Modeling subsurface drainage using steady-state and transient methods

Abstract. Computer models have been widely used to simulate the hydrologic and environmental processes at work in subsurface-drained cropland. However, some models (e.g., RZWQM2) hold back tile drainage until rainfall and infiltration have ceased, resulting in predicted drainage being seriously delayed when rainfall events extend beyond 10 h. While the steady-state Hooghoudt equation (ssH) has been widely used in simulating daily drainage flow, transient methods such as the integrated Hooghoudt (inH) and van Schilfgaarde (vanS) equations have seldom been tested. In addition to testing whether allowing soil moisture redistribution and rainfall to occur simultaneously would improve drainage flow simulation, the performance of the ssH equation was compared to that of the transient inH and vanS equations in simulating daily and hourly tile drainage. Implementing the ssH equation by default, the Root Zone Water Quality Model (RZWQM) served as a base model. Field data including drainage flow, weather, soil, and crop parameters were collected between 2006 and 2008 in a tile-drained field under a corn-soybean rotation situated in Iowa, U.S.A. Compared to maintaining drainage rate constant until rainfall ceased, allowing soil moisture redistribution to occur along with rainfall enhanced the performance of the base model (ssH equation): the percent of bias (PBIAS), Nash-Sutcliffe efficiency (NSE) and Index of Agreement (IoA) improved from 11.6%, 0.39, and 0.69, respectively, to 6.5%, 0.71, 0.76. Under the same moisture redistribution alternative the inH and vanS transient drainage equations provided similar improvement in model accuracy compared to the ssH equation. Overall, the inH and vanS equations (PBIAS = 6.4%, NSE = 0.70 and IoA = 0.76) showed little difference from the ssH equation (PBIAS = 6.5%, NSE = 0.71, and IoA = 0.76) in predicting daily drainage flow occurring concurrently with rainfall. However, when tested against hourly drainage flow data, the inH and vanS transient equations showed slightly higher drainage peaks, closer to the observed peaks, than the ssH equation. Furthermore, results were similar when applying these equations to the original unmodified RZWQM. Allowing soil moisture redistribution to occur simultaneously with rainfall significantly improved model performance in simulating tile drainage, but the use of the transient inH and vanS equations (vs. the ssH) had no appreciable effect on model accuracy.