新疆地区年降水量的空间插值方法研究

摘要

为提高新疆地区年降水量的空间插值效果,通过增设模拟气象站点、补充实测站点数据空间变异性信息,采用确定性函数和地统计学的综合插值方法进行新疆地区年降水量的空间插值与比较验证分析.结果表明:(1)利用地统计方法分析多年平均降水量的最佳抽样间距,确定气象站点影响区,建立降水量与经纬度、高程、坡度、坡向、NDVI间的多元回归方程,增设模拟气象站点,可以有效增加降水量空间变异信息,显著提高降水量空间插值效果.(2)通过对比发现,反距离加权法(IDW)、复二次径向基函数法(MRBF)和普通克里格法(OK)3种空间插值方法中,MRBF的插值效果最好,是该地区降水空间分布的最适插值方法.(3)在新疆地区采用逐步综合的方法,即以多元回归方法增设模拟气象站点和MRBF进行年降水量空间插值,全面考虑影响因子与空间变异性,去除了多元回归带来的“边界效应”,弥补了实测降水量空间信息,得到了具有较高精度的年降水量数据集.%For improving the effect of annual precipitation spatial interpolation in Xinjiang area, the spatial interpolation of annual precipitation had been performed and the precision was comparatively validated by using of deterministic function and geo-statistical interpolation methods, through adding some simulative meteorological stations as a supplement of spatial variability information of fact station data. The results showed that the precipitation spatial variability information and interpolation precision had effectively enhanced by establishing simulative weather stations through finding weather stations influence zone by investigating appropriate sampling interval distance of multi-year average precipitation with geostatistical analysis, and founding multiple regression equation between precipitation and longitude-latitude, elevation, slope, aspect and NDVI( normalized difference vegetation index). MRBF was the best and most appropriate method for spatial interpolation of precipitation among three interpolation methods of Inverse Distance Weighted (IDW), Multiquadratic Radial Basis Functions (MRBF) and Ordinary Kriging (OK) in Xinjiang area. By using stepwise and integrated methods, namely establishing simulative meteorological stations through multiple regression and conducted spatial interpolation of precipitation by MRBF method, considering significant factors and spatial variability, removing boundary effect of multiple regression, enhancing precipitation spatial geostatistical information, the annual precipitation spatial distribution dataset of this area with higher precision had been obtained.