基于统计降尺度技术预估我国未来极端降水情景

In the last decades, the frequent occurrence of extreme precipitation events has attracted wide attention owning to their huge impact on human life, the environment, economy, and society. According to projections from current global climate models (GCM), global warming will most likely continue for the future owing to increased greenhouse gas emissions from human activities, which likely leads to more frequent extreme precipitation events in many regions. Thus, understanding and studying historical change of extreme precipitation events and projecting their future scenarios is necessary and important to mitigate effectively dangerous change and cope with the change that we cannot change in China. GCMs are the most important and effective tools for predicting future climate change, however, they have limitations in simulating regional/local climate due to low spatial resolutions. Therefore, reliable and fine-spatial resolution climate change scenarios are required to assess the impact of future changes of extreme precipitation events. A common used downscaling technique is statistical downscaling technique since it needs a low computation and is easily applied. Specially, it can obtain regional climate information on a station scale. However, the current precipitation statistical downscaling approaches perform less perfectly in reproducing the extreme precipitation. Thus, the objective of this study is to improve the skill of statistical downscaling models in modeling extreme precipitation and further project future scenarios of extreme precipitation. The model skill in extreme precipitation in statistical downscaling process will be improved by the following steps. The contribution of large-scale predictors will be assessed to extreme precipitation, and different predictor sets will be chosen for extreme precipitation at each station. Probability distribution function of extreme precipitation at each station will be replaced by GEV or GPV distribution function. Through comparison of different downscaling techniques, the best appropriate downscaling technique will be adopted for each station in China. Finally, a set of future extreme precipitation indices scenarios on a station scale will be developed through applying the above improved models to GCM simulations, in order to provide a technical basis for climate impact assessment concerning regional sustainable development.

近年来极端气候事件频发，严重威胁到人类赖以生存的生态环境，造成巨大的经济损失和人员伤亡，深入理解和研究我国极端降水事件的变化规律并预估其未来可能的变化情景，对评估极端降水事件对人类活动和社会生存环境等可能带来的影响意义深远。全球气候模式（GCM）由于分辨率低，很难满足区域气候影响评估的需要，统计降尺度技术作为重要的降尺度方法不仅计算量少，易于实施，且在站点尺度气候预估方面优势突出，因此得以被国际社会重点关注。然而目前的统计降尺度模型在日尺度极端降水模拟方面存在缺陷。本研究将通过评估大尺度气候因子对我国各台站日尺度极端降水的相对贡献，通过选择最优的预报因子、极端降水分布函数和降尺度方法，提高统计降尺度对我国各台站日尺度极端降水的模拟性能，为我国各台站构建一套能良好表征极端降水特征的逐日降水统计降尺度模型，进而发展一套我国各台站未来21世纪日尺度极端降水情景数据集，为政府决策提供参考依据。

全球气候模式GCMs由于其分辨率较低，很难精细地评估我国区域或台站尺度的逐日降水分布特征，特别是逐日降水统计分布的尾部特征，因而很难精确地预估我国未来极端降水事件变化情景。本项目通过深入分析和考察我国743个台站逐日降水分布特征，发现我国大部分台站的极端降水和一般降水呈现显著不同的统计分布特征，而且它们与大尺度气候之间也具有明显不同的统计关系。因此本项目基于极端降水和大尺度气候的统计联系，构建了适合评估我国极端降水的统计降尺度模型，其结果表明该模型显著改进了对极端降水的模拟能力。这说明项目发展的极端降水统计降尺度模型对预估未来我国极端降水情景来说是一个有效的预估工具。同时通过把降水统计降尺度模型应用于多个GCMs模拟的CMIP6预估情景，集合预估了未来我国的极端降水的变化情景，结果发现尽管不同CGMs驱动，会带来我国极端降水情景预估较大的不确定性，但是集合预估结果表明：本世纪末我国降水强度、极端降水和最大降水量均成显著增加的趋势，特别是高强迫情景（SSP585）情景下增加更为剧烈，这表明未来我国大部分地区，特别是南部地区面临着较大的暴雨洪涝风险，需要加强防范。同时本项目构建的未来极端降水情景数据库对我国未来的防洪减灾政策制定，以及对农业，生态和环境等领域的气候影响评估和可持续发展战略具有重要的指导意义。