根据中国 4 0 0多个气象站 1995年的积温数据分析 ,≥ 0℃积温和≥ 10℃积温与经纬度、海拔高度的相关性分别为r2 =0 96 5 6和r2 =0 94 0 2。以≥ 0℃积温数据为例 ,利用聚类分析法 ,将全国分为 7个积温计算区 ,每个区分别构建模型 ,通过“回归方程计算 +空间残差”的方法对全国积温数据进行栅格化。验证结果为 :计算值与实测值之间的线性相关系数r2 =0 9889,平均相对误差 3 5 6 % ,相对误差在 5 %以内的气象站个数占验证气象站总数的 86 %。

Accumulated temperature is an important heat resource and one of the key factors for crop growth. Accumulated temperature is of significance in guiding agricultural production. Usually, accumulated temperature data from meteorological stations can only reflect heat regimes of areas close to meteorological stations. In areas far from meteorological stations, it is by interpolation that accumulated temperature can be calculated. Rasterizing is one of the means for interpolating accumulated temperature data. Accumulated temperature is influenced by longitude, latitude and altitude to a great extent. There is a close linear relationship between accumulated temperature and longitude, latitude and altitude. By analyzing data from more than four hundred national meteorological stations in China, it was found that the ≥0℃ accumulated temperature and ≥10℃ accumulated temperature in 1995 had correlation coefficients of r 2= 0.9656 and r 2= 0.9402 with longitude, latitude and altitude respectively. In order to explore ways to process rasterized accumulated temperature data products, the ≥0℃ accumulated temperature was taken as an example. China was divided into seven different accumulated temperature regions by means of cluster analysis, seven models suitable for each region were established respectively, and the method of "model based computation result plus spatialized residues "was used to rasterize accumulated temperature data in China. Data from 436 of 481 meteorological stations were used to establish rasterizing models, and accumulated temperature data in whole China were calculated. 45 meteorological stations, which were distributed countrywide evenly and were not used to establish the models and to calculate accumulated temperature, were selected to test and verify precision of the rasterizing method. The result showed that there was a close relationship of r 2 = 0.9889 between actual observed accumulated temperatures and calculated temperature by rasterizing with mean relative deviation equal to 3.56%, and the meteorological stations with deviation less than 5% occupied 86% of the total meteorological stations used for testing and verifying. It was demonstrated that the rasterizing method used in the study is more accurate and it can be used for rasterizing of accumulated temperature data at large scale.