Evaluation of grid point data

For forecast ENS temperature data, all locations within each grid box surrounding a grid point are considered to have the same values as that forecast at the central grid point.  The fluxes of heat, moisture and momentum which in turn determine the surface values of temperature, dewpoint and wind at the grid point are calculated using the proportion of land within the surrounding area (where HTESSEL will be used) and lake/coastal seas (where FLake will be used).  For a sea grid point well offshore NEMO is be used to determine the surface fluxes of heat, moisture and momentum.

Energy flux information at each grid point is governed by the "fraction of land cover" assigned to the area surrounding it (see Fig8.1.4.1-1).   Thus grid points in rectangles that are coloured:

  • dark green are land points and HTESSEL will supply 90-100% of the flux information.
  • mid-green are land points (but with 10-20% water surface) so HTESSEL will supply 80-90% and FLake 10-20% of the flux information.
  • light green are land points (but with 20-30% water surface) so HTESSEL will supply 70-80% and FLake 20-30% of the flux information.
  • turquoise are land points (but with 30-40% water surface) so HTESSEL will supply 50-60% and FLake 30-40% of the flux information.
  • cyan are land points (but with 40-50% water surface) so HTESSEL will supply 50-60% and FLake 40-50% of the flux information.
  • blue are sea points (i.e. >50% water surface) so FLake will supply 100% of the flux information in coastal waters.  NEMO will supply 100% of the flux information in oceanic waters.

Users should note, for flux information: 

  • At coastal locations where there is less than 50% land cover in a grid box the water proportion is treated as a lake (using FLake) rather than as an ocean (which would use NEMO).  

  • Some water surfaces (e.g. The Great Lakes) are classed as lakes rather than sea and FLake is used exclusively.


Fig8.1.4.1-1: An example over southern England of "fraction of land cover" values showing the proportion of land and water within each 9km x 9km square centred on each grid point.  At grid point X the fluxes of heat, moisture and momentum  will be determined by 70%-80% by HTESSEL and 20%30% by FLake.  At grid point Y the fluxes of heat, moisture and momentum will be determined by 100% by FLake,  even though the grid point lies over land. 

Selection of ENS grid point relevant for a chosen location:

For land locations:

  • The nearest ENS grid point is selected from among the four ENS grid points surrounding the selected location.  Within these four grid points:
    • if there is at least one land grid point then the nearest land grid point is chosen (even though a sea grid point may be nearer).  A land grid point is one where the "fraction of land cover"  is greater than 50%.   
    • if there is no land grid point then the nearest ENS grid point is chosen (which will be a sea grid point).

For sea locations:

  • The nearest ENS grid point is selected from among the four ENS grid points surrounding the selected location.  

The process of selecting which gridpoints ENS that are used on meteograms is illustrated below, using relatively complex but informative examples.


Example of grid point data on meteograms

Fig8.1.4-2: 10-day medium-range meteogram for Oslo from HRES/Ensemble Control (blue line) and ENS members (box and whiskers) data time 00UTC 26 June 2023.   The map shows a close up of Oslo city.  The nearest land grid point to central Oslo is at 59.93N 10.83E which lies some 5km away from and some 141m higher than Oslo city centre.  This grid point may well be representative of Haugerud on the fringes of Oslo, but temperatures are reduced to near sea level using 6.5K/km lapse rate.

Examples of selection of grid point for meteograms

Example1: A medium sized island.

The Isle of Wight in southern England.   The island is approximately 40km long by 25km wide.  Coastal areas are strongly influenced by the sea while central parts are not.

Fig8.1.4.1-3: ENS grid points over part of southern England.  Rectangles surrounding each grid point are coloured according to the "fraction of land cover" assigned to each grid point and shown by the scale on the right.  Within each rectangle all locations are considered to have the same values.  The fluxes of heat, moisture and momentum which in turn determine the surface values of temperature, dewpoint and wind at the grid point are calculated using the proportion of land (where HTESSEL will be used) and lake/coastal seas (where FLake will be used for lakes or shallow coastal water), or NEMO alone for grid points over open sea. Towns mentioned below are Ventnor (V), Bembridge (B), Freshwater (F) and the city of Portsmouth (P) and locations are marked by a cross.

