基于组合模型的庐山森林土壤有效铁光谱间接反演研究

发布时间：2018-02-15 06:57

本文关键词： 土壤光谱有效铁预测 RBF神经网络偏最小二乘回归(PLSR) 组合模型　出处：《土壤学报》2017年03期 　论文类型：期刊论文

【摘要】：铁是植物生长的重要微量营养元素之一,土壤有效铁含量对林地环境起着重要的影响,利用土壤光谱预测技术获取土壤有效铁含量信息具有重要意义。而要通过土壤光谱直接预测土壤有效铁含量是难以实现的,因此提出利用土壤有机质含量与有效铁含量之间的相关性,探讨间接估算土壤有效铁含量的可行性。以庐山森林土壤样本为研究对象,研究基于偏最小二乘回归(PLSR)和径向基函数(RBF)神经网络的组合模型预测土壤有机质含量的适用性,并且通过构建有机质含量与有效铁含量的二项式线性模型,对土壤有效铁含量进行间接反演,探讨不同权重下的最优组合模型。结果表明,组合模型的预测效果优于偏最小二乘回归和RBF神经网络单个模型,并且熵值组合为最优组合模型,其中,土壤有机质的反演模型验证的决定系数(R~2)为0.81,均方根误差(RMSE_p)为11.54 g kg~(-1),测定值标准差与标准预测误差的比值(RPD)为2.18;有效铁的间接反演模型R~2为0.70,RMSE_p为21.60 mg kg~(-1),RPD为1.77。通过土壤有机质构建土壤有效铁含量的光谱反演间接模型,在光谱反演模型中,组合模型能较大限度地利用各种预测样本信息,能有效减少单个预测模型中随机因素的影响,增强预测稳定性,提高模型的预测能力。因此,组合模型可对土壤有机质含量的光谱预测及土壤有效铁的间接预测发挥更好的作用。
[Abstract]:Iron is one of the most important micronutrients in plant growth. It is of great significance to obtain soil available iron content information by means of soil spectral prediction technology, but it is difficult to predict soil available iron content directly by soil spectrum. Therefore, using the correlation between soil organic matter content and available iron content, the feasibility of indirectly estimating soil available iron content was discussed. The applicability of the combined model based on partial least square regression (PLSRS) and radial basis function (RBF) neural network for predicting soil organic matter content was studied. The binomial linear model of organic matter content and available iron content was constructed. Indirect inversion of soil available iron content was carried out to discuss the optimal combination model with different weights. The results show that the prediction effect of the combined model is better than that of the partial least square regression and the single RBF neural network model. And the entropy value is combined as the optimal combination model, where, The determination coefficient of soil organic matter inversion model was 0.81, the root mean square error (RMSEp) was 11.54 g / kg ~ (-1), the ratio of standard deviation to standard prediction error was 2.18, and the indirect inversion model of effective iron was 0.70 ~ (-1) mg 路kg ~ (-1) 路kg ~ (-1) ~ (-1) ~ (-1) ~ (-1) ~ (-1) ~ (-1) ~ (-1) and 0.70 ~ (-1) mg 路kg ~ (-1) 路kg ~ (-1) 路kg ~ (-1) 路kg ~ (-1) 路L ~ (-1) of RMSEp, respectively. The indirect spectral inversion model of soil available iron content was constructed by soil organic matter. In the spectral inversion model, the combined model can make use of all kinds of prediction sample information to a large extent, can effectively reduce the influence of random factors in a single prediction model, enhance the prediction stability and improve the prediction ability of the model. The combined model can play a better role in the spectral prediction of soil organic matter content and the indirect prediction of soil available iron.
【作者单位】：江西农业大学江西省鄱阳湖流域农业资源与生态重点实验室/林学院;
【基金】：国家自然科学基金项目(41361049) 土壤与农业可持续发展国家重点实验室项目(0812201202)资助~~
【分类号】：S714

【相似文献】