SWAT modeling with uncertainty and cluster analyses of tillage impacts on hydrological processes

The impacts of tillage practices, majorly conventional tillage (CT) and no-till (NT), on soil hydraulic properties have been studied in recent decades. In this paper, we incorporated an auto-calibration algorithm into the Soil and Water Assessment Tool (SWAT) model and calibrated the model at eight field sites with soil water content (SWC) observations in the Pataha Creek Watershed, WA, USA. The Green-Ampt method in SWAT was chosen to determine infiltration and surface runoff. Parameter uncertainty was quantified by relatively optimal parameter sets filtered by a critical objective function value. Cluster analysis was adopted to obtain equal-sized parameter sets for each site and to compare parameter sets between tillage practices. The centers of these clusters were employed as a sample of parameter values. The clustered parameter sets were then used in scenario analysis to examine the impacts of cropland tillage practices on lateral flow, runoff and evapotranspiration (ET). The model parameters (e.g., soil hydraulic properties) were significantly different between CT and NT. In particular, higher bulk density, larger available water capacity, and higher effective hydraulic conductivity were found for NT than for CT. SWCs at three depths of the NT sites were significantly higher than those of CT sites, which could be attributed to tillage practices. However, higher available water capacity at NT sites indicated that the NT soil had a higher capacity to hold water. Thus the mean net changes in SWC during a year were not significantly different between CT and NT. The statistically different model parameters neither resulted in statistical differences in annual outputs (e.g., runoff and ET) nor substantial differences in monthly outputs. Our study indicates that the tillage impacts on hydrological processes are site-specific and scale-dependent.