Abstract. Model-data comparisons are always challenging, especially when working at a large spatial scale and evaluating multiple response variables. We implemented the Soil and Water Assessment Tool (SWAT) to simulate water quantity and quality for the Tennessee River Basin. We developed three innovations to overcome hurdles associated with limited data for model evaluation: 1) we implemented an auto-calibration approach to allow simultaneous calibration against multiple responses, including intermediate response variables, 2) we identified empirical spatiotemporal datasets to use in our comparison, and 3) we compared functional patterns in landuse-nutrient relationships between SWAT and empirical data. Comparing monthly SWAT-simulated runoff against USGS data produced satisfactory median Nash-Sutcliffe Efficiencies of 0.83 and 0.72 for calibration and validation periods, respectively. SWAT-simulated water quality responses (sediment, TP, TN, and inorganic N) reproduced the seasonal patterns found in LOADEST data. SWAT-simulated spatial TN loadings were significantly correlated with empirical SPARROW estimates. The spatial correlation analyses indicated that SWAT-modeled runoff was primarily controlled by precipitation; sedimentation was controlled by topography; and NO₃ and soluble P were highly influenced by land management, particularly the proportion of agricultural lands in a subbasin