and elastic network regression to screen the key indicators of response to disasters.It is found that the density of water supply pipes in built-up areas is the strongest indicator of response to rainfall disaster loss.Secondly,a tree Gaussian process model was constructed to analyze the interactions of the characteristic variables and their respective change trends.It was found that the response of daily domestic water consumption indicator to disaster effect increased nearly four times under the interaction.As the daily domestic water consumption and domestic waste disposal rate increased,the economic loss caused by the disaster increased.Finally,the results were verified by random forest sensitivity.Small changes in sensitivity elements can lead to large changes in economic losses from disasters,so it can be used as a risk consideration grip for flooding in seaside cities.Meanwhile,the sensitivity elements can be used to further analyze the spatial differentiation of cities.