The performance of four popular surface closure schemes commonly used in limited area and large-scale numerical models are evaluated using data obtained from the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE) which took place in central Kansas during 1987--1989 and the Hydrologic Atmospheric Pilot Experiment in the Sahel (HAPEX-Sahel) experiment which took place in Niger, West Africa, in 1992. The analysis uses the 21-day FIFE data set from the 1989 IFC 5 intensive field campaign (July 23 to August 12) and golden day 283 (October 9) from the HAPEX-Sahel intensive operational period. The surface heat flux closure schemes are based on (1) the bulk aerodynamic method of Laval et al. (1981), (2) the parametric bulk Richardson number scheme of Louis (1979), (3) the modified parametric Louis scheme of Louis et al. (1981), and (4) the level 2 closure scheme of Mellor and Yamada (1982). The analysis of surface flux RMS error statistics over the entire multistation, half hourly, 21-day FIFE record indicates that all four schemes exhibit significant RMS errors, particularly for unstable conditions, although the day-to-day average RMS errors fluctuate considerably. The biases and histograms of RMS error distribution are examined in the context of the parameterization formulations, in order to identify specific weaknesses in the assumptions used in the closure schemes.

In the composite diurnal cycle framework for unstable conditions in the FIFE analysis, large RMS errors exceeding 120 W m-2 are found for all four schemes during midday, with daytime fluxes always overestimated. However, when the schemes are tested on an individual day characterized by light wind, dry convection conditions from FIFE, the four schemes continue to overestimate fluxes but with considerably smaller RMS error. This result is corroborated with a similar case from HAPEX-Sahel. The smaller errors for these cases are consistent with the idealized conditions under which the schemes are calibrated. Recalibration of the original bulk Richardson number closure scheme, based on the FIFE data, leads to a 40% RMS error improvement, suggesting there are alternate approaches that could be used to optimally calibrate the schemes. However, the recalibrated scheme has not been tested independently and thus should not be applied beyond a tallgrass prairie setting. The study concludes with explanations as to why the original calibrations would be expected to overestimate fluxes under unstable conditions given that meteorological conditions generally deviate from the idealized boundary layer situation. ¿ American Geophysical Union 1995