【摘要】 降水是最重要的水循环过程,在全球气候变化影响下,降水过程出现显著时空变异,极端气象事件频发,对我国社会经济发展造成了严重影响.除降水观测数据以外,CMIP5数据已成为未来降水预估研究的重要数据源,而目前CMIP5数据空间分辨率低,如何对CMIP5数据进行降尺度研究,已成为区域降水研究的关键.本文以中国北部农牧交错带为研究区,提出了基于站点的非线性统计降尺度模型(station-based non-linear statistical downscaling model,SNSDM),并得出了以下结论:1)SNSDM降尺度降水序列相对于BCSD(bias corrected spatial downscaling)降尺度降水序列,在同等分辨率的情况下,SNSDM提高了对降水低值的模拟精度,可更为准确地模拟中国北部农牧交错带降水时空特性;2)相比于BCSD降尺度方法,SNSDM模拟结果与实测降水相关性提高最高达1.66%,且明显减少了对实测降水过高估计的误差,最大误差仅为0.2～0.3mm·d-1(每月6～9mm);3)CMIP5降水产品在较为湿润地区对于降水强度及趋势模拟精度要普遍高于对较为干旱地区的模拟精度.本研究提出的SNSDM方法对CMIP5降水数据过高估计实测降水的改进,进一步提高了利用CMIP5数据集对未来气候变化预估的精度及研究结果的可信度.更多还原

【Abstract】 Precipitation is a critical component in the hydrological cycle.Precipitation processes are subject to remarkable spatiotemporal alterations in the backdrop of global climate changes and high frequency extreme weathers.Warming climate-induced amplification of weather extremes has apparent impacts on socioeconomic development.It should be noted here that,other than in situ precipitation observations,CMIP5 precipitation products can act as a major data source for prediction of precipitation changes.Lower spatial resolution of CMIP5 precipitation dataset however,enables only limited theoretical and practical application of CMIP5 precipitation datasets in the evaluation of regional precipitation variations.In the current study,a new stationbased Non-linear Statistical Downscaling Model,SNSDM,was proposed for the agricultural and livestock farming transitional zone in northern China.Downscaling performance of this model was evaluated.It was found that downscaled precipitation data by SNSDM,when compared to Bias Correlated Spatial Downscaling(BCSD),greatly improved the spatial resolution of lower precipitation value and could better describe the spatiotemporal pattern of precipitation regimes across the whole agricultural and livestock farming transitional zone in northern China.In comparison with spatial resolution of BCSD downscaled precipitation datasets,the improvement of spatial resolution was 1.66% with SNSDM.This also resulted in greatly reduced estimation error in downscaled precipitation data when compared to in situ precipitation observations.CMIP5 datasets could better quantify precipitation processes in terms of precipitation intensity and precipitation trends in humid regions when compared to arid regions.In comparison to BCSD,SNSDM could greatly improve prediction accuracy and validity for CMIP5 precipitation datasets.更多还原