中文摘要：

高精度土壤理化属性的空间分布图件是环境保护、土壤管理及精准农业不可或缺的要件,土壤特性空间异质性的普遍存在是土壤理化特性精确预测的一大难题。以土壤阳离子交换量（CEC）为对象,通过比较不同的空间插值方法,来探讨土壤属性的高精度预测方法。通过采样实测表层土壤阳离子交换量,利用普通线性回归、地理权重回归及二者残余的克里格进行空间插值,来比较不同方法的插值精度和制图效果。结果表明,地理权重回归残余克里格法的精度最高,其预测值与实测值的回归决定系数达90%,普通线性回归的精度最低。在回归残余分析中,普通线性回归残余对CEC预测精度提高的幅度达34%,而地理权重回归残余对预测精度提高程度为9%,相对较小。说明地理权重回归及回归克里格法也是较好的空间预测方法。在制图效果上,2种残余法的预测图在空间上过渡自然,平缓,能较好地反映出CEC在空间上的变化细节,与实际情况吻合较好。对于那些环境影响因素明确,且与之存在线性相关的土壤属性,地理权重回归残余克里格法及回归克里格法可作为其高精度空间预测及制图的有效工具。

英文摘要：

Soil property maps of high quality and accuracy are important for environmental protection, soil management and precision agriculture. It is challenging to predict soil chemical ＆ physical properties due to the spatial heterogeneity features of soil. Cation exchange capacity （CEC） as a target soil property was used in this study for acquiring its high accuracy spatial distribution map compared with various interpolating approa- ches. The predicting accuracy and mapping effects of soil CEC interpolated by differential approaches based on the observed CEC samples in surface soil were evaluated, these approaches were ordinary least squares （OLS）, geographically weighted regression （GWR）, RK （regression Kriging, that is OLS and its regression residuals with ordinary Kriging）, and GK （GWR and its regression residuals with ordinary Kriging）. There- suits showed that the prediction accuracy of GK ranksed the first place among the four used methods, and the coefficient of determination between its predicted and observed values reached to 90 percent. While the worstone in prediction accuracy was OLS. In the analysis of residuals, OLS residuals had 34% contribution to the improvement of CEC prediction accuracy, larger than 9% by GWR residuals. It was discovered that GWR and RK were both good methods in spatial prediction for soil properties. In the aspectlof mapping effects, RK and GK could better reflect the marks of environmental factors on soil CEC, and maps were more continuous in transition and less affected by extreme sample values. The maps interpolated by GK and RK were more realis- tic and reflected the details of soil CEC patterns impacted by the environmental factors in nature. RK and GK are eligible tools for predicting and mapping soil properties with high accuracy combined with related environ- mental factors which have explicit linear relation with soil properties.