| dc.description.abstract | Open cell convection is a type of cellular convection with cloud-free interiors and cloud walls,
occurring more frequently in winter than other seasons. The cloud-free interior has a large
e!ect on the radiative balance, and together with the precipitation often occurring in the cell
walls, open cell convection can cause dangerous icy road conditions.
In most weather models convection has scales of motion that cannot all be resolved by the
model. The e!ects of convection are thus parametrized. Due to increasing model resolutions,
more and more convection can be explicitly resolved, and therefore parametrizations should not
parametrize all convection.
In cycle 46 of the HARMONIE-AROME numerical weather prediction model, open cell
convection is often ill represented and even missing.
Therefore, in this study we test an adjustment to the convection scheme of HARMONIE-
AROME, causing the model to explicitly resolve instead of parametrize moist updrafts for
clouds of less than 2 km thick in cold conditions. We test the adjustment in the winter period of
2024-2025, showing that the representation of showers and open cell convection is significantly
improved. Analyzing a case of open cell convection, the reference version of HARMONIE-
AROME transports too much moisture from the lifting condensation level further into the
atmosphere. This prevents the buildup of instability, precipitation and deeper convection. In
the adjustment, the instability builds up and leads to deeper convection, precipitation, and
open cell convection. A spectral analysis shows the adjustment more accurately reproduces the
typical scales found in satellite observations than the reference. In a verification comparing
station observations to forecasts over two two-weekly periods in the winter of 2020 and the
summer of 2018, we show that the forecast quality is not degraded in the adjustment.
These results suggest allowing the model to explicitly resolve convection in cold conditions
with clouds less than 2 km thick improves the forecast produced by the model. | |
