沉积物粒度组分空间预测的地理加权回归残余Kriging方法

Geographically weighted regression residual Kriging method applicable for the spatial prediction of sediment grain size compositions

针对沉积物粒度组分空间分布的自相关性和环境相关性特征,提出了基于地理加权回归残余Kriging的近海沉积物粒度组分空间预测方法,并以苏北海州湾表层沉积物的粒度数据为基础,分析和评估了其在沉积物粒度组分空间预测和底质类型制图应用中的可行性。结果表明,地理加权回归残余Kriging方法可获得比普通Kriging方法更高的粒度组分空间预测精度和底质类型制图精度,其底质类型的总体制图精度达到了89.6%,相应的Kappa系数为0.873,显示其底质的制图类型与样本的实测类型有着较好的一致性。新方法综合考虑了变量的空间自相关性和环境相关性,在定量化的沉积物粒度组分空间预测和底质类型制图中具有实用价值。

In view of characteristics of autocorrelation and environmental correlation of the spatial distribution of grain size components in offshore surface sediments,a geographically weighted regression residual Kriging(GWRRK)method is proposed with its application feasibility analyzed and evaluated in spatial prediction of grain size compositions and sediment types using the grain size composition data of surface sediment in Haizhou Bay,north Jiangsu Province.The results showed that GWRRK method can obtain a higher spatial prediction accuracy of sediment grain size components and mapping accuracy of sediment types than the ordinary Kriging method.The overall predictive mapping accuracy of sediment types by GWRRK method reached 89.6%.Its corresponding Kappa coefficient is 0.873,indicating that the mapping types of sediments are in good agreement with their actual types.As the new method can comprehensively consider the spatial autocorrelation and environmental correlation of variables,it has practical value in quantitative spatial prediction of sediment particle size components and mapping of sediment types.