Assessment of Infiltration Rate of Soil Using Empirical and Machine Learning-Based Models

The application of several infiltration models in evaluating the infiltration rate of soil is subject to the spatial variability of soil. Infiltration plays an important role in designing and evaluating surface irrigation systems as well as subsurface groundwater recharge systems. The main theme of this paper is to compare the empirical equation-based models which are used to estimate the infiltration rate of various locations in Kurukshetra, India. Infiltration experiments were conducted at 20 different locations using a mini disc infiltrometer. The least square fitting method was used for each location separately to estimate the equation parameters and infiltration rate of three different empirical models, namely Kostiakov, modified Kostiakov and Novel models. The performance of these empirical infiltration models was further compared with the machine learning-based adaptive neuro-fuzzy inference system (ANFIS) and random forest regression (RF) techniques. Statistical performance evaluation parameters (R-2, CC, RMSE and MAE) are used for the performance comparison of various models. The Novel infiltration model was observed to be the most reasonable amongst the empirical models tested. However, for machine-learning methods, the RF approach is observed to be the most appropriate technique for the estimation of the infiltration data. (c) 2019 John Wiley & Sons, Ltd.