不同覆盖类型土壤电阻率影响因子及其PLS和BP模型的预测研究  

Influencing Factors of Soil Electrical Conductivity of Different Coverage Types and Its Prediction Based on PLS and BP Neural Network Model

作　　者：冯旭宇[1,2] 刘晓东[2,3] 石湘波 李庆君[3] 王卫红[5] 宋昊泽 博格 刘翠[5] 

机构地区：[1]内蒙古自治区生态与农业气象中心,内蒙古呼和浩特010051 [2]内蒙古自治区气象科学研究所 [3]内蒙古自治区雷电预警防护中心 [4]浙江省宁波市气象局 [5]巴彦淖尔市气象局

出　　处：《现代农业科技》2017年第9期198-201,208,共5页Modern Agricultural Science and Technology

基　　金：内蒙古自治区自然科学基金(2015MS0410);内蒙古自治区气象局科技创新项目(nmqxkjcx201408)

摘　　要：土壤电导率是反映土壤质量和物理性质的重要参数。本研究通过对试验区不同覆盖类型下土壤温度、含水量及电导率的测试,探讨土壤温度和含水量对土壤电阻率的影响。结果表明,不同土壤覆盖类型土壤温度的变化对土壤电阻率的影响不同,土壤电阻率随着土壤含水量的增加而逐渐变小。将偏最小二乘回归模型(PLS)与BP神经网络模型应用于土壤电阻率的预测,PLS模型及BP神经网络模型对土壤电阻率预测皆有较好效果,偏最小二乘回归模型对沙地和草地土壤电阻率预测的误差较小,BP神经网络对农田土壤电阻率建模精度较为理想。The soil electrical conductivity is an important parameter to reflect soil quality and physical properties.ln this study,the soil temperature, water content and electrical conductivity of different coverage types of farmland, grassland and sand were tested to explore the effect of soil temperature and water content on soil electrical conductivity. The results showed that different coverage types had different effects on soil electrical conductivity, the soil electrical conductivity decreased gradually with the increase of soil water content. The partial least squares regression （PLS）model and BP neural network model were applied predict soil electrical conductivity, and got good prediction effects. The PLS model had little error in the prediction of soil electrical conductivity of grassland and sand ,and BP neural network model was more ideal for modeling farmland soil electrical conductivity.

关 键 词：土壤电阻率 覆盖类型 偏最小二乘回归 BP神经网络 预测 

分 类 号：X43[环境科学与工程—灾害防治] P4[天文地球—大气科学及气象学]