Modeling subsurface drainage in Clermont silt loam using the finite element technique

Mauro Aparecido Martinez, Purdue University

Abstract

A finite element model was used to predict drain flow hydrographs and water table depths in an artificially drained Clermont silt loam soil. The field data were obtained from a drainage project at the Southeast Purdue Agricultural Center (SEPAC). Vertical and horizontal saturated hydraulic conductivities measured in 6 soil layers at 6 vertical sections showed a high spatial variability for that property and an anisotropic soil profile. The model incorporates rectangular and triangular linear elements; it uses two methods (influence coefficient and numerical integration) for calculating the element matrices; and two methods (Brooks and Corey and van Genuchten) for estimating the unsaturated hydraulic conductivity. The model also uses a finite difference technique to approximate the ordinary differential equation in time resulting from the finite element discretization of the spatial domain. The time is also automatically controlled. The simulated drain flow hydrographs for 5, 10 and 20 m drain spacings, predicted the shape of the measured hydrographs quite well. However, the model underestimated the peak flow in a dry year (1987) and overestimated the hydrographs in a wet year (1989). The predicted water table for the wet year was higher than the measured water table. Also, when deep seepage was included in the model, both the predicted water table depth and shape were in better agreement with the field conditions. A probabilistic method, the modified point estimate method, was combined with the finite element model to transfer the variability in the measured saturated hydraulic conductivity to the output of the model. The simulations showed that the flow velocities were more sensitive to the variability in conductivity than were the pressure potentials. While the large coefficients of variation for velocities occurred at the region far away from the drain where the velocities were low, the large coefficients of variation for pressure potentials occurred close to the drain tube. The coefficients of variation for the drain flow rates decreased as the water table depth increased for both 5 and 10 m drain spacings, but were lowest for the 5 m drain spacing. (Abstract shortened with permission of author.)

Degree

Ph.D.

Advisors

Monke, Purdue University.

Subject Area

Agricultural engineering|Hydrology

COinS