The IFS atmospheric model has many levels in the lower atmosphere to capture the all important boundary layer but precision is difficult.  This is because there are difficulties modelling the detail of radiation exchanges at the surface the lack of uniform and widespread observations.  This affects the development and persistence of cloud, and hence also the albedo and radiative balance between surface and boundary layer air.  

Cloud containing super-cooled liquid water (SLW) is frequently observed by aircraft and by remote sensing.  But the processes and consequent effects associated with super-cooled liquid water in the cloud are difficult to model precisely because:

• super-cooled liquid water is important for radiation considerations.
• super-cooled liquid water can increase cloud lifetime (liquid drops can remain suspended while ice crystals grow and fall out).
• there is a fine balance between turbulent production of water droplets, nucleation of ice, deposition growth and fallout.
• there are uncertainties in turbulent mixing, ice microphysics, vertical resolution.

The structure of the boundary layer is crucial.  This is particularly so where there is a well-marked inversion, with or without a sharp change in humidity.  Users should note:

• stratocumulus tends to be under-predicted over land in anti-cyclones or may dissipate too quickly.

• incorrect definition of the boundary layer in the physics schemes can mean incorrect identification, formation or dispersal of low cloud.
• observed, analysed and forecast temperatures can be very different to one another in hilly or mountainous regions.
  

Problems in handling low cloud can have a significant impact on the temperature and moisture structure of the boundary layer and impact 2 m temperatures.  A revised warm-phase microphysics and revised boundary layer clouds and shallow convection were introduced in 2018.

The user should assess carefully the model representation of temperature and moisture structure in the lower atmosphere.

Fig9.1.1: A comparison of observed (orange) and model analysed/forecast (green) temperature and dewpoint structures.  Errors are due to assimilation issues coupled with the difficulties handling the cloud physics.  In this case the surface cool and moist layer was analysed to be slightly deeper than in reality.   This retarded fog clearance and therefore delayed heating and overturning of the boundary layer through the morning.  So by 12UTC the forecast inversion was too low compared with reality and it had also not captured the stratocumulus from the convective overturning within the boundary layer.  Consequently the true radiation balance around midday was not captured.

Fig9.1.2: Examples of the difficulty of describing the boundary layer temperature and moisture structure.  Dew-point is used here as a measure of moisture.  Values as analysed or forecast shown in blue; observed values shown in orange.

At Stuttgart, Budapest and Nis the lowest ~500 m is poorly represented.  Solar radiation in the atmospheric model is depleted/enhanced by the presence/absence of fog or low stratus cloud.  At Bucaresti the temperature structure of the lowest layers is modelled quite well, but the moisture structure near the subsidence inversion is not.  Differences in boundary layer moisture have a strong influence on the development of stratocumulus as surface temperatures rise during the morning.   

Fig9.1.3:  Comparison of vertical profiles at Nottingham, England at 05UTC 13 Nov 2022, as observed (black) and HRES T+17 data time 12UTC 12 UTC 12 Nov 2022 (red).  This is fairly typical vertical profiles for anticyclones with inversions in the lowest part of the troposphere. HRES does show inversion but lacks detail ands didn't manage to form low clouds/fog.  It is apparent how sensitive the cloud formation is to subtle changes in the environment near the surface.

Fig9.1.4: Comparison of vertical profiles at Bucuresti, Romania at 00UTC 16 Nov 2022, as observed (black), First Guess (blue) and analysed (red).  The observed profile shows boundary layer cloud between 980 hPa and 940 hPa beneath a strong inversion associated with an anticyclone. The cloud is absent on the first guess profile and only just identified on the analysis as a thin layer under the inversion.

Fig9.1.5: Comparison of vertical profiles at Bucuresti, Romania at 12UTC 16 Nov 2022, as observed (black) and as forecast (red).  The observed profile shows the strong inversion has descended and the boundary layer cloud has persisted between 970 hPa and 960 hPa. HRES has not captured the descent of the inversion so well.  Also the analysed moisture structure of the boundary layer at 00UTC was too dry and thus little or no cloud has been forecast.  This resulted in forecast 2 m temperatures at 12 UTC being far too high over much of Romania.