Note: HRES and Ensemble Control Forecast (ex-HRES) are scientifically, structurally and computationally identical.  With effect from Cy49r1, Ensemble Control Forecast (ex-HRES) output is equivalent to HRES output where shown in the diagrams.   At the time of the diagrams, HRES had resolution of 9km and ensemble members had a resolution of 18km.

Assimilation of 2m temperature

Screen level temperature (T2m) observations are assimilated into the model analyses.  Assimilation is done using 4D-Var as for observations of other variables.  

The process used is:

  • The observed screen level temperature is adjusted using a lapse rate of 5.5 K/km to take account of the height difference between station height and model height in 4D-Var.  This fits the data slightly better than the standard lapse rate of 6.5 K/km.
  • Only stations between 400 m below and 200 m above the model height are used.  The lower height limit is because, on average, stations are slightly lower than the model height as they are more likely to be in valleys.
  • No bias correction applied, because of the complexity of observation–background T2m biases, and in many cases the background biases are larger than observation biases. 
  • Large differences of the adjusted temperature from the background T2m temperature field are given a lower weighting.  Temperature differences of more than 7.5 K are not used. 
  • Observations are limited to the first six hours of the 12‑hour 4D‑Var window in order to produce more localised increments.

Screen level temperatures were not assimilated in earlier model cycles (Cy48 and earlier).


Assimilation of 2m specific humidity

Screen level specific humidity (q2m) observations, both day and night, are assimilated into the model analyses in a similar way to the assimilation of screen level temperature observations.


Forecast of 2m temperatures

Model output of forecast of screen level temperatures (T2m) is not a direct output from the atmospheric model.  

Instead, screen level temperatures (T2m) are derived by interpolation between: 

  • the model forecast temperature at the lowest model level (L137 at 10m) and 
  • the model forecast temperature of the underlying surface (the skin temperature).  This is determined using the land surface scheme HTESSEL, or the lake surface scheme FLake, or the sea-surface temperature (from NEMO). 

Stability in the lowest layers is taken into account using an interpolation function (α) derived using Monin-Obukhov similarity theory.  The stability measure is taken as the ratio of height above ground (z) to the Monin-Obukhov length (L).  The Monin-Obukhov length (L) is itself a function of, among other parameters, horizontal wind speed and upward ground heat flux.

  • With low stability, z is small compared with L.  The interpolation function (α) tends to 0 and T2m tends to the skin temperature. 
  • With high stability, z is large compared with L . The interpolation function (α) tends to 1 and the T2m tends to the T at the lowest model level.

In practice these extreme values of the interpolation function (α) are not realistic and the function that is used to interpolate between the temperature at 10m and the skin temperature is shown in Fig2.1.9.4-2.  This interpolation function gives rather better results than that used in earlier model cycles (Cy48 and earlier). 

  • in some winter regions 850 hPa temperatures are degraded. This is partly due to unrealistic coupling in stable conditions.


Forecast of 2m humidity

Model output of forecast of screen level humidity (q2m) is not a direct output from the atmospheric model.   It is interpolated in a similar way to screen level temperature.


Fig2.1.4.9-1:  Values at the 2m level (e.g. temperature) are not taken from a model level but are interpolated between model forecasts of temperature at the lowest atmospheric model level (level 137) and surface skin temperature.  The 2m level dew point is derived from the model forecast specific humidity interpolated in a similar way.  The nature of the interpolation profile used depends on other factors, such as stability and/or wind speed.


Fig9.1.4.9-2: The interpolation function (α) shown as a function of stability.    The stability measure is taken as the ratio of height above ground (z) to the Monin-Obukhov length (L).  The Monin-Obukhov length (L) is itself a function of, among other parameters, horizontal wind speed and upward ground heat flux.   α=1 implies that T2m equals the temperature at the lowest model level (TL137  at 10m); α=0 implies that T2m equals the surface (skin) temperature.  For practical purposes, the orange line shows the function that is used to interpolate between the temperature at 10m and the skin temperature.

Fig9.1.4.9-3: Illustration of temperature observations which are accepted for use in analysis of T2m by 4D-Var.  Temperatures are adjusted by 5 K/km from station height up or down to the model orography height.  Temperatures at stations >200m higher or >400m lower than model orography height are not used.  Stations on mountain tops and in deep valleys are thus excluded while retaining the majority of observations, including those in shallow valleys where many stations are located.

Assimilation of other surface variables

Winds from ships and moored buoys continue to be assimilated.  However, 10 m wind observations over land are not assimilated as it has proved difficult to get a positive impact on forecasts.

Surface pressure observations remain the most important surface variable for global NWP.


Considerations in using the forecast values

In land surface modelling (HTESSEL):

  • an urban tile models the fluxes of heat, moisture and momentum at the surface and allows a more realistic representation the heat island effects of towns and cities.
  • separation of tiles into high and low vegetation gives a more accurate seasonal variability of vegetation and incorporates the differing albedo effects of underlying snow cover.

In soil structure modelling:

  • forecasts of the relative availability of water in the soil for plant uptake allows plant roots to better extract water, especially in relatively dry conditions.  This can affect the surface specific humidity.

In the ensemble of data assimilations (EDA):

  • the background error estimate still appears too small in the lowest levels although stochastic physics in Cycle 49r1 have partially addressed this.  A larger spread of error estimates near the surface would tend to increase the size of analysis increments from T2m assimilation at the lowest model level and to reduce them a few levels up.  This would be more realistic for winter cases.

In general:

  • actual and model station altitude in mountainous areas may differ.
  • strong surface inversions, particularly over snow, may not be well modelled.
  • extent of low cloud cover may not be captured by the model.

Users should assess the potential for deficiencies in low-level parameters and adjust forecast values as necessary.


Additional sources of information

(Note: In older material there may be references to issues that have subsequently been addressed)


  • No labels