基于GIS的“一带一路”地区气温插值方法比较研究

“一带一路”倡议是新时期中国为加强对外开放提出的全球化合作倡议,资源环境的优化配置对全球化发展意义重大。气温作为重要的基础数据和输入要素,对其进行空间化处理是实现大尺度区域资源环境优化配置的前提。本文基于地理信息技术（GIS）,运用距离平方反比法（IDS）、协同克里格法（CK）、回归距离平方反比法（RIDS）和回归协同克里格法（RCK）,对“一带一路”地区1980—2017年的2679个气象站点的月平均气温和年平均气温数据进行插值,获得了“一带一路”地区10 km分辨率的气温空间分布数据。交叉验证结果表明：① IDS、CK、RIDS和RCK插值法在整体上均较好地展示了“一带一路”地区气温的地理空间分布规律,4种插值方法的月均气温的均方根误差分别在1.93~2.43、1.78~2.14、1.31~2.23和1.23~1.92 ℃之间;年均气温的均方根误差分别为1.94、1.83、1.37和1.27 ℃;② 在“一带一路”地区,加入协变量分析的CK插值精度整体优于IDS,并且削弱了IDS的极值现象;③ RIDS和RCK对年均气温的插值精度分别较IDS和CK提高了29.4%和30.6%,表明加入地理要素并进行残差修正的插值精度得到了进一步提高。总体来看,RCK插值法对气温数据的插值精度最高,可以考虑将此方法作为“一带一路”地区温度等气象要素的插值方法。

Comparative Study of Different Temperature Interpolation Methods in the Belt and Road Regions based on GIS

The Belt and Road initiative was a globalization cooperation initiative put forward by China to strengthen the opening-up in the new era. With the development of globalization, it is of great significance to optimize the allocation of resources and environment. As an important reference dataset and input factor, the result of temperature interpolation is the basis for optimal allocation of regional resources and environment in large scale study area. Here, taking the Belt and Road (BR) regions as the study area, the monthly and annual mean temperature data in 2679 meteorological stations from 1980 to 2017 were interpolated based on Geographic Information Technology (GIS), using Inverse Distance Squared (IDS), CoKriging (CK), Regression-IDS (RIDS) and Regression-CK (RCK) interpolation methods. The 10 km map of spatial interpolation were generated using aforementioned four methods. The results showed: (1) In the BR regions, the geographical distribution of temperature were better displayed by IDS, CK, RIDS and RCK. The Mean Square Root Error (RMSE) of monthly mean temperature were 1.93~2.43 ℃, 1.78~2.14 ℃, 1.31~2.23 ℃ and 1.23~1.92 ℃, IDS, CK, RIDS and RCK, respectively. And the RMSE of annual mean temperature were 1.94 ℃, 1.83 ℃, 1.37 ℃ and 1.27 ℃, IDS, CK, RIDS and RCK, respectively. (2) The accuracy of CK interpolation with covariates was better than that of IDS, and the peak values produced by IDS were corrected. (3) After considering the impact of terrain, the accuracy of interpolation in temperature based on Residual correction were improved by 29.4% and 30.6%, RIDS compared to IDS and RCK compared to CK, respectively. In summary, The Regression-CK performed better than other three methods in this study area and it can be considered as temperature and climate data interpolation methods in the BR regions.