Geologic control of soil-infiltration rate based on artificial neural network models

Abstract

The interconnected porosity of soil provides conduit channels for the downward infiltration of water into the subsurface; this occurs in soil layers and within soil-less areas or geologic formations. The lithology and geological structure significantly influence the infiltration capacity of soils and are crucial in determining whether the infiltration water continuously reaches an aquifer or becomes stagnant in the saturated soil. An artificial neural network (ANN) algorithm was employed to model the actual infiltration rate, incorporating soil texture and soil moisture along with geological scores as inputs and actual infiltration rates as outputs. This study aimed to quantify qualitative geological data and incorporate it into ANN model parameters. The development of the ANN infiltration model involved two serial trial-and-error experiments to determine the optimal number of nodes in the hidden layer, ranging from nodes c(4,2) to c(12,2), one serial experiment withgeological input, and the other without geological input. Throughout the model testing, metrics such as MAE, RMSE, and MSE were recorded, and the first and second optimum models were identified when employing c(9,2) nodes of hidden layers. The resulting model can be used to predict actual infiltration and will be beneficial for hydrometeorological-disaster mitigation and city-development planning.