Comparison of performance of tile drainage routines in SWAT 2009 and 2012 in an extensively tile-drained watershed in the Midwest

Abstract. Subsurface tile drainage systems are widely used in agricultural watersheds in the Midwestern US and enable the Midwest area to become highly productive agricultural lands, but can also create environmental problems, for example nitrate-N contamination associated with drainage waters. The Soil and Water Assessment Tool (SWAT) has been used to model watersheds with tile drainage. SWAT2012 revisions 615 and 645 provide new tile drainage routines. However, few studies have used these revisions to study tile drainage impacts at both field and watershed scales. Moreover, SWAT2012 revision 645 improved the soil moisture based curve number calculation method, which has not been fully tested. This study used long-term (1991–2003) field site and river station data from the Little Vermilion River (LVR) watershed to evaluate performance of tile drainage routines in SWAT2009 revision 528 (the old routine) and SWAT2012 revisions 615 and 645 (the new routine). Both the old and new routines provided reasonable but unsatisfactory (NSE  <  0.5) uncalibrated flow and nitrate loss results for a mildly sloped watershed with low runoff. The calibrated monthly tile flow, surface flow, nitrate-N in tile and surface flow, sediment and annual corn and soybean yield results from SWAT with the old and new tile drainage routines were compared with observed values. Generally, the new routine provided acceptable simulated tile flow (NSE  =  0.48–0.65) and nitrate in tile flow (NSE  =  0.48–0.68) for field sites with random pattern tile and constant tile spacing, while the old routine simulated tile flow and nitrate in tile flow results for the field site with constant tile spacing were unacceptable (NSE  =  0.00–0.32 and −0.29–0.06, respectively). The new modified curve number calculation method in revision 645 (NSE  =  0.50–0.81) better simulated surface runoff than revision 615 (NSE  =  −0.11–0.49). The calibration provided reasonable parameter sets for the old and new routines in the LVR watershed, and the validation results showed that the new routine has the potential to accurately simulate hydrologic processes in mildly sloped watersheds.