In this study, the forecast products of the National Meteorological Center's Eta model during the period August 1993 to March 1994 are critically examined to gain insight into the quality of the first guess fields used in the model's four-dimensional data assimilation and the usefulness of the forecast products in diagnostic studies involving unmeasured hydrological variables. We find that the Eta model 12--36 hour forecasts produce patterns of monthly precipitation that are a reasonably good approximation of the monthly observed precipitation fields for the period under consideration. The diurnal cycle of the water vapor fluxes and their seasonal changes obtained from the model forecasts are also in good agreement with observations. During the warm months the stationary component of the forecast fluxes includes a good representation of the low-level jet (LLJ) that is an important means for the transport of moisture from the Gulf of Mexico into the Great Plains. The LLJ decays during autumn, when it is replaced by the transient component of the fluxes as the means of transport of water vapor from the Gulf of Mexico region into the United States. Finally, evaporation estimates were derived from the convergence of the fluxes of the forecast fields and model and observed values of precipitation. All estimates agree within 0.5 mm d-1 and showed a reasonable seasonal cycle, with maximum evaporation during the warm season and a minimum evaporation during February. ¿ American Geophysical Union 1996