Developing Innovative Flood Frequency Estimation for a Resilient nation (DIFFER)

洪水是人类历史上一直在与之斗争的一种极其有害的自然灾害。时至今日，在任何特定地区消除洪水风险仍是不可能的，但可以做很多事情来更好地了解洪水发生的频率和程度之间的关系，并利用这些信息使人类住区更具弹性。要减轻气候变化对极端事件的影响并制定有效的防洪计划，就需要更好地了解频率-震级关系以及这种关系在未来可能如何变化。拟议的项目通过开发用于洪水频率分析的先进统计模型来解决这一紧迫的科学和社会挑战，即研究在任何给定年份中预计超过给定规模的事件的频率。目前，这种分析通常是基于系统测量的年最大峰值（即任何一年记录的最大事件）的一系列。对于大样本来说，这种类型的分析得到了完善的理论结果的证明，但由于在任何给定年份只使用一个数据点，因此浪费了长期记录中可用的信息。另一种方法是基于更现代和成熟的极端事件理论，基于对河流流量记录中超过一定高阈值的数据点的分析，因此可以被认为是极端的。这类数据的使用在水文学中一直被忽视，部分原因是从连续流量记录中提取阈值超出的障碍，以及将分析的输出转化为传统上用于设计和维护基础设施的年度超出概率的信息所需的不太直观的步骤。然而，正如2016年洪水国家评估所证明的那样，越来越多的人需要更先进的工具来获得更精确（即更少的不确定性）的洪水概率估计，并评估由于气候变化和其他外部因素（如土地利用变化）导致的流量记录变化的存在。基于阈值超出分析的新统计方法的引入有可能成为洪水频率分析的一个阶梯式变化。先前的研究表明，这种类型的数据可以提供较少变化的估计，并且对于检测集水区变化对洪水产生过程的影响很有用。拟议的研究项目将调查如何利用这类数据对全国范围内洪水行为的变化和洪水频率的总体估计做出更可靠的估计。首先，将评估国家河流流量档案中现有阈值超标记录的适用性，并确定高质量的记录。其次，使用最高质量的记录，将调查数据中存在的趋势，并仔细审查外部驱动因素（如降雨或气候变率指数）解释任何已检测到的变化的潜在能力。在进行这一分析的同时，还将研究阈值超过的性质可能发生的变化与目前估计年超过概率的模式之间的关系。最后，基于对阈值超出特性的研究结果，提出了全国范围内洪水频率估算的新方法，并与实践者和决策者讨论了其可行性。虽然拟议的项目侧重于洪水频率估计方法，但预计其结果将与其他极端环境的分析相关，例如浪高或阵风。

Flooding is a highly detrimental natural hazard with which man has struggled throughout history. Still today, it is not possible to eliminate the risk of flooding in any given area, but much can be done to gather a better understanding of the relationship between the frequency and magnitude of flooding and use this information to make human settlements more resilient. Mitigation of climate change impacts on extreme events and the development of effective flood protection schemes requires improved understanding of the frequency-magnitude relationship and of how this is likely to change in the future. The proposed project addresses this pressing scientific and societal challenge by developing advanced statistical models for flood frequency analysis, i.e. the study of the frequency with which events of given sizes are expected to be exceeded in any given year. At present, this analysis is typically based on series of the annual maximum (i.e. the largest recorded events for any given year) of systematically measured peak flow. This type of analysis is justified by well-established theoretical results for large samples, but is wasteful of the information available in long term records since only one data point for any given year is used. Another approach, which is based on more modern and well-established theory for extreme events, is based on the analysis of the data points in the river flow record which have exceeded a certain high threshold and can therefore be considered to be extreme. The use of this type of data has been neglected in hydrology, due partially to the hurdle of extracting the threshold exceedances from the continuous flow record and to the less intuitive steps needed to translate the outputs of the analysis to information on the annual exceedances probabilities traditionally used in the designing and maintaining of infrastructures. Nevertheless, as evidenced in the 2016 Flood National Review, there is a growing request for more advanced tools to obtain a more precise (i.e. less uncertain) estimation of flood probabilities and to assess the presence of changes in the flow records attributed to climate change and to other external factors (e.g. land-use changes).The introduction of new statistical methods based on the analysis of threshold exceedances has the potential to be a step-change in flood frequency analysis. Previous research has shown that this type of data can deliver less variable estimates and is useful for detecting the effect of changes in the catchment on the flood generating processes. The proposed research project will investigate how to make use of this type of data to make more reliable estimates of both changes in the behaviour of floods and in flood frequency estimation in general across the whole country. At first the suitability of the existing threshold exceedances records from the National River Flow Archive will be assessed and gauged records of high quality will be identified. Secondly, using the highest quality records, the presence of trends in the data will be investigated and the potential ability of external drivers such as rainfall or climatic variability indices to explain any detected changes will be scrutinised. In parallel to this analysis, the relevance of the potential detected changes in properties of the threshold exceedances in relation to the current models for the estimation of annual exceedance probabilities will also be studied. Finally, based on the findings on the properties of threshold exceedances, new methods for flood frequency estimation at a national scale will be proposed and their feasibility discussed with practitioners and decision makers.Although the proposed project is focussed on methods for flood frequency estimation, it is expected that its outcomes will be relevant for the analysis of other environmental extremes, for example wave heights or wind gusts.