Multi-objective optimization was used to calibrate a regional surface water-groundwater model of the Yaqui Valley, a 6800 km<sup>2</sup> irrigated agricultural region located along the Sea of Cortez in Sonora, Mexico. The model simulates three-dimensional groundwater flow coupled to one-dimensional surface water flow in the irrigation canals. It accounts for the spatial distribution of annual recharge from irrigation, subsurface drainage, agricultural pumping, and irrigation canal seepage. The main advantage of the calibration method is that it accounts for both parameter and model structural uncertainty. In this case, results show that the effect of including the process of bare soil evaporation is significantly greater than the effects of parameter uncertainty. Furthermore, by treating the different objectives independently, a better identification of the model parameters is achieved compared to a single-objective approach, since the various objectives are sensitive to different parameters. The simulated water balance shows that 15–20% of the water that enters the irrigation canals is lost by seepage to groundwater. The main discharge mechanisms in the Valley are crop evapotranspiration (53%), non-agricultural evapotranspiration and bare soil evaporation (19%), surface drainage to the Sea of Cortez (15%), and groundwater pumping (9%). In comparison, groundwater discharge to the estuary was relatively insignificant (less than 1%). The model was further refined by identifying zonal <i>K<sub>v</sub></i> and <i>K<sub>h</sub></i> values based on a spatial analysis of the model residuals.