Hydrology models DRAINMOD and SWIM applied to large soil lysimeters with artificial drainage

Abstract

Two hydrological models, which used different methods to determine the soil water distribution in a soil profile, were evaluated against 4 years of data from large soil lysimeters. SWIM determines soil water distribution from a finite difference implementation of the Richards" equation. DRAINMOD uses a soil-specific relationship between the air volume in a profile and the watertable height to locate the depth to the saturated zone. An ‘equilibrium’ relationship between soil water tension and depth is then assumed to distribute the soil water in the unsaturated zone. Predicted values and measured values for drainage and watertable heights were compared for 3 drainage treatments. The drainage in the lysimeters was achieved by installing an outlet tube on the slowly permeable layer at 0·75 m from the soil surface. The conventional drainage treatment allowed gravity drainage to occur directly from this drainage outlet tube. The other 2 drainage treatments employed controlled drainage, where a step (or weir) is installed in the outlet tube. No drainage can occur from the lysimeters until the water table within the lysimeters reaches the step height. Two different step heights provided 2 controlled drainage treatments. Independently determined model parameters were used without additional calibration for the analysis. Both models performed well. DRAINMOD over-predicted the 4-year cumulative drainage for all treatments, with the largest error being 7%. SWIM conversely under-predicted cumulative drainage, with a maximum error of 16%. The standard error of estimation for the watertable height over the full 4-year data period was lower for SWIM, ranging from 0 ·06 to 0·12 m. DRAINMOD’s standard error over the same period for the watertable height ranged from 0·09 to 0·21 m. Generally, error values from this work were smaller than comparable values from other studies. The hydrology of the lysimeters where there was no lateral inflow, surface runoff, or deep seepage losses, coupled to an essentially 1-dimensional flow domain, probably contributed to the lower errors. Furthermore, limitation of the maximum watertable heights by the controlled drainage regime in the lysimeters also reduces the maximum possible magnitude of the standard error term.