海南岛日最低、最高气温空间插值研究

Spatial Interpolation of Daily Minimum Temperature and Daily Maximum Temperature in Hainan Island

为探索适用于海南岛日最低、最高气温空间插值的方法,研究以经度、纬度、海拔、坡度、坡向、海陆距离、NDVI等为环境变量,采用多元线性回归(multiple linear regression,MLR)、地理加权回归(geographically weighted regression,GWRK)、多元线性回归克里格(multiple linear regression-Kiging,MLRK)和地理加权回归克里格(geographically weighted regression-Kriging,GWRK)等4种方法对海南岛2016年1月1日—6月30日的日最低、最高气温进行了插值。结果表明:4种方法对日最低气温插值的总的平均绝对误差:MLR>GWR>GWRK>MLRK,但GWR、GWRK、MLRK对日最低气温插值的总的平均绝对误差十分接近,对日最高气温有相同的规律。MLRK对日最低、最高气温的总体平均绝对误差分别为0.50℃和0.73℃。GWRK、MLRK对逐日最低气温插值的平均绝对误差也十分接近,对日最高气温也有相同的规律。无论是对日最低气温还是对日最高气温,MLRK、GWRK插值空间分布的主要差异均在站点稀疏的山区。因此,在海南岛,宜采用多元回归克里格(MLRK)对日最低、最高气温进行空间插值。

The paper aims to find the suitable spatial interpolation method for daily minimum temperature and daily maximum temperature in Hainan Island.Multiple linear regression（MLR）,geographically weighted regression（GWRK）,multiple linear regression-Kiging（MLRK）and geographically weighted regressionKriging（GWRK）were compared to interpolate daily minimum temperature and daily maximum temperature in Hainan Island during January 1-June 30,2016 by using longitude,latitude,elevation,slope,aspect,closest distance to the seaside,NDVI as covariates.The results showed that：the mean absolute errors（MAE）of the four methods for all daily minimum temperature samples were MLRGWRGWRKMLRK;however,the MAEs of GWR,GWRK,and MLRK for all daily minimum temperature samples were very close,the MAEs of GWRK and MLRK for daily minimum temperature were also very close;the daily maximum temperature was the same;the MAE of MLRK for daily minimum temperature and daily maximum temperature was 0.50℃ and 0.73℃,respectively;the main differences of the spatial distributions interpolated by GWRK and MLRK for the daily minimum temperature were in the site sparse mountainous areas;the daily maximum temperature was the same.Therefore,in Hainan Island,the daily minimum temperature and daily maximum temperature should be interpolated by using MLRK.