Rosen, R. D., D. A. Salstein, J. P. Peixoto, A. H. Oort, and N-C. Lau, 1985: Circulation statistics derived from level III-b and station-based analyses during FGGE. Monthly Weather Review, 113 (1), 65-88.

Abstract: A number of Northern Hemisphere circulation fields and statistics are derived for the months of January and June 1979 from level III-b analyses produced by GFDL using a 4-dimensional data assimilation scheme which incorporates measurements from a wide variety of sources. In particular, hemispheric maps and zonal cross sections of the wind, specific humidity, and the eddy fluxes of momentum, heat and moisture are examined. Certain quantities related to the atmosphere's energy cycle are also considered. These fields and statistics are compared with those derived from analyses that rely solely on the conventional rawinsonde station data taken during the same months. In the case of the monthly mean zonal and meridional winds, we also present results based on the level III-b analyses of the ECMWF. The station-based analyses yield zonal mean statistics and hemispheric integrals that are generally comparable to those from the level III-b analyses. For example, the intensity of the Northern Hemisphere Hadley cell in January produced by the station analyses lies between those of the III-b analyses, which differ by as much as 35%. On regional scales, however, there are some large differences in the circulation fields between the station-based and level III-b analyses over areas of sparse station coverage. For example, the station-based analysis of the 200 mb field of transient eddy momentum flux in January does not include a significant region of northward flux over the northeast Pacific that is contained in the GFDL analysis. It is not yet clear, though, to what extent model biases may be affecting the GFDL analysis in this or in other station-sparse areas. In the case of the subtropical Pacific jet in January, the station-based analysis appears to underestimate its extent, but there are also considerable differences between the two level III-b analyses in this region. In addition, the GFDL analyses often appear to be noisy. Improvements in the level III-b analyses need to be made before full confidence can be placed in results based on modern data assimilation techniques.