Execution of the technique 

Example sites are shown on the diagram:

  • An inland location - Newport (location N).  The ENS grid is scanned for the grid points surrounding the location (ENS grid points NPJR) and the nearest land point is chosen (Point J).  This is a land point where the "fraction of land cover" is 100% and the surface energy fluxes are determined by HTESSEL.
  • A coastal city location - Portsmouth (location P).   The ENS grid is scanned for the grid points surrounding the location (ENS grid points ABCD) and the nearest land point is chosen (Point A).  This is a land point where the "fraction of land cover" is 100% and the surface energy fluxes are determined by HTESSEL.  There will be no influence of a water surface.  HTESSEL does not take into account the urban nature of the city. 
  • A coastal location - Freshwater (location F).  The ENS grid is scanned for the grid points surrounding the location (ENS grid points MNRS) and the nearest land point is chosen (Point R).  This is a land point where the "fraction of land cover" is 60%-70%.  Surface energy fluxes are determined 60%-70% by HTESSEL and 30%-40% by FLake.
  • A coastal location - Bembridge (location B).  The ENS grid is scanned for the grid points surrounding the location (ENS grid points EFGH).  None is a land point and a sea point is chosen (Point E).  At this point the "fraction of land cover" is less than 50% and the surface energy fluxes are determined by FLake.  There will be no influence of land energy fluxes.  In fact any land location within grid box EFGH will be treated similarly, no matter how far away from the coast. 
  • A coastal location - Ventnor (location V).  The ENS grid is scanned for the grid points surrounding the location (ENS grid points JHLK) and the nearest land point is chosen (Point J).  This is a land point where the "fraction of land cover" is 100% and the surface energy fluxes are determined by HTESSEL.  No adjustment is made for the influence of the sea and the effect of the sea may not be evident on ENS meteograms.  This grid point is the same as selected for the inland town of Newport (location N) even though the town of Ventnor (location V) is right on the coast. 
  • A location near land - offshore of Ventnor (location S).  The ENS grid is scanned for the grid points surrounding the location (ENS grid points JHLK) and the nearest sea point is chosen (Point L).  The surface energy fluxes are determined by FLake.  The influence of the sea will be more evident on ENS meteograms for location S than at location V. 

Users should note:

  • Inspection of meteograms for nearby offshore locations (e.g. location S) may add useful information for nearby coastal locations (e.g. location V).
  • The same ENS grid point is used for both locations N and V (even though location N is inland and location V is coastal).  Differences between the inland location N and the coastal location V will not be apparent.  
  • Temperature is adjusted to reflect the differences in height between the altitude of each location and the corresponding ENS orography, using a lapse rate of 6.5K/km.

In the above example, if winds were light and from the East (i.e. wind blowing from sea to land at Ventnor) the influence of the sea point S is helpful in the derivation of temperatures.  However, if the winds were from the north (i.e. wind blowing from land to sea at Ventnor) then the influence of the sea point S may not be relevant.


Example2: A lake surrounded by rugged orography.

Eastern Lake Geneva.  Vevey and Montreux are lakeside towns which are not far apart but have different grid points; one grid point has an altitude near lake level, the other has an altitude associated with the nearby mountains.

Fig8.1.4.1-4: ENS grid points over Lake Geneva.  Rectangles surrounding each grid point are coloured according to the "fraction of land cover" assigned to each grid point and shown by the scale on the right.  Within each rectangle all locations are considered to have the same values.  The fluxes of heat, moisture and momentum which in turn determine the surface values of temperature, dewpoint and wind at the grid point are calculated using the proportion of land (where HTESSEL will be used) and lake (where FLake will be used). Towns mentioned below are Montreux (M) and Vevey (V); locations are marked by a cross.

Example sites are shown on the diagram:

  • A lake side location - Montreux (location M).  The ENS grid is scanned for the grid points surrounding the location (ENS grid points DEFG) and the nearest land point is chosen (Point D).  This is a land point where the "fraction of land cover" is 50%-60%.  Surface energy fluxes are determined 50%-60% by HTESSEL and 40%-50% by FLake.  The grid point D actually lies at Montreux itself.  However it has a model altitude of 801m while Montreaux has a geographical altitude of 582m. 
  • A lake side location - Vevey (location V).  The ENS grid is scanned for the grid points surrounding the location (ENS grid points ABCD) and the nearest land point is chosen (Point A).  This is a land point where the "fraction of land cover" is 60%-70%.  Surface energy fluxes are determined 60%-70% by HTESSEL and 30%-40% by FLake.  The grid point A has a model altitude of 693m while Vevey has a geographical altitude of 412m. 

The difference in geographical altitude reflects the hilly nature of land and towns near the lake.


