A multi-gene genetic programming model for estimating stress-dependent soil water retention curves

Soil water retention curve (SWRC) is an important parameter required for seepage modelling in unsaturated soil and is used for analysing rainfall-induced slope failures, design of waste contaminant liners and cover, etc. The influence of stress, which is one of constitutive variables that governs unsaturated soil behaviour on the SWRC, has been well recognised by researchers. Stress is essential for study as it drastically alters the soil fabric which includes macropores, minipores and micropores and thus affects the ability of soil to retain water. Various computational modelling techniques that formulate models based on existing databases such as UNSODA, ISRIC and HYPRES for the estimation of SWRC do not take into account the stress influence on soil behaviour. In the present work, three artificial intelligence (AI) methods of support vector regression, artificial neural network and multi-gene genetic programming (MGGP) have been applied to formulate the mathematical relationship between the water content and input variables such as stress and suction (i.e. stress-dependent soil water characteristic curves (SDSWRCs)). The results indicate that the MGGP model outperforms the other two models and is able to extrapolate the water content values satisfactorily along the stress value of 800 kPa. This MGGP model can then be deployed by experts for the estimation of SDSWRCs, thus eliminating the need for conducting costly and time-consuming experiments.