Climate and Environmental Risk Analytics for Resilient Critical Infrastructure Finance

联合王国所有受监管的金融机构都要接受所谓的压力测试，即对其资产负债表的特定部分进行一系列繁重的设想，并对其当前生存能力的后果进行评估。在目前的方法中，所使用的压力的大小和来源是基于对金融部门发生的事情的狭隘观察的经验。然而，英国央行（Bank of England）下属的审慎监管局（Prudential Regulatory Authority）等监管机构已将气候变化视为对金融业当前稳定及其长期生存能力的最重大威胁之一。目前关于如何考虑气候风险的建议见naïve。没有关于(a)有害气候情景对金融体系的传导机制和(b)对资产负债表等造成压力的频率和程度的方法。为了解决这个问题，需要开发一种方法，将关键金融机构的资产负债表与气候情景明确联系起来。该项目的重点将是英国金融机构的大量实物资产库存用作贷款抵押品。第一阶段：暴露本研究将考虑的主要暴露情景是洪水和地面不稳定。在大卫·舒尔茨的指导下，学生将考虑一系列强加的（和可能的）未来气候情景，以及这些情景对英国各地河流和海平面的影响。这项演习将用于识别受洪水和地面不稳定威胁最大的地区。这些地点与实物资产的交集将被用来确定具体的案例研究地点。阶段2：响应将考虑三种不同类别的实物资产，并对其实物响应进行精确建模。此练习的目的是评估第1阶段中描述的物理外部行为与建模的物理资产的响应之间的直接联系。第3阶段：脆弱性第1阶段确定的暴露将使用第2阶段开发的基本模型来确定选定的实物资产的脆弱性。在基于绩效的评估中，用脆弱性曲线表示实物资产的脆弱性已成为惯例，脆弱性曲线表示给定暴露强度下的累积失效概率。在数学上，脆弱性曲线将以对数正态累积分布函数的形式表示。阶段4：风险通过将问题视为随机过程来计算所选资产对给定暴露水平的风险。这意味着暴露和脆弱性都将通过它们的概率密度函数来考虑，而不是它们的平均值。此外，通过根据均值、标准差和相关长度定义的概率空间分布，对资产的关键机械特性（如建筑类型、建筑类别等）的固有异质性进行建模也很重要。由于这种过程的随机处理产生了大量的分析，因此将使用拉丁超立方体方法和蒙特卡罗等采样技术来减少总体计算时间。估计的风险将以未来四种气候情景中不可接受的表现概率的累积分布来表示。第五阶段：财务分析在项目合作伙伴JBA Risk Associates的参与下，将使用一个案例研究，使用第一到第四阶段新开发的损失数据来检查一个或两个资产负债表的弹性。将审查当前方法的乐观/保守主义。第1至第5阶段将作为进一步发展更专业的气候相关金融分析的拟议框架提出。

All regulated financial institutions in the United Kingdom are subject to so-called stress tests, whereby specific elements of their balance sheets are subjected to a range of onerous scenarios and the consequences for their immediate viability assessed. In the present approach, the magnitude and source of the stresses used is based upon experience from insular observations of what happens in the financial sector. However governing bodies, such as the Bank of England's Prudential Regulatory Authority, have identified climate change as being one of the most significant threats to the immediate stability of the financial sector, as well as its long-term viability. Current proposals for how climate risks are to be factored in are naïve. There is no methodology surrounding (a) the transmission mechanism of deleterious climate scenarios into the financial system and (b) the frequency and magnitude of the resulting stresses on balance sheets etc. To resolve this, an approach that can explicitly link the balance sheets of critical financial institutions to climate scenarios needs to be developed. The focus of this project will be UK financial institutions with large inventories of physical assets used as lending collateral.Stage 1: ExposureThe primary exposure scenario that will be considered in this study is flooding and ground instability. Working under the supervision of David Schultz, the student will consider a range of imposed (and probable) future climate scenarios and the consequences these will have for river and sea levels across the UK. This exercise will be used to identify areas at greatest risk from flooding and ground instability. The intersection of these locations with the presence of physical assets will then be used to determine specific case study locations.Stage 2: ResponseThree different classes of physical asset will be considered and their physical responses modelled with precision. The purpose of this exercise is to evaluate the direct link between the physical external actions described in Stage 1 and the response of the modelled physical assets.Stage 3: VulnerabilityThe exposures determined in Stage-1 will be used to determine the vulnerability of the selected physical assets using the fundamental models developed in stage-2. In performance-based assessments, it has become customary to express the vulnerability of physical assets in terms of fragility curves, which express the cumulative probability of failure for a given exposure intensity. Mathematically, the fragility curves will be expressed in the form of lognormal cumulative distribution functions. Stage 4: RiskThe risk of the selected assets to a given level of exposure will be computed by treating the problem as a stochastic process. This implies that both esposure and vulnerability will be considered through their probability density function, rather than their mean values. Furthermore, it will be important to model the inherent heterogeneity of key mechanical properties of the assets, eg, construction typology, building class etc., through probabilistic spatial distributions defined in terms of mean, standard deviation and correlation lengths. As this stochastic treatment of the process produce a large number of analyses, sampling techniques such as Latin-hypercube method and Monte-Carlo will be used to reduce the overall computation time. The risk estimated will be expressed in terms of cumulative distribution of probability of unacceptable performance for four future climate scenarios.Stage 5: Financial analyticsA case study will be used, with the involvement of project partner JBA Risk Associates, to examine the resilience of one or two balance sheets using the newly developed loss data in Stages 1 to 4. The optimism/conservatism of current approaches will be examined. Stages 1 to 5 will be presented as a proposed framework for developing further, more specialised climate dependent financial analytics.