近50a中国降水格点数据集的建立及质量评估

Establishment and assessment of the grid precipitation datasets in China for recent 50 years

基于2012年6月更新的高质量2 400个台站降水资料,采用薄盘样条法,制定了采用3个自变量(经度、纬度、海拔高度)、降水量开平方预处理、3次样条的插值方案,并引入数字高程资料,以减弱中国独特地形条件下高程对降水空间插值精度的影响,并对1961—2010年中国区域地面降水站点资料进行了空间内插,得到了中国地面降水0.5°×0.5°格点数据集。经数据集的质量评估结果表明:分析值与站点观测值均方根误差平均为0.49 mm,相关系数平均达0.93(通过0.01的显著性检验),夏季插值误差高于冬季,东南地区误差普遍高于其他地区。冬、春、夏、秋季绝大多数台站绝对误差在±10 mm/月以内。冬、春、夏、秋季分别有60%、82%、54%、77%的台站相对误差在±10%之间。插值后的格点化降水资料能够比较细致、准确地描述中国大陆年平均降水场的东南多、西北少的主要空间特征,但也平滑掉了范围很小的降水极值中心。台站分布越密集的地方,插值效果越好,并且最近距离小于40 km的台站插值精度较高,大于40 km插值精度衰减较快。

Based on the high-quality precipitation data from 2,400 stations over China through the ANUSPLIN software developed by Australian National University with thin plate smooth spline method,interpolation scheme adopting three variables （longitude,latitude,altitude）,precipitation square pretreatment and three times spline was set up,then the digital altimetric data was introduced in order to weaken the effects of elevation on interpolation precision of precipitation square under the condition of Chinese special landscape,finally the datasets of daily and monthly 0.5°×0.5° grid-based precipitation over China from 1961 to 2010 were established by interpolating the precipitation data.Cross-validation tests show that this gauge-based analysis worked quite well with annual interpolation error typically less than 0.49 mm and the relation coefficient amounting to 0.93; monthly error depended on period and region,it was more in summer than in winter,and it was more in the east and west than in other areas.The mean bias error of the most gauges was between-10mm/month and 10mm/month.The relative bias error of 60％,82％,54％ and 77％ gauges were between-10％ and 10％.This kind of datasets is helpful to explore the spatial and temporal distributions of the precipitation,but it still ignores narrow-range rainfall extreme centers.It can be used widely in weather/climate monitoring,climate analysis,numerical model verifications,ecological assessment and hydrological studies.