高山区多时间尺度Anusplin气温插值精度对比分析

Contrastive Analysis of Temperature Interpolation at Different Time Scales in the Alpine Region by Anusplin

以中国地形起伏较大的横断山区为研究区,利用研究区内51个国家气象站点1960-2014年气温数据,以高程为协变量,采用Anusplin对不同时间分辨率(年代、年均、季节、月、日)的气温数据分布进行空间插值。通过气象站点交叉验证,结合绝对误差(平均绝对误差和均方根误差)和相对误差(平均相对误差)等量化指标,对比分析各个时间尺度气温数据的插值精度。结果表明:(1)夏季插值的绝对精度和相对精度最优,年代、年均和秋季气温插值的绝对精度水平相近,春季次之,冬季气温插值的绝对精度和相对精度最差。(2)在区域气候变化趋势方面,各个时间尺度气温插值数据的变化趋势和变化率与观测值结果一致。(3)年内各月均温和各月内日均温插值结果的相对精度和绝对精度在夏季月份精度最佳,在冬季月份精度最差,且各月内日均温插值结果在绝对精度和相对精度方面均低于各月均温插值结果的精度水平。分析显示,利用Anusplin的山区气温插值,气温海拔梯度性的优劣是造成不同时间尺度气温插值精度水平不一致的主要原因。

Anusplin is a common method that considering terrain effect during the temperature interpolation and it also has a high accuracy for temperature interpolation in the alpine area. But it is not clear whether there will be different accuracy based on the temperature data at different time scales by this method, especially for mountain temperature interpolation, which is very important for mountain science research. In order to determine the accu- racy of temperature interpolation in mountain area at different time scales, we took the most complex mountain area（Hengduan Mountains, a mountain range in Sichuan and Yunnan） as study area. Based on the temperature data recorded by 51 meteorological stations during 1960- 2014 and SRTM in the study area, Anusplin was used to interpolate regional temperature at different time scales （decadal, annual, seasonal, monthly, and daily）. With the cross-validation and comparison method, the interpolation accuracy of temperature at each time scale was compared by mean absolute error （MAE） , mean relative error （MRE） and root mean square error （RMSE）. The results showed that ： （ 1 ） The interpolation of summer temperature shows the minimum value in MAE, MRE and RMSE（ O. 41℃, 0.03 ℃ and 0.66 ℃ ）, while the interpolation of winter temperature shows maximum val- ue in MAE, MRE and RMSE （0. 79 ℃, 0. 81 ℃ and 1. 02 ℃）. The MAE value and RMSE value of decadal, annual and autumn temperature interpolation are similar. （2） The temperature interpolation by using Anusplin can well express the climate trends at decadal, annual and seasonal scale. The trend obtained from interpolation data are consistent with the tendency calculated from observed data, and their change rate shows no discrepancy [ error 〈0. 1 ℃ · （10a） -1]. （3） The MAE, MRE and RMSE value for daily and monthly temperature interpolations in summer are also minimum while in winter are maximum. In addition, the MAE, MRE and RMSE value of daily temperature interpolation are higher than the monthly temperature interpolation. Especially for the mean relative error in winter, the value of daily temperature interpolation error is about twice than that of monthly temperature interpolation error. It is worth noting that the inconsistency of temperature interpolation accuracy at decadal, an- nual and seasonal scale is mainly attributed to the difference in the relationship between temperature and elevation at different time scale. Moreover, the accuracy error of elevation data and the regional special climatic condition can also cause error for temperature interpolation by using Anusplin.