Assimilating radiosonde relative humidity data directly into the Met Office operational forecast models can result in spurious depletion of analysed cloud cover. For this reason, a bias correction, derived statistically in 1994, is applied to these measurements prior to their assimilation. Recent work has focused on developing a real-time correction scheme to assess individually the biases for each radiosonde ascent.

In essence, the bias correction scheme compares surface reports of cloud cover with the total cloud implied by the radiosonde, which is calculated using the model cloud scheme relationship between humidity and cloud followed by an overlap assumption. The methods of deriving either a statistical or real-time correction from this comparison will be outlined. Since the model's humidity-cloud relationship is not constant for ice cloud, a 'calibration' of the relationship used in the bias correction is attempted from a local assessment of the model background field. The bias correction is tested and tuned in assimilation and forecast experiments that focus on verification of cloud amount, screen temperature, precipitation and visibility.

Pre-operational trial results will be presented.