Prediction of Vegetation-Induced Soil Suction Using Numerical Modelling and AI

Abstract

Study of soil suction is important in design and implementation of slope stability and erosion control measures. In order to conduct a realistic analysis of performance of sustainable green infrastructure, it is essential to address the uncertainties in suction induced by vegetation due to variability in their leaf and root characteristics, evapotranspiration (ET) and initial conditions of the soil. The objective of this study is to investigate the combined influence leaf area index (LAI), root depth, ET rate and initial suction of soil on root water uptake-induced soil suction. A parametric numerical study was performed with 480 simulations using HYDRUS to carry out the finite element analysis. The study was done on completely decomposed granite (CDG) soil and vegetation species used was Schefflera heptaphylla. It was observed that although if independently considered, vegetation with higher LAI provided greater mechanical stability, when combined with higher ET rates or initial suction, the suction induced may lead to wilting of the vegetation. Artificial intelligence technique such as Artificial neural network (ANN) was used to predict matric suction at any given depth using the results obtained from the numerical simulations. Performance of the best model indicated that ANN was able to successfully predict the vegetation-induced matric suction.