We calibrated SPARROW (Spatially Referenced Regression on Watershed Attributes) surface water-quality models using measurements of total nitrogen and total phosphorus from 37 sites in the 13,900-km2 Waikato River Basin, the largest watershed on the North Island of New Zealand. This first application of SPARROW outside of the United States included watersheds representative of a wide range of natural and cultural conditions and water-resources data that were well suited for calibrating and validating the models. We applied the spatially distributed model to a drainage network of nearly 5000 stream reaches and 75 lakes and reservoirs to empirically estimate the rates of nutrient delivery (and their levels of uncertainty) from point and diffuse sources to streams, lakes, and watershed outlets. The resulting models displayed relatively small errors; predictions of stream yield (kg ha-1 yr-1) were typically within 30% or less of the observed values at the monitoring sites. There was strong evidence of the accuracy of the model estimates of nutrient sources and the natural rates of nutrient attenuation in surface waters. Estimated loss rates for streams, lakes, and reservoirs agreed closely with experimental measurements and empirical models from New Zealand, North America, and Europe as well as with previous U.S. SPARROW models. The results indicate that the SPARROW modeling technique provides a reliable method for relating experimental data and observations from small catchments to the transport of nutrients in the surface waters of large river basins.