季节性冻融土壤水盐动态预测BP网络模型研究

BP network model research on water and salt transfer forecast in seasonal freezing and thawing soils

冻融土壤水盐运移模拟预测一直是冻土物理学和土壤水科学研究的热点和难点,而冻土水盐运动的特殊规律与分配特性是影响河套灌区土壤盐渍化发生、发展和演变的重要因素。该文基于内蒙古河套灌区5个冻融期（1994～1999）的水分、盐分和温度的田间实测资料,利用人工神经网络对冻融土壤水盐耦合运移进行了联合预测。结果表明：初冻期地下水埋深、初冻期含水率、初冻期含盐量、秋浇水量、11月到翌年4月的月平均地温等10个影响因子可以有效表征冻融土壤水分和盐分之间的强烈耦合关系,神经网络预测模型具有较好的精度。该研究为冻融条件下土壤水盐动态预报和灌区灌溉管理的研究开辟了新的途径。

Prediction and simulation of water and salt transfer in freezing and thawing soils have always been a considerable attention and difficult research on the physics of frozen soils and soil water science. It is considered that the movement and distribution rule of frozen soil water and salt is a significant factor on the evolution and transition of soil salinity and alkalinity. Based on field measurements of water, salt and temperature for five years （1994- 1999） overwintering in Inner Mongolia Hetao Irrigation District, the model of artificial neural network on soil water and salt transfer under freezing-thawing conditions was established. The results show that BP network of ten factors i.e. groundwater table, water content, salt content, irrigation amount and average monthly soil temperature from November to April next year, can reflect the coupling relation between soil water and salt under freezing and thawing conditions. The BP network may forecast field coupling transfer of soil water and salt under freezing and thawing conditions accurately. This study offer an effective and feasible method for manages irrigation district and forecast of soil water and salt transfer under freezing-thawing conditions.