Fig8.1.4.1-5: 10-day medium-range meteogram for Vevey (on the shores of Lake Geneva) from HRES/Ensemble Control (blue line) and ENS members (box and whiskers) data time 00UTC 26 June 2023.   The nearest land grid point to Vevey is at 46.50N 6.79E which lies some 5km away from and some 281m higher than Vevey city centre.  This grid point may well be representative of the mountains to the northeast of Vevey, but temperatures are reduced to Vevey level using 6.5K/km lapse rate.


Fig8.1.4.1-6: 10-day medium-range meteogram for Montreaux (on the shores of Lake Geneva) from HRES/Ensemble Control (blue line) and ENS members (box and whiskers) data time 00UTC 26 June 2023.   The nearest land grid point to Montreaux is at 46.43N 6.92E which is almost coincident with the city.  However, the model altitude is some 219m higher than Montreaux city centre.  Temperatures are reduced to Montreaux level using 6.5K/km lapse rate.

Because of the complexities of the orography around the location users should note: 

  • The area is very mountainous and different areas of each town have different altitudes.
  • The model orography, though fairly detailed, is somewhat smoothed and does not exactly follow true land altitude at each grid point.
  • Significant temperature adjustments are required from ENS grid point forecast values to better represent temperatures at the altitude of each town.  These adjustments make assumptions about the structure of the lower atmosphere.
  • Speeds cannot be relied upon as winds are strongly modified by orography and local effects.

Example3: A mountainous oceanic island.

Canary Islands

Fig8.1.4.1-7: ENS grid points around the Canary Islands.  Rectangles surrounding each grid point are coloured according to the "fraction of land cover" assigned to each grid point and shown by the scale on the right.  Within each rectangle all locations are considered to have the same values.  The fluxes of heat, moisture and momentum which in turn determine the surface values of temperature, dewpoint and wind at the grid point are calculated using the proportion of land (where HTESSEL will be used) and coastal water (where FLake will be used), or NEMO alone for grid points over open sea.  Locations mentioned below are St Cruz de Tenerife and Mount Tiede; locations are marked by a cross.


Example sites are shown on the diagram:

  • A coastal town - St Cruz de Tenerife.  The ENS grid is scanned for the grid points surrounding the location (ENS grid points ABCD) and the nearest land point is chosen (Point A).  This is a land point where the "fraction of land cover" is 90%-100% and the surface energy fluxes are determined by HTESSEL.  
  • A mountain location - Mount Tiede.  The ENS grid is scanned for the grid points surrounding the location (ENS grid points EFGH) and the nearest land point is chosen (Point F).  This is a land point where the "fraction of land cover" is 100% and the surface energy fluxes are determined by HTESSEL.  The grid point F actually lies at the peak of Mount Tiede itself.  However it has a model altitude of 1977m while Mount Tiede has a geographical altitude of 3385m.  

There are wide variations in orography within the islands (the islands are quite mountainous) and the representativeness of a grid point can be uncertain.


Fig8.1.4.1-8: 10-day medium-range meteogram for Santa Cruz de Tenerife from HRES/Ensemble Control (blue line) and ENS members (box and whiskers) data time 00UTC 26 June 2023.   The nearest land grid point to Santa Cruz is at 28.51N 16.28W which lies some 5km away from and some 173m higher than Santa Cruz.  This grid point may well be representative of the hills to the northeast of Santa Cruz, but temperatures are reduced to Santa Cruz level using 6.5K/km lapse rate.

Fig8.1.4.1-9: 10-day medium-range meteogram for Mount Tiede from HRES/Ensemble Control (blue line) and ENS members (box and whiskers) data time 00UTC 26 June 2023.   The nearest land grid point to Mount Tiede is at 28.30N 16.63W which is almost coincident with the mountain peak.   However, the model altitude is some 1408m lower than the height of the mountain.  Temperatures are corrected to mountain peak level using 6.5K/km lapse rate.


There are wide variations in orography within the islands (the islands are quite mountainous) and the representativeness of a grid point can be uncertain.  Local uncertainty in forecast temperatures at altitude can have a large impact of model precipitation especially over mountainous islands and coasts.  


Example4: Isolated small islands.

Isole Eolie.  A set of small volcanic islands near southwest Italy.  The islands are roughly 5km x 5km or smaller.

