Estimation of Some Soil Hydraulic Characteristics of Jefara Plain Using Artificial Neural Networks

Abstract

Direct estimations of some soilproperties are costly and time-consuming. Therefore, they can be estimated using some readily available soil characteristics through pedotransfer functions technique. In this study, a number of pedotransfer functions were developed by artificial neural networks to estimate some of the soil hydraulic characteristics from some physical soil properties, which can be easily obtained in the Jefara plain. Data were collected for different soil samples from several sites in Jefara plain area. These data were extracted from many studies, reports and projects during the previous periods and have been used to develop the models for estimating field capacity, permanent wilting point, saturated hydraulic conductivity and hygroscopic water. The collected data were divided into two groups. The first group included training and internal calibration. The second group of data which was not used in developing neural network models was used to test and evaluate these models. Many statistical parameters were used to assess the accuracy of the developed models. They included the mean absolute error (MAE), root mean square error (RMSE), (RMSE %) and Nash and Sutcliffe Efficiency (NSE). The linear regression equation relating predicted data with measured data with intercept equals zero and determination coefficient (R²) were also used for evaluation purpose. According to the statistical parameters of the evaluation criteria, the results revealed that the developed artificial neural networks models for estimating field capacity, permanent welting point, saturated hydraulic conductivity and hygroscopic water provided reliable and high performance. However, the study strongly recommends using the developed models for estimating the field capacity, permanent welting point, soil saturated hydraulic conductivity and hygroscopic water for Jefara Plain soil

Keywords: Filed Capacity, Permanent Wilting Point, Hydraulic conductivity, Artificial Neural Networks.