Developing a simple artificial intelligence fuzzy-based model for estimating saturated hydraulic conductivity of soil

Saturated hydraulic conductivity is one of the important physical properties of soil in modeling water and solute transport, irrigation management, and drainage issues. Laboratory and field methods for directly measuring this parameter are time-consuming and costly. In recent years, the use of intelligent systems for estimating various soil parameters has significantly increased. Therefore, this research aims to utilize Fuzzy Inference Systems (FIS), Artificial Neural Networks (ANN), and Linear Regression (LR) to create a mapping between soil texture parameters and saturated hydraulic conductivity. The data used in this study includes physical properties related to 331 soil samples from the UNSODA soil database (170 samples) and existing data from soils in the cities of Amol, Babol, Karaj (50 samples), and Shahrekord (111 samples). After examining different models and combinations of available data, three models were proposed for estimating saturated hydraulic conductivity. In these models, saturated hydraulic conductivity was estimated using soil texture characteristics (percentage of clay, silt, and sand) and bulk density, and the performance of the models was evaluated using statistics such as root mean square error (RMSE), mean bias error (MBE), and coefficient of determination (R²). Comparing the results of the proposed nonlinear models with different input parameters showed that fuzzy systems can estimate saturated hydraulic conductivity with acceptable accuracy. During the training phase, the fuzzy model with four input variables (percentage of clay, silt, sand, and bulk density) had the highest correlation (r = 0.92), and considering other evaluation parameters (R² = 0.84, MBE = 0.28 cm/hr, and RMSE = 1.64 cm/hr), it showed a good fit with the measured values. In the testing phase, similar results were obtained, and the fuzzy model with four parameters had the best fit. Based on the results of this research, the fuzzy model can be introduced as one of the methods for estimating saturated hydraulic conductivity with suitable accuracy.