Testing an Artificial Neural Network for Predicting Soil Hydraulic Conductivity

Multilinear regression has been used extensively to predict soil hydraulic properties, both the θ( h ) and K ( h ) relationships, from easily obtainable soil variables. As an alternative, this study investigated the performance of an artificial radial basis neural network in predicting some K ( h ) values from other variables. This kind of neural network may be seen as a multivariate interpolation technique, which can theoretically fit any nonlinear continuous function. Neural networks are characterized by parameters that must be optimized to solve a given problem. We used a fitting procedure requiring only two parameters to ensure a unique solution. These two parameters were determined by data splitting. Hypothetical data bases with uncertainties were simulated to analyze the performance of the neural network in predicting a nonlinear relation derived from a classical model for K ( h ). A soil data base covering a broad spectrum of soil horizons was used to test the neural network in solving multivariate problems. Numerical simulations showed that the neural network was sensitive to large uncertainties in the data base. It was more efficient than a multilinear regression when the uncertainties were small. Experimental results showed that the neural network was more efficient than the multilinear regression for predicting K ( h = −1 m) or K ( h = −2.5 m) from two qualitative and five quantitative soil variables. It was also more efficient than two independent multilinear regressions, one for the sandy samples and the other for the loamy and clayey samples. Provided that a large data base with accurate K values is available, artificial neural networks can be useful to predict θ( h ) and K ( h ) over a broad spectrum of soils.