Effect of assimilating microwave imager observations in the presence of a model bias in marine stratocumulus

Decks of stratocumulus clouds over the easternmost parts of the maritime continent show a positive bias in the microwave brightness temperatures in the order of a Kelvin. We think the cause of this bias is an insufficient representation of drizzle and cloud liquid water path in the forecast model under stable conditions and an insufficient amplitude of cloud or drizzle diurnal variation in stratocumulus areas. This bias is not easy to correct with variational bias correction so biases in stratocumulus regions remain and are being assimilated, which led to concerns that these could have potentially affected the quality of the forecast. The most recent model cycle 43R1 of the Integrated Forecast System generates more stratocumulus clouds through a change in the entrainment in the shallow convection scheme, which corrects part of the bias. In this study, the estimated inversion strength is shown to be a good measure to identify and screen stratocumulus clouds before being assimilated. With such a screening the root-mean-square errors in relative humidity and temperature are reduced, which is caused by a change in the analysis and in the short-range forecast in very localised stratocumulus areas. Medium-range forecast scores are unaffected. Furthermore, fits to other humidity sensitive observations are slightly degraded by the screening. Hence, microwave observations in stratocumulus areas are probably useful for the assimilation system and should not be screened, despite the presence of biases.