Comparisons of humidity observations and Eta model analyses and forecasts for water balance studies

Abstract

Direct comparison was performed between observed radiosonde humidity and vapor transport profiles and the National Centers for Environmental Prediction (NCEP) (formerly the National Meteorological Center) mesoscale Eta model analyses and forecasts. The data for this comparison were gathered during the Global Energy and Water Cycle Experiment's Continental‐Scale International Project Integrated Systems Test (GIST) for the period July 15 to August 31, 1994, which fell within the preoperational evaluation phase of the new NCEP mesoscale Eta model and associated data assimilation system. These intercomparisons show good agreement between Eta model analyses and observations in humidity and the zonal and meridional components of the humidity transport. Typical root‐mean‐square (RMS) errors relative to observed standard deviations were of the order of 0.4–0.5 at all levels between the surface and 300 hPa. The relative RMS humidity errors associated with the forecasts ranged between 0.9 and 1.4 in the lowest 300 hPa for all forecast times examined (3 to 36 hours). The RMS humidity transport errors for the forecasts tend to be somewhat less. During the GIST period, this study found the Eta model performance for water balance studies to be somewhat degraded over mountainous terrain. On the basis of this GIST finding and other corroborating evidence, NCEP corrected an error in the Eta model data assimilation system. The Eta‐model‐derived, vertically integrated moisture flux divergence compares favorably with similar estimates based on radiosonde observations alone. In these comparisons, the Eta model estimates of evaporation minus precipitation agree with the radiosonde estimates within 10% of the mean precipitation over the Mississippi River Basin. The agreement is less (around 50% of the mean precipitation) for a smaller basin. The analysis suggests that the Eta model forecasts may be useful over larger basins for various water budget study issues, such as diurnal variability and impacts of observation sampling density.