关键词: BP神经网络, RBF神经网络, 土壤盐渍化, 预测

Abstract: Aiming at the problem indynamic monitoring the soil salinity in the oasis of the Ugan Kuqa River Delta in Xinjiang, we analyzed the factorsaffectingthe soil salinization withthe gray-correlation-degree model, and determinedthe degree of the association between the factors and the soil salinity. The artificial intelligence technology was utilized in the soil salinity prediction. After several adjustments on the network structure and parameters, we established a BP neural network model and a RBF neural network model to predict surface soil salinity. The results showed that the BP network model and RBF network model, takingthe evapotranspiration, groundwater depth, groundwater mineralization, soil conductivity, total dissolved salts, pH value, slope and land-use type as input factors andsoil salinity as output factor, couldeffectively simulate soil salinity and its impact to the inherent complexity of the relationship between the factors.The prediction error of the BP network model was less than that of the RBF. The present study could provide an effective and viable new way for analyzing and predicting the soil salinization. Itcould be a complement to the traditional dynamicsmonitoringon soil salinity.

Key words: BP neural network, RBF neural network, soilsalinity, prediction