中国区域温度和降水不同空间插值方法精度对比

摘要

The precision and time series of different spatial interpolation data for temperature and precipitation over the mainland of China were studied,using 1. 0°×1. 0°and 0. 5°×0. 5°gridded,monthly temperature and precipi-tation data based on the optimal interpolation (OI),ANUSPLIN V2. 0 (Australian National University′s thin plate Smoothing SPLINe,AV 2. 0)interpolation,and ordinary kriging (OK)approaches from the China National Mete-orological Information Center and the GHCN (Global Historical Climatology Network)5. 0°×5. 0°gridded, monthly precipitation data from the U.S.NCDC (National Climatic Data Center)from 1961 to 2004.The results show that under an average climatic state from 1961 to 2004,there are consistent spatial distributions and annual cycle change in temperature and precipitation in China using different interpolation approaches.The absolute errors of precipitation fields in the whole mainland,the eastern and western China determined using the OI and AV 2. 0 approaches are 2. 15 mm,1. 28 mm and 0. 00 mm,respectively.The absolute errors of temperature fields for these regions using the OK and AV 2. 0 approaches are 0. 20 ℃,0. 05℃ and 0. 45 ℃,respectively.For time series of precipitation field in China,there is a relatively consistent seasonal variation obtained from AV 2. 0,OI and GHCN approaches,with a larger difference occurring in summer and a smaller one in winter.The correlation coefficient of precipitation fields between AV 2. 0 and OI approaches ranges from 0. 22 to 0. 98 from 1961 to 2004,with a high correlation occurring in winter and spring,and a low one in summer and autumn.The difference of precipitation a-mount is larger in some individual autumn and winter,with a maximum value of 3. 08 mm. The difference of mean precipitation amount from 1961 to 2004 is 0. 64 mm.The difference of mean annual temperature obtained from AV 2. 0 and OK approaches is less than 0. 54 ℃,and most of time,the change trends of temperature from the two ap-proaches are consistent.%利用国家气象信息中心基于最优插值法（Optimal Interpolation，OI）、ANUSPLIN插值法（AV 2.0）、普通克里格法（Ordinary Kriging，OK）的1.0°×1.0°与0.5°×0.5°格点化的1961—2004年中国区域月温度和月降水资料及1961—2004年美国NCDC的GHCN 5.0°×5.0°月降水资料，对中国大陆地区温度和降水不同插值方法空间插值数据的精度及时间序列进行了对比研究。结果表明：在1961—2004年平均气候态下，中国区域不同插值法插值后的降水和温度空间分布型较一致，年循环变化也较一致。在中国区域、东部区域和西部区域，OI与AV 2.0方法插值的降水场绝对误差分别为2.15 mm、1.28 mm和0.00 mm，OK与AV 2.0方法插值的温度场绝对误差分别为0.20℃、0.05℃和0.45℃。对于中国区域降水场时间序列， AV 2.0和OI方法插值的降水与GHCN不同季节的降水变化趋势较一致，且不同插值方法插值的夏季降水量差异较大，冬季降水量差异较小。1961—2004年AV 2.0与OI方法插值的降水场相关系数在0.22—0.98之间变化，冬季和春季降水场相关性较高，夏季和秋季降水场相关性较低；个别年份秋季和冬季插值后降水量的偏差稍大，最大偏差达3.08 mm，1961—2004年平均降水量偏差为0.64 mm。AV 2.0与OK方法插值的年平均温度差值小于0.54℃，且多年时间序列变化趋势较一致。