Uma Aplicação dos Algoritmos Genéticos e do Método dos Volumes Finitos para Simular o Fluxo da Água na Zona Vadosa

ABSTRACT

The water flux in porous media obeys a nonlinear partial differential equation (the Richards equation), whose the mixed form contains the variables volumetric water content and matric potential. The soil water retention curve is the nonlinear relation between the volumetric water content and matric potential, essential for analysing the soil water dynamics in the vadose zone. The main purpose of this paper is to evaluate the best fit model for the retention curve considering experimental analysis realized at Universidade Federal Rural do Rio de Janeiro. In this work, a genetic algorithm (GA) is proposed in order to search the design variables that optimize the coefficient of determination (R Squared), considering three soil horizons and the van Genuchten, Brooks-Corey and Haverkamp models to describe the retention curve. The GA proposed was evaluate comparing the results with the software SWRC Fit , which performs nonlinear fitting of soil water retention curves by Levenberg-Marquardt method. GA adjusted the measured data more precisely and the Haverkamp model presented the highest coefficient of determination. In addition, the discretization for the equation Richards was more stable using the Haverkamp model as a soil water retention curve.

Keywords:
laboratory tests; Richards equation; non-linear regression