Background error correlation between surface skin and air temperatures: Estimation and impact on the assimilation of infrared window radiances

This paper makes use of ensemble forecasts to infer the correlation between surface skin temperature T-s and air temperature T-a model errors. The impact of this correlation in data assimilation is then investigated. In the process of assimilating radiances that are sensitive to the surface skin temperature, the T-s - T-a error correlation becomes important because it allows statistically optimal corrections to the background temperature profile in the boundary layer. In converse, through this correlation, surface air temperature data can substantially influence the analysis of skin temperature. One difficulty is that the T-s - T-a correlation depends on the local static stability conditions that link the two variables. Therefore, a correlation estimate based on spatial or temporal averages is not appropriate. Ensembles of forecasts valid at the analysis time provide a novel means to infer the correlation dynamically at each model grid point. Geostationary Operational Environmental Satellite ( GOES)- 8 and - 10 surface-sensitive imager radiances are assimilated with and without the inferred correlations in a 3D variational analysis system. The impact of the correlation on analyses is assessed using independent radiosonde data. The impact on 6-h forecasts is also evaluated using surface synoptic reports. The influence of the correlation extends from the surface to about 1.5 km. Temperature differences in the resulting analyses on the order of 0.3 - 0.6 K are typical in the boundary layer and may extend over broad regions. These difference patterns persist beyond 6 h into the forecasts.