Developing and evaluating Impact-based Forecasting (IbF) for flooding

概述：洪水对全球人口的影响超过任何其他自然灾害，洪水事件的数量正在增加，最近20年期间（2000-2019年）的洪水事件数量是前20年的两倍多。在印度，极端风暴的频率在过去70年里增加了三倍，风暴的强度和洪水的发生也在增加。再加上人口增长、城市快速扩张和面临洪水风险的资产价值，近几十年来，洪水对社会经济的影响急剧增加，而在气候变化的影响下，这种影响只会加剧。改进洪水（风险）预报已被确定为提高印度的洪水抵御能力和支持有效的洪水风险管理的关键优先事项。国际研究界和研究机构一直在朝着基于影响的预报方法发展，这种方法不仅预测灾害，而且还预测灾害的潜在影响，以帮助决策——正如世界气象局所建议的那样。关键的挑战是如何考虑水文气象预报中的不确定性，以及如何开发和评估IbF端到端系统，以确保输出可由用户操作。水文气象灾害综合预报是水文气象灾害综合预报的一个基本组成部分，但如何最好地形成这些综合预报以及如何对其进行评价仍是有待研究的问题。其次，评估和交流基于影响的预测技能是一个新兴的研究领域，需要新的技术。印度洪水的一个特别热点是西南沿海的喀拉拉邦，那里有多山的西高止山脉和向西流入阿拉伯海的集水区。在季风季节（6月至9月），强降雨可能导致严重的洪水，2018年、2019年和2020年的近期事件造成了重大影响。监督小组一直在与英国气象局和业务伙伴NCMRWF（国家中期天气预报中心）和IMD（印度气象部门）合作，为喀拉拉邦开发一个IbF原型系统，这将成为研究的基础，并为与国家气象机构合作提供一个令人兴奋的机会，使研究具有现实效益。方法：由监督小组和印度合作伙伴共同开发的印度洪水灾害影响模型（FHIM-India）将为研究提供框架。它使用栅格对栅格（G2G）分布式水文模型，配置在印度南部约1km尺度上，并结合高分辨率（<30m） HiPIMS（高性能集成水动力建模系统）来模拟城市尺度的洪水淹没。在计算冲击严重程度时，对水深和速度的估计结果可以直接与感受器联系起来。洪水风险矩阵框架结合了集合降雨预报的不确定性，如NEPS-R （NCMRWF区域集合预报系统），以及洪水预报的严重程度。然后在空间（例如病房或村庄区域）和时间（例如预测范围）的适当整合水平上总结风险。该研究将探索和评估可选择的集合预测方法，并为IbF输出开发新的验证方法。需要使用高性能计算环境（例如伯明翰的JASMIN或BEAR）。

Overview: Flooding affects more people globally than any other natural hazard and the number of events is increasing with the recent 20 year period (2000-2019) recording more than double the number of flood event than the previous 20 years. Within India, the frequency of extreme storms have increased three-fold during the last 70 years with the storm magnitudes and occurrence of flooding increasing too. When combined with population growth, rapid urban expansion and the value of assets exposed to the risk of flooding, the social-economic impacts of flooding have increased dramatically over recent decades with this only set to worsen under climate change. Improved flood (risk) forecasting has been identified as a key priority for increasing flood resilience and supporting effective flood risk management in India. The international research community and research agencies have been moving towards Impact- based Forecasting approaches that not only forecast the hazard but also the potential impact of the hazard to help with decision making - as recommended by the World Meteorological Agency. Key challenges are how to account for uncertainty in the hydro-meteorological forecasts, and how to develop and evaluated the IbF end-to-end system to ensure the outputs are actionable by users. Use of ensemble hydro-meteorological hazard forecasts are a fundamental component but there are open research questions as to how best to form these ensemble forecasts and how to evaluated them. Secondly, evaluating and communicating the skill in Impact-based Forecasts is a new and emerging research area that requires novel techniques. A particular hotspot for flooding in India is the south west coastal state of Kerala that has the mountainous Western Ghats and catchments draining westward to the Arabian sea. During the monsoon season (June-September), heavy rainfall can produce serious flooding with recent events in 2018, 2019 and 2020 causing major impacts. The supervising team have been collaborating with the UK Met Office and operational partners NCMRWF (National Centre of Medium Range Weather Forecasting) and IMD (Indian Meteorological Department) to develop a prototype IbF system for Kerala that will form the basis for the research and provide an exciting opportunity to work with national Meteorological Agencies and for the research to have real-world benefits. Methodology: The Flood Hazard Impact Model for India (FHIM-India), co-developed by the supervising team and Indian partners, will provide the framework for the research. It uses the Grid-to-Grid (G2G) distributed hydrological model configured at ~1km scale over southern India and combined with the high-resolution (<30m) HiPIMS (High-Performance Integrated hydrodynamic Modelling System) to model city-scale flood inundation. The resulting estimates of water depth and velocity can be linked directly to receptors in the calculation of impact severity. The Flood Risk Matrix framework combines uncertainty from ensemble rainfall forecasts, such as NEPS-R (NCMRWF Regional Ensemble Prediction System), together with the severity of flooding forecast. The risk is then summarises at appropriate levels of integration in space (e.g. Ward or Village area) and time (e.g. forecast horizon). The research will explore and evaluate alternative ensemble forecasting methods and develop novel verification approaches for IbF outputs. Use of HPC environments (e.g. JASMIN or BEAR at Birmingham) will be required.