Temperature,Relative Humidity,and Cloud Fraction Predicted by the NCEP Global Forecast System at the ARM SGP Site during 2001-2008:Comparison with ARM Observations
NCEP 全球预报系统在ARM SGP站点预报大气温度、湿度和云量的检验

This study evaluates the performance of the Global Forecast System(GFS) of the U.S.National Centers for Environmental Prediction(NCEP) against the Climate Modeling Best Estimate(CMBE) observational dataset made by the U.S.Department of Energy Atmospheric Radiation Measurement(ARM) Program at the southern Great Plains(SGP) site for the years of 2001-2008.The investigation focuses on the vertical distributions of air temperature(T),relative humidity(RH),and cloud fraction.The major findings are as follows:(1) NCEP GFS was able to largely capture the seasonal variations of T and RH.However,on seasonal average,the model overestimated T at the heights of 1.5-12 km,while underestimated T at 13-16 km in spring and winter and at 0-1.5 km in autumn and winter,by less than 1℃.Both the predicted and observed RH had double peaks located near the surface and around 12 km,respectively.However,the model overestimated RH in the upper and middle troposphere(4-12 km).Increase of model resolution from T170L42 to T254L64 significantly improved the prediction of RH at 14-18 km.(2) NCEP GFS generally underestimated cloud fraction at heights below 10 km and slightly overestimated cloud fraction at 11-13 km.Moreover,the prediction missed the daytime nonprecipitating low-level clouds and underestimated precipitating cloud amounts below 8 km,indicating that activities of shallow convection and deep convection in the model were not active enough.(3) Using the observed RH and the predicted cloud water/ice mixing ratio(qc) to calculate cloud fraction with the diagnostic method in the NCEP GFS model,the result shows that cloud fraction from this calculation is more significantly underestimated compared to the NCEP GFS predicted cloud fraction,suggesting that the underestimation of cloud cover at heights below 11 km by the NCEP GFS is probably contributed by an underestimate of qc at these altitudes.(4) Improvements in the prediction of T,RH,and cloud fraction were insignificant during 2001-2008.The inaccurate prediction of cloud fraction and qc is probably related to uncertainties of parameterizations of deep and shallow convection,as well as cloud microphysics,in the NCEP GFS model.