Since the spatial resolutions between MERIS water vapor products and ASAR data are greatly different, the elevation-dependent ONN water vapor correction model is not sufficient in precision for spatial interpolation, especially in the area with large local elevation changes. In order to solve the problem, spaital interpolation methods, i.e. ordinary Kriging and von Karman Kriging, are proposed to combine with the ONN model for spatial interpolation. In order to correct the local over-smoothing in the Kriging model, Yamamoto's modified Kriging method was applied for spatial interpolation of regional atmospheric residuals. The results show that total atmospheric distribution in the studied area is the sum of the local atmospheric residual estimation and global ONN interpolation result. According to the cross-validation of the three different interpolation methods, von Karman Kriging for residual interpolation+ONN terrain model has better results than the other two methods in Ermse, Emae, Emre and ESTD. Additionally, von Karman Kriging for residual interpolation is more fit to the spatial distribution characteristics of the atmosphere, which is beneficial to correct the local over-smoothing in the von Karman Kriging for residual interpolation+ONN terrain model.%对于分辨率差别较大的MERIS水汽产品与ASAR数据,直接利用ONN地形模型进行大气空间插值对局部地形变化较大的区域插值精度不高.针对这一问题,在基于ONN地形水汽空间插值的基础上,提出应用普通Kriging和von Karman Kriging 2种插值模型对局部大气进行空间插值.为修正Kriging模型引起的局部过度平滑问题,运用这2种Kriging模型的基础上,提出运用Yamamoto修正的Kriging法对区域大气残差进行空间插值.将全局ONN插值结果与局部大气残差估计值相加,得到区域的大气分布状况.对3种不同的插值方法进行交叉验证比较发现:利用ONN地形模型+基于残差的von Karman Kriging方法空间组合插值方法精度,无论是均方根误差ERMSE、平均绝对误差EMAE、平均相对误差EMRE还是平均相位标准偏差都远低于ONN地形模型+基于残差OK的空间组合插值和简单的ONN地形模型的插值误差,而且ONN地形模型和基于残差的von Karman Kriging 的空间组合插值方法能进一步克服ONN+基于残差OK模型的区域平滑问题,更符合大气的空间分布特征.
展开▼
