基于ArcGIS不同空间插值方法的降雨量预测效果对比

A Comparison of Precipitation Prediction Effect Based on Different Spatial Interpolation Methods in ArcGIS

利用美国Idaho州的实测年均降雨量数据,基于ArcGIS平台分别采用趋势面拟合、IDW、Splines和普通Kriging四种插值方法进行了降雨量预测建模,使用留出验证法对所建的四种模型实施了预测效果评估。基于20个预留站点的验证结果显示:趋势面拟合、IDW、Splines和普通Kriging四种插值方法的拟合R^2分别为0.74,0.71,0.83,0.81;RMSE分别为5.17,5.30,4.11,4.37in。即在该例中Splines插值模型的预测效果最优,而后基于Splines插值模型生成了美国Idaho州年均降雨量分布图。

Trend surface fitting, inverse distance weight, thin-plate splines and ordinary kriging interpolation methods were carried out in ArcGIS to model the distribution of precipitation in Idaho, US based on annual averaged precipitation data. Hold-out validation was used to evaluate prediction effect of these four methods. The validation result based on twenty hold-out stations showed that: the fitting R^2 and RMSE of these four models were 0.74, 0.71, 0.83, 0.81 and 5.17, 5.30, 4.11, 4.37 in respectively. The model of Splines interpolation was the best method in this study. Then, a distribution map of Idaho’s annual precipitation was generated using Splines interpolation method in ArcGIS.