Verification of FGGE Assimilations of the Tropical Wind Field: The Effect of Model and Data Bias

Abstract

We examine the tropical wind field analyses produced by a recent assimilation of the Final FGGE II-b dataset. Our aim is to estimate the effects, on the tropical wind analyses, of biases in the data and biases in the assimilation system. The assimilation system was similar to that used operationally at ECMWF in the first half of 1985. The period studied is the first Special Observing Period (SOP-1). Important differences occur in the intensity of divergence at upper and lower levels in the western Pacific, as measured by cloud-track winds (SATOBs) and by rawinsondes (TEMPs). There appear to be important biases also in the SATOB estimates of the zonal flow at upper and lower levels in the eastern Pacific. There are substantial biases in the wind directions at some west African stations. The 6-hour forecasts which provide the background fields for the analyses show important underestimates of the mean intensity of the tropical divergence field, particularly in the equatorial western Pacific. The errors in the background field probably occur because of underestimation of the intensity of tropical convection in the diabatic initialization and in the course of the forecast; the heavy spatial smoothing applied to the convective heating in the initialization probably also plays a role. Data were available in sufficient quantities that the analysis algorithm corrected the mean errors in the background field to a very large extent. As a result, any residual uncertainty in the mean analyses is within the uncertainty of the observations. The analysis algorithm has a rather poor response to divergent information even on large scales, so the analyzed divergence field agrees best with the observational data showing the weakest divergence, both in the upper and lower troposphere. The mean analyzed divergence field in the west Pacific agrees with the 850 mb TEMP data but is weaker than the intensity suggested by the low-level SATOBs and the SHIPs. In the upper troposphere the analyzed divergence is weaker than that suggested by the TEMPS, but agrees with that suggested by the (probably less reliable) SATOBs. Thus in this important area in the tropics the biases in the new analyses of the mean divergent wind field appear to be within the range of biases in the data, but the divergence is probably still underestimated in the upper troposphere and near the surface.