INSURANCE and WATER: Estimating uncertainty in future flood risk analysis for insurance and re-insurance markets

保险业迫切需要对预测的气候变化导致的洪水风险未来可能发生的变化进行前瞻性预测，以便就风险管理做出长期战略决策，允许进行未来的财务规划，并提高业务对任何增加的大型事件的应变能力。通常，洪水保险损失被视为影响利润率的自然减损事件，但不太可能导致公司破产和无法履行保险和再保险保单义务。然而，大洪水事件的频率和事件集群的变化可能会改变这种情况，并增加对保险市场造成严重破坏的可能性。重新确定保险组合的优先次序和谈判新的再保险合同以应对这种威胁，必须以现有证据为基础，并需要有长远的眼光，因为这种变化需要相当长的时间才能实施，并可能产生大量额外费用。然而，目前缺乏对气候变化对未来洪水风险的影响的强有力的科学理解，因为全球和区域气候模型输出变量(降雨量、温度和蒸散量)的不确定性的影响尚未一直级联到洪水淹没模型。特别是，这种分析尚未尝试用于大多数风险资产所在的城市环境，因为这里需要精细的水力模拟的计算成本。更有甚者。虽然已经开发了将可能性分配给不确定的RCM和GCM输出的方法(例如，Rougier，2007)，但这些方法需要扩展到级联可能性，直到水力预测，以便可以产生似然加权淹没图。为此，总部位于伦敦的全球领先再保险公司Willis Re与布里斯托尔大学和伦敦国王学院接洽，考虑研究这一问题的方法。这种工作非常适合案例学生资助，因为它建立在现有研究的基础上，这些研究已经证实了可行性。它还需要大量的专注于个人研究的时间，以及案例奖所促进的产业界和学术界之间的密切合作。因此，此类研究的下一步是：1.对GCM和RCM输出进行进一步分析，以更好地了解此类模型在模拟当前和未来极端降雨方面的优势和局限性，并确定为这些模型分配可能性的最佳可能方法。2.检验偏差校正是否可以减少气候模式预测降雨的系统时空误差，并增加对这些输出的置信度。3.通过城市和区域尺度的集水区水文学和水力模型，将不确定的降雨量估计串联起来，并测试似然加权方法的使用，以生成不确定的未来洪水风险图。4.与保险业紧密合作，确保以最适当的形式传达这些复杂的分析，并分析这些地图对长期策略决策的潜在影响。学员的成果将是：(1)评估未来洪水风险的新工具，适用于广泛的问题，主要包括保险部门；(2)对洪水风险未来可能发生的变化及其相关不确定性有新的科学理解；(3)注重复杂模型输出的可视化和交流，以确定如何最有效地将已形成的理解传递给行业用户。这本身就代表着高质量的科学，可能会对英国工业产生重大影响。这名学生将由布里斯托尔大学的保罗·贝茨教授和吉姆·弗里尔博士、伦敦国王学院的汉娜·克洛克博士和威利斯再保险公司的马特·富特指导。

A forward view of likely future changes in flood risk resulting from predicted climate change is critically required by the insurance industry in order to take long term strategic decisions on risk management, allow future financial planning and improve business resilience to any increased frequency of large events. Typically, flood insurance losses are regarded as attritional events that impact on profit margins but which are unlikely to result in company bankruptcies and an inability to meet both insurance and reinsurance policy obligations. However, changes to the frequency of large flood events and event clustering may alter this situation and increase the potential for severe disruption to insurance markets. The re-prioritization of insurance portfolios and negotiation of new re-insurance contracts to deal with such threats must proceed on the basis of available evidence and requires a long term view as such changes take considerable time to implement and may incur substantial additional costs. A robust scientific understanding of the impact of climate change on future flood risk is, however, currently lacking because the impact of uncertainty in Global and Regional Climate Model output variables (rainfall, temperature and evapotranspiration) has yet to be cascaded all the way through to flood inundation models. In particular, such analyses have yet to be attempted for urban settings where the majority of at-risk assets are located because of the computational cost of the fine scale hydraulic simulations that are here required. Moreover. whilst methods have been developed to assign likelihoods to uncertain RCM and GCM outputs (e.g. Rougier, 2007) these require extension to cascade likelihoods through to hydraulic predictions such that likelihood-weighted inundation maps can be produced. For this reason University of Bristol and King's College London were approached by Willis Re, a leading global re-insurance firm based in London, to consider ways to research this issue. Such work is ideally suited to CASE studentship funding as it builds on existing research which has confirmed feasibility. It also requires both the substantial period of focussed individual research and close collaboration between industry and academics that a CASE award facilitates. The next steps for such research are therefore to: 1. Undertake further analysis of GCM and RCM output to better understand the strengths and limitations of such models for simulating current and future extreme rainfalls and determine best possible methods to assign likelihoods to these. 2. To test whether bias correction can reduce systematic spatio-temporal errors in climate model predicted rainfalls and increase confidence in these outputs. 3. To cascade uncertain rainfall estimates through catchment hydrology and hydraulic models for urban settings and at regional scales, and test the use of likelihood weighted methods to produce uncertain future flood risk maps. 4. To work closely with the insurance industry to ensure that such complex analyses are communicated in the most appropriate form and analyse the potential impact of such maps on long term strategic decision making. The outputs from the studentship will be: (1) new tools for assessing future flood risk with applicability to a wide range of problems including, primarily, the insurance sector; (2) new scientific understanding of likely future changes in flood risk and their associated uncertainty; and (3) a focus on visualization and communication of complex model outputs to determine how the developed understanding can be most effectively transferred to industry users. This represents high quality science in its own right, with potentially significant impact for UK industry. The student will be supervised by Professor Paul Bates and Dr. Jim Freer of University of Bristol, Dr. Hannah Cloke of King's College London and Matt Foote of Willis Re.