It is a spatial regression (nonlinear) model that is developed using observed information in the US to predict loading on a raster scale. It is data-dependant and need to be assessed with data from your location. It doesnot model spatial flows and is for distributed parameter assessment. If you have monitoring data spread over representative areas, it will e useful to apply this model to quantify coefficients for you location.
SPARROW, a modeling tool for the regional interpretation of water-quality monitoring data. The model relates in-stream water-quality measurements to spatially referenced characteristics of watersheds, including contaminant sources and factors influencing terrestrial and aquatic transport. SPARROW empirically estimates the origin and fate of contaminants in river networks and quantifies uncertainties in model predictions. You will find interesting information in the following links
As with all models, sufficient field data is needed for calibration and validation. For SPARROW, this usually means stream or riverine load data, which can be time intensive/costly to collect. Without enough validation data, your modeled numbers don't mean very much.
My opinion is that models like SPARROW can be very useful to help understand the watershed (e.g. what subbasin is the source of the most phosphorus flux) or for use in management scenario modeling, but cannot replace field based measurements.