基于多源辅助变量和极限学习机的蔬菜地土壤有机质预测研究

发布时间：2018-03-13 02:05

本文选题：土壤有机质　切入点：极限学习机　出处：《土壤通报》2017年01期 　论文类型：期刊论文

【摘要】：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。
[Abstract]:Using multiple auxiliary variables to predict the spatial distribution of soil organic matter can effectively improve the prediction accuracy. 422 soil samples were collected in Xi'an vegetable production area. Using extreme learning machine learning machine, stepwise linear regression model, support vector machine support vector machine (SVM) and random forest random for stave (RFRF) model, combining slope, slope direction, planting years, planting type, irrigation method, nitrogen fertilizer application rate, phosphate fertilizer application rate, slope, slope direction, planting years, planting type, irrigation mode, nitrogen fertilizer application rate, phosphate fertilizer application rate, The soil organic matter content of vegetable land in the study area was predicted by applying potassium fertilizer, soil type, alkali-hydrolyzed nitrogen, available phosphorus, available potassium, salt, nitrate pH value, and so on. The results show that the root-mean-square error is 0.631 g 路kg ~ (-1) and the ratio of root mean square error to the average value of prediction set is 0.037, which is lower than that of the other three models. The results show that the root-mean-square error of soil organic matter is 0.631 g 路kg ~ (-1) and the ratio of mean square root error to mean value of prediction set is 0.037. The correlation coefficient is 0.716, which is significantly higher than the spatial prediction results of SLR- SVM and RFNELM, which is closer to the true situation of soil organic matter content. According to the results of ELM analysis and the essence of the algorithm, this paper explains the geographical significance of its application in the field of soil attributes, and introduces a new method for spatial prediction of other soil attributes.
【作者单位】：西北大学城市与环境学院;西安市农业技术推广中心;西安市农产品质量安全检验监测中心;西北大学信息科学与技术学院;
【基金】：教育部人文社会科学研究规划项目(10YJA910010) 陕西省农业科技攻关项目(2011K02-11) 西安市科技计划项目(NC1402,NC150201) 西北大学“211工程”研究生自主创新项目(YZZ15013)资助
【分类号】：S153.6

【相似文献】