Improving supercooled liquid water representation in LIMA using ICICLE data

Abstract A realistic supercooled liquid water (SLW) representation in numerical weather prediction models is necessary to accurately simulate clouds, their effects on precipitation and radiative balance, and to improve atmospheric icing forecast. Our objectives are to assess the SLW representation in the Liquid Ice Multiple Aerosol (LIMA) double‐moment microphysics scheme, to identify the physical processes responsible for the main biases, and to suggest some improvements. To this end, numerical simulations of the Meso‐NH research model were compared with observations from the In‐Cloud Icing Large‐Drop Experiment airborne measurement campaign. A total of 20 days, including 25 research flights of the campaign, were simulated at a 2 km horizontal resolution. A comprehensive evaluation of the simulations was performed, both statistically and flight‐by‐flight. Several classical freezing rain and lake effect situations were sampled, allowing a more specific evaluation of these situations. A detailed analysis of microphysics process budgets was then conducted on one freezing drizzle case in order to suggest an improved SLW modelling. LIMA produced SLW at temperatures as cold as . However, SLW contents were underestimated, especially at temperatures between 10 and . Furthermore, the SLW was modelled in mixed clouds, containing both liquid water and ice, whereas purely supercooled liquid clouds were observed. The budget analysis showed that supercooled freezing rain was overconsumed by riming. A delayed ice nucleation and a snowflake size distribution depending on the temperature diminished the riming of snow, but the SLW was still consumed by graupel riming. Therefore, we allowed the riming of graupel only for graupel mean volume diameters larger than 125 m, which enabled the freezing rain to reach the ground. Those modifications were tested over the entire campaign and improved the results statistically.