In 1998 the U.S. Geological Survey, in cooperation with the Trinity River Authority, did a study to estimate storm-runoff pollutant loads using two models—a deterministic model and a statistical model; the estimated loads were compared to loads computed from measured data for a large (118,000 acres) basin in the Dallas-Fort Worth, Texas, metropolitan area. Loads were computed and estimated for 12 properties and constituents in runoff from two 1997 storms at streamflow-gaging station 08048543 West Fork Trinity River at Beach Street in Fort Worth. Each model uses rainfall as a primary variable to estimate pollutant load. In addition to using point rainfall at the Beach Street station to estimate pollutant loads, areal-averaged rainfall for the basin was computed to obtain a more representative estimate of rainfall over the basin. Loads estimated by the models for the two storms, using both point and areal-averaged rainfall, generally did not compare closely to computed loads for the 12 water-quality properties and constituents. Both models overestimated loads more frequently than they underestimated loads. The models tended to yield similar estimates for the same property or constituent. In general, areal-averaged rainfall data yielded better estimates of loads than point rainfall data for both models. Neither the deterministic model nor the statistical model (both using areal-averaged rainfall) was consistently better at estimating loads. Several factors could account for the inability of the models to estimate loads closer to computed loads. Chief among them is the fact that neither model was designed for the specific application of this study.
