天津表土DDT浓度的空间插值方法研究

Spatial Interpolation Methods of DDT in Tianjin Topsoil

分别运用反距离加权插值法(IDW)、反距离平方加权插值法(ISDW)、多项式回归法(PR)、双线性插值法(BI)以及普通克里格法(OK)共5种方法对天津表土188个采样点的DDT浓度数据进行了空间插值,并对插值精度进行了评估和比较。结果表明,上述5种方法均无系统误差,且方均根误差(RMSE)差别不大,从误差分布来看,普通克里格法表现稍好。188点中,普通克里格法插值误差在0.5和1.0个对数单位以内的百分比分别为62.2%和91.5%。在表土DDT浓度突变的局部,所有插值方法的误差都超过一个对数单位。相对于其他4种插值方法而言,普通克里格法生成的空间插值图,更能准确表达天津表土DDT浓度的空间分布特征,其细部的浓度变化也更清晰。

DDT concentration data from 188 topsoil samples in Tianjin, China, was interpolated using five methods： inverse distance weighting （IDW）, inverse square distance weighing （ISDW）, polynomial regression （PR）, bilinear interpolation （BI）, and ordinary Kriging （OK）. Interpolation results were cross-validated and evaluated. For each method of estimation we computed two indices, mean error （ME） and root mean square error （RMSE）, from the observed and predicted values of DDT concentration data at 188 sample sites. All methods presented unbiased estimation. ME was close to zero and similar for all five methods： OK（-0.002）, BI（-0.003）, PR（-0.005）, IDW（0.006）, IDSW（0.010）. According to analysis of error＇s distribution, OK gave the best interpolation result, in which 62.2% and 91.5% RMSE were less than 0.5 and 1.0 log unit respectively. All methods showed deficiency at locations where DDT concentration varies sharply and dramatically, at these locations RMSE were higher than 1.0 log unit. IDW and ISDW methods produced raster map with poor spatial structure. PR and BI showed clear spatial structure, but with few details of DDT concentration in Tianjin topsoil. Contrastively, OK produced raster map with clear spatial structure and local details of DDT concentration in Tianjin topsoil.