Comparison between the land surface response of the ECMWF model and the FIFE‐1987 data

Abstract An averaged time series for the surface data for the 15 × 15 km FIFE site was prepared for the summer of 1987. Comparisons with 48‐Hour forecasts from the ECMWF model for extended periods in July, August and October 1987 identified model errors in the incoming short‐wave radiation in clear skies, the ground heat flux, the formulation of surface evaporation, the soil‐moisture model, and entrainment at boundary‐layer top. The model clear‐sky short‐wave flux is too high at the surface by 5–10%. The ground heat flux is too large by a factor of 2 to 3 because of the large thermal capacity of the first soil layer (which is 7 cm thick), and a time truncation error. The surface evaporation was near zero in October 1987, rather than of order 70 W m −2 at noon. The surface evaporation falls too rapidly after rainfall, with a time‐scale of a few days rather than the 7‐10 days (or more) of the observation. On time‐scales of more than a few days the specified ‘climate layer’ soil moisture, rather than the storage of precipitation, has a large control on the evapotranspiration. The order 2g Kg‐1 in forecasts from an experimental analysis with nearly realistic surface fluxes; this because there is insufficient downward mixing of dry air.