Assessment of the land surface and boundary layer models in two operational versions of the NCEP Eta Model using FIFE data

Data from the 1987 summer FIFE experiment for four pairs of days are compared with corresponding 48·h forecasts from two different versions of the Eta Model, both initialized from the NCEP-NCAR (National Centers for Environmental Prediction-National Center for Atmospheric Research) global reanalysis. One used the late 1995 operational Eta Model physics, the second, with a new soil and land surface scheme and revisions to the surface layer and boundary layer schemes, used the physics package that became operational on 31 January 1996. Improvements in the land surface parameterization and its interaction with the atmosphere are one key to improved summer precipitation forecasts. The new soil thermal model is an improvement over the earlier slab soil model, although the new skin temperature generally now has too large a diurnal cycle (whereas the old surface temperature had too small a diurnal cycle) and is more sensitive to net radiation errors. The nighttime temperature minima are often too low, because of a model underestimate of the downwelling radiation, despite improvements in the coupling of the surface and boundary layer at night. The daytime incoming solar radiation has a substantial high bias in both models, because of some coding errors (which have now been corrected), insufficient atmospheric shortwave absorption, and underestimates of cloud. The authors explore evaporation before and after a midsummer heavy rain event with the two models. The late 1995 operational model uses a soil moisture bucket physics, with a specified annual-mean fixed field soil moisture climatology, to the surface evaporation responds primarily to the atmospheric forcing. While the surface fluxes in the new model show this strong rain event more dramatically, because its soil moisture comes from the global reanalysis rather than climatology, there remain problems with soil moisture initialization. It appears that a fully continuous Eta data assimilation system (wihch is under develo