基于优化参数的陕西省气温、降水栅格化方法分析

基于统计、地统计基本原理,运用传统插值法、地统计插值法、多元回归法和模拟气象站点法4 大类11 种方法,对陕西省2003-2012 年平均气温、降水量数据进行栅格化。结果表明:① 多元回归法和模拟气象站点法均能大幅提高气温数据的估测准确度,且多元回归法对气温的表示更加精细,其中以“回归+残差IDW(Inverse Distance Weighting)”法精度最高,MAE、RMSE、R²分别为0.498、0.775 和0.954 8;② OK(Ordinary Kriging)法对降水数据的估测准确度最高,MAE、RMSE、R²为46.934、69.251 和0.947 8;③ 气温、降水栅格化方法具有明显的区域性,不存在绝对普适的最优方法,应根据数据类型、地域特征对原始数据进行探索性分析,从而获得区域最适合的栅格化方法;④ 陕西2003-2012 年多年平均气温为10.925 ℃,标准差2.221 ℃,气温随纬度、海拔的增加而降低,具有鲜明的纬度和垂直地带性;平均降水量为664.446 mm,标准差213.226 mm,降水呈现由南向北逐渐递减的趋势,纬度地带性较强。

Comparison of Temperature and Precipitation Rasterization Methods Based on Optimized Parameters in Shaanxi Province

According to the basic principles of statistics and geo-statistics, the paper uses 11 different methods of four categories including traditional interpolation method, geo-statistical interpolation method, multiple regression and suppositional meteorological station method, to rasterize the annual average temperature and precipitation in Shaanxi province from 2003 to 2012. The results show that: 1) Multiple regression and suppositional meteorological station methods can greatly improve the temperature estimating accuracy. Furthermore, multiple regression method is more refined in calculating temperature details, especially the“Regression+Residual IDW (Inverse Distance Weighting)”method, whose MAE, RMSE and R² are 0.498, 0.775 and 0.9548 respectively. 2) OK (Ordinary Kriging) method is the most accurate method in estimating precipitation, whose MAE, RMSE and R² are 46.934, 69.251 and 0.9478. 3) The rasterisation methods of temperature and precipitation are obviously regionally adapted. There is no a universal optimal rasterization method. So it's necessary to explore and analyze the original data before acquiring the most suitable rasterization method. 4) The perennial average temperature of Shaanxi from 2003 to 2012 is 10.925 ℃, with the standard deviation of 2.221 ℃. The temperature decreases with the increasing of latitude and elevation, having distinct latitudinal and vertical zonality. The average precipitation is 664.446 mm, with the standard deviation of 213.226 mm. The latitudinal zonality of precipitation is strong, showing a decreasing trend from south to north.