Reducing the drying effect through a water vapour correction to the all-sky error model

In the past the assimilation of microwave imagers caused a small drying around 850 hPa, additional to a dry model climate. The additional drying effect has mostly been reduced due to the activation of 150/166 GHz channels at IFS cycle 46R1, and to a lesser extent due to changes in the humidity background error formulation at IFS cycle 43R3. This report investigates whether specific aspects of the observation error model can explain the remaining drying effect. First, using bias corrected values to compute the symmetric cloud amount C37 barely affect the observation error model. Second, the drying mainly occurs in almost cloud-free scenes, which led to a further hypothesis that the representation of water vapour in the observation error model in these cases could be related to the dry bias. Using reconstructed values for clear sky brightness temperatures at 37 GHz as part of the observation error model allowed a more symmetric treatment of how water vapour affects the clear sky polarisation difference at 37 GHz (P37clr) for observations and first guess. Combining both updates in the computation of C37 delivers the largest reduction in the additional drying effect associated with almost cloud-free conditions. However, the overall effect of these modifications of the observation error model is quite small, as seen in the neutral impact on forecast scores and neutral fits to independent observations. Having in mind the technical work required, we do not recommend the modifications in the computation of symmetric cloud amount for an operational model.