回归分析框架下洪涝灾害脆弱性曲线构建方法综合比较研究

脆弱性曲线定量揭示灾害危险性强度与承灾体损失程度的关系,链接灾害的自然属性与社会属性,是灾害经济影响评估和风险管理的重要支撑。当前研究提出了多种洪涝灾害脆弱性曲线,多为基于案例的实证性研究,缺少系统性的理论整理。本文在回归分析理论框架下,对脆弱性曲线模型进行一致性的考证。明晰了损失率脆弱性曲线与损失状态脆弱性曲线的区别与联系,分析两种脆弱性曲线10种不同构建形式的基本假设、优缺点、适用性、拟合方法及选取标准。同时,利用湖北省恩施市“7.17”洪灾调研数据,构建商业停滞损失脆弱性曲线模型,对理论进行了验证分析。研究建议：损失率脆弱性曲线可以更好地链接损失评估与资产计算,在有准确的损失率数据情况下选用。其中,简单线性模型和双对数模型具有更好的参数可解释性,可优先考虑。损失状态脆弱性曲线更容易对应风险分级与对策分类,在损失率数据存在较大不确定性时选用。其中,Lognormal模型具有更好的理论适用性,建议优先考虑。

Comprehensive comparative study on the construction method of flood vulnerability and fragility curves under the framework of regression analysis

Vulnerability/fragility curve reveals the relationship between hazard level and loss degree, and links natural and social attributes of disasters.It is an important basis for disaster economic impact assessment and risk management.Current studies have proposed a variety of flood vulnerability/fragility curves, but most of them are empirical studies based on cases, lacking systematic summary.Therefore, in this paper, under the theoretical framework of regression analysis, the vulnerability and fragility curve models are verified consistently.The difference and relationship between the vulnerability curves and the fragility curves are clarified.The basic assumptions, advantages and disadvantages, applicability, fitting methods and selection criteria of 10 different construction forms of the vulnerability/fragility curves are discussed.Meanwhile, using the "7.17" flood survey data in Enshi City, Hubei Province, the vulnerability and fragility curve models of business stagnation loss are constructed to demonstrate the theoretical summarization.The study shows that vulnerability curve accounting for loss rate can better support the loss assessment and asset estimation.It can be selected if accurate loss rate data available.Linear model and Log-log model are preferred vulnerability curve models due to better parameter interpretability.Fragility curve accounting for loss state is easier to correspond to risk classification and countermeasure making.It is recommended when loss rate data are with large uncertainty.The Lognormal model is preferred fragility curve model due to better theoretical applicability.