Comparison of NCEP-NCAR Reanalysis with 1987 FIFE Data

Data from the FIFE experiment of the summer of 1987 are used to assess the diurnal and seasonal cycles of the surface energy budget and boundary layer in the NCEP-NCAR reanalysis, which used the summer 1995 version of the Medium-Range Forecast model. The seasonal agreement is quite good, reflecting the improvements in land-surface parameterizations in recent years. Detailed studies, however, identify several places where still further improvements in model parameterizations are possible. Clear-sky shortwave absorption and cloudiness may be underestimated in this model as has been noticed in other global models. More frequent updates of the model cloud cover (currently every 3 h) would improve the land-surface interaction after the initial onset of precipitation. The model produces a realistic well-mixed boundary layer, but underestimates boundary layer deepening by entrainment. For much of the summer, precipitation is close to that observed: however, for a period in June and early July, the reanalysis model has excess precipitation, which comes from daytime interactions between the surface evaporation, boundary layer, and convection schemes. The model, which has a deep 2-m soil reservoir, has adequate soil moisture storage for several weeks without rain, although surface evaporative fraction fluctuates more strongly after rain than is observed. Surface evaporation is generally too high at night, especially in high winds. This analysis suggests several improvements to the model parameterizations, some of which have already been implemented in the operational MRF model.