New prediction models for infiltration rate in soil using multi-gen genetic programming and artificial neural networks

Infiltration is the process of water flow through the soil, characterization of which is of high importance for many engineering applications. The determination of infiltration rate is required for estimating effective rainfall, and groundwater recharge. The design of irrigation system would not be possible without knowing this parameter. Furthermore, the knowledge of infiltration rate is essential for modeling hydro-mechanical coupling phenomenon in soil. In the integrated hydrological models and slope stability analysis tools where the analysis of rainfall induced landslides is targeted, it is an important ingredient. There are various empirical models presented in the literature to estimate infiltration rate. Although several physical quantities such as soil porosity, soil moisture content, soil mineralogy, soil grain size distribution, and soil organic content can influence infiltration rate, the available models for estimating this parameter do not fully take these parameters into account, or often overlook them. In this study, artificial neural network and multi-gen genetic programming were used to establish predicting models for infiltration rate, accounting for soil structure (soil grain size distribution and porosity), as well as organic matter.

Keywords:

Infiltration rate , multi-gen genetic programming , artificial neural network , Kostiakov model