Developing enhanced impact models for integration with next generation NWP and climate outputs

目前最好的估计表明，大约有500万人生活在200万财产面临洪水的风险，造成极端风暴在英国。在这些家庭中，约有200，000个家庭没有受到政府建议的最低保护水平，即每75年发生一次的事件的保护。当发生大洪水时，总损失成本很高（2007年夏季洪水为35亿英镑），每年防洪总支出接近8亿英镑。保护这些人口并最大限度地减少未来的成本需要开发强大的水文和水力模型，将数值天气预报（NWP）和气候模型的输出转化为有意义的影响估计（具有不确定性）。这些影响预测可用于规划投资决策，提供实时预警，设计洪水防御计划，并通常有助于更好地管理风暴风险和减轻危险的气候变化的影响。在联盟成员开发的基础上，作为极端事件的NERC洪水风险（免费）和EPSRC/NERC洪水风险管理研究联盟（FRMRC）计划的一部分，我们在这里提出了一个综合的研究计划，这将导致我们在短期和长期内量化风暴影响的能力逐步改善。根据上述方案中获得的知识，我们建议通过我们独特的定位来实现四个相互关联的目标，以改善风暴影响建模。具体而言，这些是：1.水文模拟结构的尺度缩小、不确定性传播和评价。2.发展数据同化和遥感方法以加强风暴影响模型的预测。3.通过沿海风暴潮模型更好地传播误差。4.开发一种新的水力模型，可用于将洪水径流或沿海极端水位的预测转换为整个城市区域的LiDAR数据分辨率（水平分辨率约1 - 2米）的洪水范围和深度的估计，使用真正的动量守恒方法。在本提案中，我们评估了上述四种方法的潜力，以减少风暴影响的集合预测的不确定性，因为NWP和气候模式输出中的典型误差被用作影响建模链的边界强迫。我们的初步表征的误差预测风暴功能（空间降雨量和风速场）在目前实施的数值预报和气候模式将根据现有的研究进行的英国气象局和阅读大学。随着项目的进行，我们将利用正在为这一呼吁的可预测性1和2开发的风暴建模方面的进展，以增强我们的错误特征，并确保我们开发的技术适用于当前和未来的气象建模技术。我们将通过使用布里斯托和阅读大学正在开发的风暴影响的独特基准数据集，严格评估我们提出的方法的成功。

Current best estimates indicate that approximately 5M people living in 2M properties are at risk of flooding resulting from extreme storms in the UK. Of these approximately 200,000 homes are not protected against a 1 in 75 year recurrence interval event, the Government's minimum recommended level of protection. When major floods do occur then total damage costs are high (£3.5Bn for the summer 2007 floods) and the total annual spending on flood defence approaches £800M. Protecting this population and minimizing these costs into the future requires the development of robust hydrologic and hydraulic models to translate the outputs from Numerical Weather Prediction (NWP) and climate models into meaningful estimates of impact (with uncertainty). These predictions of impact can then be used to plan investment decisions, provide real-time warnings, design flood defence schemes and generally help better manage storm risks and mitigate the effects of dangerous climate change. Building on foundations developed by consortium members as part of the NERC Flood Risk from Extreme Events (FREE) and EPSRC/NERC Flood Risk Management Research Consortium (FRMRC) Programmes, we here propose an integrated programme of research that will lead to step change improvements in our ability to quantify storm impacts over both the short and long term. Based on the knowledge gained in the above programmes, we suggest that improvements in storm impact modelling can be achieved through four linked objectives which we are uniquely positioned to deliver. Specifically, these are: 1. Downscaling, uncertainty propagation and evaluation of hydrologic modelling structures. 2. The development of data assimilation and remote sensing approaches to enhance predictions from storm impact models. 3. Better error propagation through coastal storm surge models. 4. The development of a new class of hydraulic model that can be used to convert predictions of rainfall-runoff or coastal extreme water levels to estimates of flood extent and depth at the resolution of LiDAR data (~1 - 2m horizontal resolution) over whole city regions using a true momentum-conserving approach. In this proposal we evaluate the potential of the above four approaches to reduce the uncertainty in ensemble predictions of storm impact given typical errors in the NWP and climate model outputs which are used as boundary forcing for impact modelling chains. Our initial characterization of the errors in predicted storm features (spatial rainfall and wind speed fields) in current implementations of NWP and climate models will be based on existing studies conducted by the UK Met Office and the University of Reading. As the project proceeds we will use the advances in storm modelling being developed for Deliverables 1 and 2 of this call to enhance our error characterizations and ensure that the techniques we develop are appropriate for current and future meteorological modelling technologies. We will rigorously evaluate the success of our proposed methods through the use of unique benchmark data sets of storm impact being developed at the Universities of Bristol and Reading.