浅地下水埋深条件下土壤水盐动态BP网络模型研究

BP network model for soil water-salt movement with a shallow groundwater table

针对浅地下水埋深条件下作物生育期内根系层土壤水盐动态模拟中存在的问题,将人工神经网络引入水盐动态的模拟和预报中,建立了根系活动层0～60 cm和0～100 cm深度内土壤水盐动态的BP网络模型.结果表明,以生育时段初平均土壤含水率、平均土壤盐分指标、地下水水位埋深、地下水盐分指标、时段内水面蒸发量、降雨量(包括灌水量)、生育期日序列7个因素为输入因子,以生育时段末平均土壤水分、平均土壤盐分指标为输出因子的BP网络模型可有效表征土壤水盐动态及其影响因素之间的内在复杂关系,并且有较高的精度.该研究为分析浅地下水埋深条件下作物生育期内土壤水盐动态规律的分析提供了一种有效可行的方法,是对传统土壤水盐动态研究的补充.

Aiming at some problems of modeling soil water-salt movement with a shallow groundwater table, a BP network model was introduced for modeling and forecasting field soil water-salt in the crop root region at the depth from 0 to 60 c m and from 0 to 100 cm. The BP network input factors are the mean values of soil water content and salinityin preliminary period, groundwater level and salinity, evaporation of water surface, rainfall （including the amount of irrigation） during the period, and the day series of crop growth duration, the output factors are the mean values of soil water content and salinity. The results show that BP network model may deal with the complex relation well between soil water-salt movement and its influencing factors. The model may forecast field soil water-salt movement accurately. This study offers an effective and feasible method for analyzing soil water-salt movement and is a supplement for the traditional dynamic study of soil water-salt movement.