青藏高原多源雪深数据适用性综合评估北大核心CSCD

青藏高原积雪对区域气候及水循环有重要影响,现有积雪数据集在该区域存在很大不确定性,适用性评估工作不可或缺。基于气象站观测数据(OBS),采用秩评分方法对一套被动微波遥感(CHE)和两套再分析(ERA5-Land和MERRA2)积雪深度数据进行了多变量、多评价指标的综合定量评估。结果表明:从年平均积雪深度、年最大积雪深度、年积雪日数三个变量分别评价各数据,MERRA2对年最大积雪深度、年积雪日数模拟最好,CHE对年平均积雪深度描述最好;各数据在不同评价指标上的得分排名存在较大差异,CHE在描述线性变化趋势上具有优势,ERA5-Land在描述年际变化上具有优势,MERRA2在描述季节循环、多年平均值、极大值、标准差上具有优势;综合考虑,MERRA2在青藏高原适用性综合评分最高、ERA5-Land次之、CHE最低。三种数据均存在明显不足之处,MERRA2对积雪线性变化趋势的定性描述与OBS相反,对积雪年代际变化的模拟有待优化;ERA5-Land对各变量的多年平均值存在严重高估;CHE刻画积雪空间分布特征能力较差。由于青藏高原西部站点稀少,相关评估结论仅适用于高原中东部。基于遥感及再分析数据得到高原西部积雪变化趋势存在较大不确定性。

Comprehensive applicability evaluation of multi-source snow depth datasets over the Tibetan Plateau

Snow cover over the Tibetan Plateau has an important impact on the regional climate and water cycle. At present， the existing snow cover datasets have great uncertainty across this region， so the applicability assessment is indispensable in order to make best use of the advantages and bypass the disadvantages. In this study， a comprehensive quantitative evaluation of multiple variables and multiple evaluation indicators was carried out for three snow depth datasets over the Tibetan Plateau against the meteorological station observations （OBS）. The three snow depth datasets include one passive microwave remote sensing dataset （CHE） and two reanalysis datasets （ERA5-Land and MERRA2）. The variables are the annual mean snow depth， the annual maximum snow depth， and the annual snow cover days. In addition， the evaluation indicators are seasonal cycle， climatology， maximum value， standard deviation， interannual variation， and trend. A rank score （RS） value of 0~1 is computed for each evaluation indicator of each variable， the larger value of RS indicate relatively better performance of a snow depth dataset. Assessment results imply that， comprehensively considered， MERRA2 exhibits best agreement with OBS， followed by ERA5-Land， and finally CHE. Evaluate based on the RS of each variable， MERRA2 shows better performance on annual maximum snow depth and annual snow cover days， CHE shows better performance on annual mean snow depth. Evaluate based on the RS of each evaluation indicator， CHE shows advantages in describing trend， ERA5-Land exhibits better agreement with OBS on interannual variation， and MERRA2 show better performance on the rest of the indicators including seasonal cycle， climatology， maximum value and standard deviation. The RS statistics in terms of regional average and spatial distribution show that CHE performs better in the former， and ERA5-Land performs better in the latter. On the other hand， there are obvious deficiencies in all three snow depth datasets. MERRA2 has insufficient ability to characterize the interdecadal variation in snow cover， and its qualitative results for trend in snow cover is inconsistent with OBS， the reason for the first deficiency needs to be further studied and the second deficiency may be mainly related to its simulation capability to precipitation trend. ERA5-Land significantly overestimates the snow cover over the Tibetan Plateau， this may be mostly related to its data assimilation scheme. CHE has poor ability to characterize the spatial distribution of snow cover， coarse spatial resolution of passive microwave remote sensing may be the main reason. The conclusions are only applicable to the central and eastern part of the Tibetan Plateau due to the scarcity of meteorological station in west part of the Tibetan Plateau. Based on the remote sensing and reanalysis data， there is great uncertainty in the trend of snow cover in the western part of the Tibetan Plateau. These systematic classification evaluation of the three representative snow depth datasets provides information on data selection and data refinement.