论文部分内容阅读
应用多源辅助变量预测土壤有机质的空间分布,能有效提高预测精度。以西安市蔬菜产地为研究区域,共采集422个土壤样品,运用极限学习机(extreme learning machine,ELM)、逐步线性回归(stepwise linear regression,SLR)、支持向量机(support vector machine,SVM)和随机森林(random forest,RF)模型,结合坡度、坡向、种植年限、种植类型、灌溉方式、氮肥施用量、磷肥施用量、钾肥施用量、土壤类型、碱解氮、有效磷、速效钾、盐分、硝酸盐、pH值等15个多源辅助变量,对研究区蔬菜地土壤有机质含量进行空间预测,并通过100个实测点验证预测结果。结果表明:ELM对土壤有机质预测结果的均方根误差为0.631 g kg-1,均方根误差和预测集平均值的比值为0.037,二者均低于其他3种模型,ELM的相关系数为0.716,显著高于SLR、SVM和RF,ELM的空间预测结果更接近土壤有机质含量的真实情况。同时,根据ELM分析结果及算法本质阐释其在土壤属性领域应用的地理学意义,也为其他土壤属性空间预测引入了一种新方法。
Predicting the spatial distribution of soil organic matter by using multi-source auxiliary variables can effectively improve the prediction accuracy. A total of 422 soil samples were collected from vegetable producing areas in Xi’an City. Extreme learning machine (ELM), stepwise linear regression (SLR), support vector machine (SVM) and Random forest (RF) model was used to study the effects of different types of irrigation on the slope, slope, planting years, planting types, irrigation methods, N application rates, P application rates, K application rates, soil types, 15 multi-source auxiliary variables such as salinity, nitrate and pH values were used to predict the soil organic matter content of vegetable fields in the study area. The results were verified by 100 measured points. The results showed that the root mean square error (RMSE) of soil organic matter (ELM) predicted by ELM was 0.631 g kg-1, and the ratio of root mean square error to the average of prediction set was 0.037, both of which were lower than the other three models. 0.716, which was significantly higher than that of SLR, SVM and RF. The spatial prediction results of ELM were closer to the real situation of soil organic matter content. At the same time, according to the result of ELM analysis and the nature of algorithm, it explains the geographical significance of its application in the field of soil properties, and also introduces a new method for the spatial prediction of other soil properties.