DEM的不确定性对温度插值精度的影响

Impacts of DEM Uncertainty on Temperature Interpolation Accuracy

以2001-2010年全国711个气象站点温度观测数据为数据源,在薄板样条插值方法和ANUSPLIN软件支持下,对比分析了使用GTOP30、SRTM3和ASTER GDEM三种DEM数据作为协变量开展空间插值得到的温度成果数据精度。研究发现:1)GDEM的平均相对误差(MRE)、平均绝对误差(MAE)和均方根误差(RMSE)明显高于其他两套数据,数据质量最差;SRTM3和GTOP30的数据精度较高,各误差评估指标大致相同。2)在全国尺度上,在MAE方面,基于ASTER GDEM所得温度插值结果比其他两组DEM得到的温度插值结果高0.4℃左右;在RMSE方面,基于ASTER GDEM获得的温度比其他两组DEM得到的温度高0.5℃左右。基于ASTER GDEM所得温度插值数据的误差明显高于其他两套数据。3)在温度插值误差的空间分布格局上,在我国中东部地区,三种DEM温度插值误差分布规律基本相同;在我国西南部地区,基于GTOP30和SRTM3的温度插值结果的数据精度明显好于基于ASTER GDEM的温度插值结果。4)温度插值误差与DEM高程误差呈现明显的耦合特性,这表明DEM数据的精度是影响温度插值结果的重要因素。

Elevation is an important factor to impact the spatial distribution of temperature.In this paper,the mean temperature,maximum temperature,minimum temperature was extracted by using the data of 711 meteorological stations from 2001 to 2010in China.This paper compares the accuracy of interpolation results by using different DEM datasets as the covariates at the national scale and regional scale.Three global DEM datasets are used in this paper：GTOP30、SRTM3and ASTER GDEM.All interpolation work has been conducted with the aid of the software of ANUSPLIN.The analysis shows that：1）The MAE,MRE and RMSE of GDEM is higher than GTOP30 and SRTM3.It indicates that the accuracy of GDEM elevation is lower than GTOP30 and SRTM3.2）In the national scale,GTOP30 and STRM3showed relatively consistent accuracy for the three temperature variables according to the root mean square error（RMSE）and mean absolute error（MAE）.The result of ASTER GDEM is relatively bad.The MAE of ASTER GDEM is higher than GTOP30 and SRTM3by 0.4℃,the RMSE of ASTER GDEM is higher than GTOP30 and SRTM3by 0.5℃.3）In the regional scale,three DEM datasets have not obvious different error at the eastern region of China.But the interpolation accuracy of GTOP30 and SRTM3is higher than ASTER GDEM at region of the Tibetan plateau and the Sichuan basin.4）Temperature interpolation error and absolute error of DEM elevation at the national scale and the regional scale showed consistency.The reason is considered that the temperature interpolation error may be caused by the elevation error.Results suggest that SRTM3 is the best DEM datasets as the accuracy to interpolate temperature.