Sparrow(spatially referenced regressions on watershed) model is a watershed loading model which consists a kernel of nonlinear regression function. Spatial referencing is accomplished through linking nutrient source, land-surface characteristic, and loading information to a geographically defined river-reach data set. The model evaluated and predicted contaminant flux, concentration in streams and identified important contaminant sources and transport variables over large spatial scales. The mechanics of the model was comprehensively reviewed, and the model was applied to a virtual watershed in northern China for total nitrogen estimation and prediction. The result of SPARROW model indicated that agricultural sources in the watershed contribute 63.7% of total nitrogen, follow by wastewater nitrogen source (19.8%), whereas industrial source is the least contributor of total nitrogen (16.5%).