黄土土壤水分运动参数预测模型研究  

Research on Prediction Model of Loess Soil Water Movement Oarameters

作　　者：秦文静[1] 樊贵盛[1] QIN Wen-jing;FAN Gui-sheng(Taiyuan University of Technology,Taiyuan 030024,Shanxi Province,China)

机构地区：[1]太原理工大学,山西太原030024

出　　处：《中国农村水利水电》2024年第7期229-233,248,共6页China Rural Water and Hydropower

基　　金：山西省基础研究计划项目(202103021223112);山西省水利科学技术研究推广项目(2023GM26)。

摘　　要：土壤水分运动参数作为量化土壤水分保持和运动过程的必要输入参数(非饱和导水率、比水容量等),其准确获取决定着基本微分方程的计算精度,进而决定了土壤水分运动模型的拟合精度。以黄土65个取样点为研究对象,构建了基于土壤常规理化参数的原状黄土土壤水分特征曲线、土壤非饱和导水率的BP神经网络预测模型,结果表明:运用BP神经网络模型获取土壤水分运动参数是可行的;所构建的模型对训练样本和验证样本均有良好的预测效果,训练样本相对误差均小于6.4%、验证样本相对误差小于5.7%。研究成果对于推进土壤水动力学理论发展和农业土壤水管理及其水文预报理论发展具有重要的理论价值,对于提升农业水资源管理水平具有重要的实际价值。Soil water movement parameters are necessary input parameters for quantifying soil water retention and movement process.Unsatu⁃rated water conductivity,specific water capacity and accurate acquisition of relevant model parameters determine the calculation accuracy of the basic differential equation,and then determine the fitting accuracy of the soil water movement model.This paper takes 65 loess sampling points as the research object,and constructed a BP neural network prediction model of soil water characteristic curve and unsaturated water conductivity of undisturbed loess,based on the conventional physical and chemical parameters of soil.The results show that it is feasible to obtain soil water movement parameters by using the BP neural network model.The constructed model has good prediction effect on both train⁃ing samples and verification samples,the relative error of the training sample is less than 6.4%,and the relative error of the verification sam⁃ple is less than 5.7%.The research results have important theoretical value for promoting the development of soil hydrodynamics theory,agri⁃cultural soil water management and hydrological forecasting theory,and have important practical value for improving the level of agricultural water resources management.

关 键 词：黄土 土壤水分特征曲线 非饱和导水率 BP神经网络 

分 类 号：TV93[水利工程—水利水电工程] S27[农业科学—农业水土工程]