Assessment of the ECMWF Model Cloudiness and Surface Radiation Fields at the ARM SGP Site

The cloud and radiation fields produced by the operational ECMWF forecasts are assessed using observations from the Atmospheric Radiation Measurement Program (ARM) Southern Great Plains (SGP) site over the April-May 1999 period. Over the first 36 h of the forecasts, most of the model fields, taken over a 24-h time window (either 0-24, 6-30, or 12-36 h) are generally in good agreement with each other. Comparisons of model fields taken from any such 24-h time window with observations are therefore representative of the quality of the ECMWF model physical parameterizations. The surface radiation fluxes are assessed separately for clear-sky, overcast, and whole-sky situations. For clear-sky fluxes, differences between model and observations are linked to differences in humidity and temperature profiles, the characterization of aerosols, and potential problems in the radiation schemes. For clear-sky conditions, the downward longwave radiation is usually within the accuracy of the measurements. For overcast conditions, the agreement with observations is also usually good. On the other hand, the downward shortwave radiation is overestimated, whatever the conditions. Although this might be partly due to uncertainties in the aerosol content, the clear-sky overestimation of the downward shortwave radiation, when aerosols are specified from climatic values or observations, indicates an underestimation of the gaseous absorption. Model cloud occurrences and boundaries over the SGP Central Facility are compared with similar quantities derived from radar and micropulse lidar observations. Model cloud water is tentatively assessed through comparisons with the radar reflectivity measurements. Systematic deficiencies in the surface radiation fields in the presence of clouds are discussed with respect to differences between the model and observed cloud characteristics. Given the T L 319 resolution of the ECMWF model at the time of the comparisons, both the dayto-day and within-the-day temporal variability are captured reasonably well by 24-h forecasts that include cloudradiation interactions with 1-h time resolution. However, most of the differences with observations can be traced back either to deficiencies in the clear-sky shortwave radiation scheme or to problems in the cloud fraction and/ or cloud water content.