江西多年平均降水量空间插值模型的选取与比较

Spatial Interpolation Model Selection of Multi-Year Average Precipitation in Jiangxi Province

为了探究江西省多年平均降水量的空间分布格局,本文选取81个时间序列完整的气象站30a(1976—2005年)降水数据,运用泛克里格中不同半变异模型对降水数据进行拟合。分别采用65个建模站点交叉验证和16个验证站点的检验,证实了该研究区域多年平均降水量存在较强的空间相关性。通过不同半变异函数模型的对比发现,球面和指数模型在建模站点交叉验证结果中的标准均方根预测误差分别为1.024和1.023,较为接近1,表明其误差较小;其在验证站点检验结果中的标准均方根预测误差分别为1.105和1.104,表明这两种模型的拟合效果较优,能较为真实地反映江西省多年平均降水量的空间分布情况,其中以指数模型拟合效果最优。

To explore the distribution pattern of multi-year average annual precipitation in Jiangxi Province,81complete sequence time series of 81weather stations were used,which included precipitation data of 30 years（1976—2005）in Jiangxi Province.The universal kriging is used in different semivariogram model to fit the precipitation data after exploring the distribution features of the data;according to test of 65cross validation and inspection of 16stations,it is indicated that the multi-year average annual precipitation has strong spatial correlation.Through the comparison of different semivariogram function models,it is found that exponential model and spherical model in the results of cross validation stations were 1.024,1.023for RootMean-Square Standardized,which is close to 1and error was small.Exponential model and spherical model in the results of inspection stations were 1.105,1.104for Root-Mean-Square Standardized,showing that the fitting effect of the two models is better than the others,it can more perfectly reflect the spatial distribution of multi-year average precipitation in Jiangxi Province,in which the result of exponential model can obtain the better integer effect within the existing weather stations network.