In 46R1 longwave scattering has been turned on, which leads to a slight warming of the surface and was found to lead to a modest reduction in the RMS error in tropospheric temperature forecasts by around 0.5%. A key innovation has been to represent longwave scattering by clouds but to neglect it for aerosols. This captures virtually all the benefits, and enables several optimizations to be performed achieving a neutral result in terms of computational cost. Furthermore, the 2D CAMS aerosol climatology has been replaced by a new 3D aerosol climatology. The latter better matches the original CAMS distribution and facilitates future updates of aerosol fields. The change has some positive impacts on lower tropospheric temperature and winds, especially along coastlines affected by seasonal biomass burning interacting with boundary layer clouds. Bigger positive impacts can be seen in the stratosphere where the RMSE of the temperature field in the 50-100 hPa layer near the summer pole decreases by 10% due to a similar reduction in the temperature bias.
The changes in the convection scheme include an increase in the entrainment of the test parcel, a positive definite correction for the denominator in the CAPE closure (improving the tangent-linear approximation) and for shallow convection a relative humidity dependent area fraction for evaporation (previously was a constant value).
A modification in the TL/AD of the semi-Lagrangian advection scheme results in improving the departure point calculation near the polar cap area. This was a long-standing problem, related with the way that inverse trigonometric functions were calculated in the non-linear and tangent-linear versions of the code. The latter was analytically but not discretely equivalent with the potential to give rise occasionally to an instability.
The changes introduced in the land surface scheme aim to minimize the occurrence of the maximum 2m temperature spikes - a side effect of an instability issue with the 2m temperature calculation. This was done by adjusting the wet tile skin conductivity. This modification partially solves the spike problem lowering the frequency of its occurrence by almost half. The increase in the skin conductivity for the interception tile (wet skin) is also justified from the physical point of view as a wet surface (basically thin water layer) should have higher conductivity than a dry one and the flux exchange should take the surface physical characteristics changes into account even for short living process. In addition a bit identical fix to the computation of the 2m temperature diagnostic for the wet skin and lakes tiles was introduced. All the performed experiments (forecast, 4D-Var and climate runs) suggested that the overall impact for such change is neutral to slightly positive.
Correctly accounting for the interception of rain has led to improved handling of episodically occurring snow events, since only a fraction of intercepted rain can refreeze. This has been introduced in 46R1 by correctly computing the amount of rain that could refreeze when intercepted by the snowpack. Previously, unphysical accumulation of snow in rainy conditions were locally observed during winter time.
A new wave physics parametrisation for wind input and open ocean dissipation was implemented in CY46R1. It is based on the work of Ardhuin et al. (2010) and the initial implementation of it into the Météo France version of the wave model code. It has been adapted and optimised to run efficiently with the latest version of the code. Because the wave model is coupled to the atmosphere, the new configuration was set up to yield similar level of feedback in the form of a sea state dependent Charnock coefficient. Note however, that the overall distribution of the Charnock parameter is bit tighter and yields slightly larger ocean surface roughness under typical tropical wind conditions. Impact on the ocean circulation, in the fully coupled system was also evaluated.
The main benefit of the changes is on the waves parameters, partly addressing the issue of over-prediction of long swell energy and the small under-estimation in the storm tracks.
Output wave parameters available from ECMWF forecasting system comprise set of parameters for the description of the mean sea state, as well as descriptions of the different major wind sea and swell components (IFS documentation part VII, chapter 10). There is also a set of variables for the description of the single largest waves in a record, which when substantially larger than the mean state are commonly referred to as freak waves. Based on new shallow-water parametrizations of envelope skewness and kurtosis of the sea surface elevation due to waves (Janssen 2017) and a new parametrization of dynamic kurtosis developed by \cite{JaJa2018}, the freak wave parameter calculation has been updated. The main impact is an enhanced probability of larger waves in shallow water with respect to the old version.
The calling frequency of the radiation scheme in ENS is changed from 3 hours to 1 hour, now consistent with HRES. The ENS now makes use of the 50 available EDA-members, the plus-minus symmetry of the ENS initial perturbations is removed and the ENS-members are now made exchangeable.