Assimilation of Raman lidar profiles in an operational, convective‐scale numerical weather prediction model

Observing systems for numerical weather prediction lack observations in the planetary boundary layer, even though boundary‐layer conditions are important for the prediction of high‐impact weather like convective storms or fog and low stratus. Raman Lidar for Meteorological Observations (RALMO), a Raman lidar remote‐sensing system located in Payerne, Switzerland, provides profiles of temperature and water‐vapour mixing ratio with a high vertical and temporal resolution. To assess the impact of RALMO in a weather model, we include its observations in the MeteoSwiss operational, convective‐scale ensemble data assimilation and forecasting system in two‐week summer and winter experiments. In the assimilation process, we make use of the state‐dependent observation‐error estimates provided by the RALMO instrument. A verification with independent observations from a collocated, ground‐based microwave radiometer shows that the analysis‐departure standard deviation of the brightness temperatures in the humidity‐sensitive channels is reduced by 20%–40% when assimilating RALMO profiles in addition to the operational observations. In winter, the improvement is more long‐lived than in summer: after a one‐hour forecast, the background‐departure standard deviation is still about 20% lower than the analysis‐departure standard deviation of the reference experiment. In addition, we investigate the impact of assimilating RALMO instead of the collocated radiosonde observations and find that the humidity analysis is closer to the microwave radiometer observations than that of the reference assimilating the radiosonde observations. Our results demonstrate the potential for profiling instruments to add to analysis quality, especially in regions without other profile observations. A detailed verification of forecasts using RALMO and radiosonde profiles, surface stations, radar/gauge precipitation estimates, and satellite‐derived cloud estimates revealed that the impact of assimilating RALMO observations on forecasts is small. This is assumed to be mainly due to the fact that only one single Raman lidar station was available for assimilation.