Estimating the final constant infiltration rate with infiltration models and pedotransfer functions

Infiltration characteristics of the soil constitute one of the basic and most important parameters for proper design and management of field irrigation systems. The design, operation, management and hydraulic evaluation of on-farm irrigation methods depend on the water infiltration properties of the soil. Empirical equations and physically based equations used to describe vertical infiltration into simple soil systems are compared with experimental data. The direct measurement of soil hydraulic characteristics is troublesome, time-consuming and expensive. Alternative approaches called pedotransfer functions (PTFs) for the predictions of the soil hydraulic parameters have been continuously developed by many researchers in the world. The accuracy and efficiency of the existing methods for estimating soil infiltration rate requires understanding of the performance of each model and its limitations.
The objectives of this study were to evaluate the performance and efficiency of physical equations such as Green Ampt and Philip; three empirical equations including Horton, Holtan and Kostiakov and also pedotransfer functions for estimating the final constant infiltration rate in the soils near the bank of Ghezel Auzan river of Zanjan province. Infiltration rates were measured by the method of double rings on 30 points and with 3 replications. The percentage of soil component was determined by hydrometric method. The initial and saturation moisture content were obtained with gravimetric method and the suction of 33 kPa (FC) was determined by the pressure plate. Saturated hydraulic conductivity (Ks) obtained using the constant head method and EC (ECe) was read by the EC meter. All data were categorized in the database that created in SPSS software and the normalization and homogenization of the data was done with the Shapiro Wilk tests. Prediction performances of these methods were evaluated using RMSE, MAE, ME, RE and EF criteria between the predicted and the measured values. For creating the pedotransfer function models, two multiple linear regression and Ridge regression using easily accessible physical and chemical soil characteristics were used in three input data structure groups.
Results of evaluation of these equations indicated that the Holtan equation had a better performance than Horton and Kostiakov equation. The Green Ampt equation estimated the best final infiltration rate with less error (RMSE =1.89 cm.h-1 and MAE =1.37 cm.h-1) and the higher coefficient of determination (R2 =0.95) compared with other experimental, physical and pedotransfer function models. In comparing the results of the physical and empirical equations, the physical equations had higher accuracy than the empirical equations. Other researchers in their studies showed that the physical equations are better than empirical equations. The results showed that the error was reduced in Ridge regression with exclusion collinearity error in easily accessible parameters than Ordinary regression. The t-test results showed that there was no significant difference between empirical equations, physical and pedotransfer functions at P=0.05. The major differences in these models were in access to needed information and the accuracy of the estimated values. ME index showed that some of the models underestimated and some are overestimated. The estimated infiltration rate of Philippe, Grynampt, Horton and Holtan equations were evaluated with measured data and the finding showed the Grynampt and Holtan equations had the best results of estimation and the equation of Horton's had the lower accuracy. Holtan equation uses the properties of soil and vegetation and land use information so it can better estimate the infiltration rate.
The difference between the values of infiltration rate with t-student test showed that the models with empirical, physical and third group PTFs equations had no significant difference at the 5% significance level. The results indicated that all of the models mentioned in this study were able to estimate the final infiltration rate in the study area with acceptable accuracy, but the main difference between these models was the access to the required information. The results showed that Green Ampt physical equation had higher accuracy in estimating the final infiltration rate; however, other empirical and physical equations and transfer functions are well able to estimate the rate of infiltration. Estimate the final infiltration rate with the use of physical and chemical parameters of soils showed ridge regression is better than the simple regression. In the areas that high estimate of infiltration caused the runoff, especially in the design of sprinkling systems the use of models with underestimated the infiltration rate is preferred. The underestimation of these models can be considered as a safety factor in the design of hydraulic systems.