回归克里格在土壤全氮空间预测上的应用

Application of Regression Kriging on the Spatial Prediction of Total Soil Nitrogen

精确地了解表层土壤养分的空间分布,是生态环境保护、合理施肥及精准农业的基础。普通克里格法以区域化变量理论为基础,在土壤空间制图中得到了广泛的应用,但该方法过分依赖样点数据数量和质量,忽视了影响土壤养分的环境要素,因而该方法在空间预测上存在缺陷。研究以使用环境辅助变量的回归克里格法在土壤全氮空间插值效果及空间制图应用,并与普通克里格做比较,对二者的制图效果及插值精度进行评价。结果表明,回归克里法能充分地利用容易获取环境辅助变量,预测精度明显地高于普通克里格。在制图效果上,回归克里格法得到的预测图更能体现环境要素的作用痕迹,预测图在空间上过渡自然、平缓,很少有在过渡区域表现突变的现象,制图效果更符合全氮空间分布的实际。因此,回归克里格法适合环境因素影响明确,且与特定环境要素存在线性相关或特定趋势的土壤属性的预测和制图。

Precisely knowing the spatial distribution and variation of soil nutrient was the premise forecological environmental protection, rational fertilization and precision agriculture. Ordinary kriging, a method based on regionalized variable theory, was widely used in the research of soil science, such as soil mapping, prediction, etc., but the method heavily depended on quality and number of data and ignored the effect of environmental factors which were very closed to soil nutrient, hence, the method had some defects on prediction of soil properties. Compared with ordinary kriging, regression kriging, which could use many auxiliary environmental variables, was used in the study for interpolating total soil nitrogen and its mapping. The mapping effects and accuracy of interpolation by the two methods were also evaluated. The results showed regression kriging could fully employ easily acquired auxiliary variables, and the prediction accuracy of regression kriging was greatly improved by compared with ordinary kriging. The map interpolated by regression kriging was more realistic and reflected the details of the total soil nitrogen patterns impacted by the environmental factors in nature and the map was more continuous in transition and less affected by extreme sample values. The map interpolated using regression kriging was more accord with soil total nitrogen distribution in nature. Regression kriging method was eligible for prediction and mapping those soil properties which had explicit relation with environmental factors, that was to say, these factors had better have a linear correlation or specific trend with soil properties.