The application of a nonlinear least-squares regression model describing the sources and transport of dissolved solids in streams of the Upper Colorado River Basin, and that was calibrated using data from water year 1991, was evaluated for use in predicting annual dissolved-solids loads for the years 1974 through 1998. Simulations for each water year were run using annual climate data. To evaluate how well the model captures the observed annual variability across the basin, differences in predicted annual dissolved-solids loads for each simulated year and 1991 were compared with differences in monitored annual loads. The temporal trend of the differences between predicted annual loads for the simulated years and the load for 1991 generally followed the trend of the monitored loads. The model appears to underpredict the largest annual loads and overpredict some of the smaller annual loads. An underprediction bias for wetter years was evident in the residuals as was an overprediction bias, to a lesser degree, for drier years. A regression analysis on the residuals suggests that the underprediction bias is associated with precipitation differences from 1991 and with previously defined downward trends in dissolved-solids concentrations in the basin. In general, given the representative climatic conditions, the model adequately performs throughout the period examined. However, the model is most transferable to years with climatic conditions similar to 1991.
