The present study describes the atmospheric component of the sixth-generation climate models of the Centre National de Recherches Météorologiques (CNRM), namely, ARPEGE-Climat 6.3. It builds up on more than a decade of model development and tuning efforts, which led to major updates of its moist physics. The vertical resolution has also been significantly increased, both in the boundary layer and in the stratosphere. ARPEGE-Climat 6.3 is now coupled to the new version (8.0) of the SURFace EXternalisée (SURFEX) surface model, in which several new features (e.g., floodplains, aquifers, and snow processes) improve the water cycle realism. The model calibration is discussed in depth. An amip-type experiment, in which the sea surface temperatures and sea ice concentrations are prescribed, and following the CMIP6 protocol, is extensively evaluated, in terms of climate mean state and variability. ARPEGE-Climat 6.3 is shown to improve over its previous version (5.1) by many climate features. Major improvements include the top-of-atmosphere and surface energy budgets in their various components (shortwave and longwave, total and clear sky), cloud cover, near-surface temperature, precipitation climatology and daily-mean distribution, and water discharges at the outlet of major rivers. In contrast, clouds over subtropical stratocumulus decks, several dynamical variables (sea level pressure, 500-hPa geopotential height), are still significantly biased. The tropical intraseasonal variability and diurnal cycle of precipitation, though improved, remained area of concerns for further model improvement. New biases also emerge, such as a lack of precipitation over several tropical continental areas. Within the CMIP6 context, ARPEGE-Climat 6.3 is the atmospheric component of CNRM-CM6-1 and CNRM-ESM2-1.Plain Language Summary Since the early 1990s, the Centre National de Recherches Météorologiques (CNRM) has been developing a global atmosphere model for climate applications. The present work presents its latest version, ARPEGE-Climat 6.3, as prepared for the sixth phase of the Coupled Model Intercomparison Project (CMIP6). It builds up on more than a decade of model development and tuning efforts. A CMIP6 amip-type numerical experiment, in which the sea surface temperatures and sea ice concentrations are prescribed, is evaluated, in terms of climate mean state and variability. ARPEGE-Climat 6.3 is shown to have better or similar skills compared to its previous version and to rank rather high among CMIP5 state-of-the-art models by many mean-state metrics. Major improvements include the top-ofatmosphere and surface energy budgets, cloud cover, near-surface temperature, precipitation climatology and daily-mean distribution, and water discharges at the outlet of major rivers. In contrast, clouds over the eastern part of ocean basins, and a few dynamical variables, such as sea level pressure, are still significantly biased. New biases also emerge, such as a lack of precipitation over several tropical continental areas. The remaining and n...

The impact of vertical resolution on numerical fog forecasting is studied in detail for a specific case and evaluated statistically over a winter season. Three vertical resolutions are tested with the kilometric-scale Applications of Research to Operations at Mesoscale (AROME) numerical weather prediction model over Paris Charles de Gaulle Airport (Paris-CDG) in Paris, France. For the case studied, the vertical resolution has a strong impact on fog onset. The nocturnal jet and the turbulence created by wind shear at the top of the nocturnal boundary layer are more pronounced with a finer vertical resolution, and the turbulence close to the ground is also stronger with high vertical resolution. Local circulations created by the terrain induce different simulated processes during the fog onset. The fog is simulated as advection–radiation fog in the finer vertical resolution run and as radiation fog in the others. The vertical resolution has little impact on the mature and dissipation phases. A statistical study over a winter season confirms the results obtained in the fog case study. High vertical resolution simulates earlier onset, as well as longer-lasting and more spatially heterogeneous fogs. The high vertical resolution configuration simulates more fog events than are found at low resolution (LR); these fog events generally form north of Paris-CDG. No observations are available in this area, leading to many simulated but no observed fog events in the fine-resolution runs. The ceiling of low clouds is not well simulated by the numerical model no matter what vertical resolution is used.