盐碱土壤Kostiakov入渗模型参数的BP预报模型  

Predictive Model of Kostiakov Infiltration Model Parameters of Saline Soil Based on BP

作　　者：李昊哲[1] 樊贵盛[1] 

机构地区：[1]太原理工大学水利科学与工程学院,太原030024

出　　处：《中国农村水利水电》2017年第7期49-53,58,共6页China Rural Water and Hydropower

基　　金：国家自然科学基金资助项目(40671081);山西省科技攻关项目(2007031070)

摘　　要：为改良和改善盐碱地提供土壤入渗参数技术支撑,基于在山西省北部盐碱地进行的野外系列入渗试验,获取了累积入渗量与入渗历时的数据样本,并计算回归了kostiakov二参数入渗模型的入渗系数k与入渗指数α,建立了盐碱土壤基本理化参数与入渗参数之间的数据样本,利用BP神经网络的方法,建立了以土壤含水率、容重、质地、有机质、全盐量以及p H为输入变量,kostiakov入渗参数为输出变量的预报模型。结果表明:盐碱地土壤条件下,以土壤基本理化参数为输入变量,kostiakov入渗模型参数为输出变量的BP预报是可行的,入渗系数k的相对平均误差为0.29%、入渗指数α的相对平均误差为1.28%,以及根据两个入渗参数计算得到90 min累积入渗量的相对平均误差为2.37%,对所建立的模型进行检验时,以上三个参数检验误差的平均值均能控制在3%以下,确定所建立的BP预测模型能获得较好的效果。In order to provide technical support for getting soil infiltration parameters to improve the saline soils, based on a series of field infiltration experiments in the saline-alkali soils that in the north of Shanxi Province, and from that the cumulative water infiltration amount and infiltration duration data were obtained and the Kostiakov infiltration coefficient k and infiltration index α were obtained by regression calculation. Finally the data samples including the basic physical and chemical parameters and the infiltration parameters of the saline soil were established. The forecast model based on the BP neural network were built, then the soil water content, bulk density, texture, organic matter, total salt and pH were used as input variables and Kostiakov infiltration parameters were used as output variables. The research resuhs show that under the condition of saline soil, it is feasible to establish BP forecasting model by using the basic physical and chemical parameters as the input variable as well as the Kostiakov infiltration model as output variable. The relative average error of the infiltration coefficient k and the infiltration index α were 0.29% and 1.28% respectively and 90 min cumulative was 2.37%. The average error of the above three parameters can be controlled within less than 3% in the test of the established model, so that it can be concluded that the established BP predicting model is able to obtain good results.

关 键 词：盐碱土壤 Kostiakov二参数模型参数 土壤基本理化参数 BP神经网络 水分入渗 

分 类 号：S152.7[农业科学—土壤学] TV93[农业科学—农业基础科学]