Fig8.1.4.1-10: ENS grid points around southwest Italy.  Rectangles surrounding each grid point are coloured according to the "fraction of land cover" assigned to each grid point and shown by the scale on the right.  Within each rectangle all locations are considered to have the same values.  The fluxes of heat, moisture and momentum which in turn determine the surface values of temperature, dewpoint and wind at the grid point are calculated using the proportion of land (where HTESSEL will be used) and coastal water (where FLake will be used), or NEMO alone for grid points over open sea.  Locations mentioned below are marked on the diagram.

The grid points either touch the islands but with less than 50% land cover, or miss the islands completely.  All fluxes of heat, moisture and momentum are derived using  FLake.


Fig8.1.4.1-11: 10-day medium-range meteogram for the town of Malfa on Malfa Island from HRES/Ensemble Control (blue line) and ENS members (box and whiskers) data time 00UTC 26 June 2023.   The ENS grid is scanned for the grid points surrounding the location.  None is a land point and nearest sea point is chosen.  This point is actually situated on land but the "fraction of land cover" is less than 50% and the surface energy fluxes are determined by FLake.  There will be no influence of land energy fluxes.  In fact the whole island including the mountains will be treated similarly, no matter how far away from the coast.  This grid point may well be representative of the southwest coast of the island.  However, local effects may be important on other coasts (e.g. sea breezes).  Conditions at inland high ground will not be reliably indicated, particularly for Monte dei Porri which rises to 860m.   



Fig8.1.4.1-12: 10-day medium-range meteogram for the town of Stromboli on Stromboli Island from HRES/Ensemble Control (blue line) and ENS members (box and whiskers) data time 00UTC 26 June 2023.   The ENS grid is scanned for the grid points surrounding the location.  None is a land point and nearest sea point is chosen.  There will be no influence of land energy fluxes.  In fact the whole island including the mountains will be treated similarly, no matter how far away from the coast.  Local effects may be important (e.g. sea breezes).  Conditions at inland high ground will not be reliably indicated.   


Fig8.1.4.1-13: 10-day medium-range meteogram for the Stromboli volcano on Stromboli Island from HRES/Ensemble Control (blue line) and ENS members (box and whiskers) data time 00UTC 26 June 2023.   The ENS grid is scanned for the grid points surrounding the location.  None is a land point and nearest sea point is chosen.  There will be no influence of land energy fluxes.  In fact the whole island including the mountains will be treated similarly, no matter how far away from the coast.  Conditions at inland high ground will not be reliably indicated.  Note the temperature data at the sea grid point (model height -8m due to the spectral representation of altitude) is amended to that at 422m (the model height at Stromboli volcano) which is itself less than the true geographic height of 926m.    

There are wide variations in orography within the islands (the islands are quite mountainous).   Grid points are almost exclusively over the sea so land effects will not be taken into account.  The representativeness of a grid point can be very uncertain though may be appropriate for coastal parts.  Inland parts of small islands will be largely similar to the coasts but nevertheless there is likely to be large local variations in conditions.  Local effects can be very important with local sea breezes, nocturnal breezes, shelter, etc.  Many small islands are mountainous - Malfa rises to 860m and the active volcano on Stromboli rises to 926m.    The effects of volcanic activity are not dealt with by IFS).  

It is for the user to make adjustments to meteogram values, particularly temperature.

Model representation of orography

Modelling the surface orography at an appropriate resolution is crucial to an effective forecast.  However, at some level, there always will be smoothing that misses important detail.   

Fig8.1.4.1-14: Schematic of the spectral representation of orography.   Model orography matches true orography over large parts of the earth but is less exact in rugged mountainous regions.  See also Section on Model Orography for further points regarding orography.


Generally model orography matches true orography over large parts of the earth.  However, the spectral representation of orography in the IFS can:

  • smooth true orography, particularly in rugged mountainous areas where there are large variations in altitude over short distances.  Mountain peaks may be under-represented and narrow valleys may not be represented at all.
  • local effects can be under-identified where there are small scale variations in true or model orography, even where relatively low in altitude. 
  • lead to "topographic ripples" over adjacent sea/large lakes, which decay with offshore distance, and which are most prominent where there are steep-sided high mountains nearby.  It is quite possible that the model sea surface level is negative!



Considerations when viewing meteograms

The method of assessment and delivery of data for presentation on meteograms has been described in detail to give an understanding of the techniques involved.  

IFS uses a spectral representation of orography and so there is some smoothing, particularly in mountainous areas.  This means that will have model station heights that are different from the  geographic height.  For the majority of locations the differences are relatively minor.  But there can be a significant difference at locations where there are large variations in geographic heights over a relatively small distance (e.g. deep valleys in rugged terrain, isolated steep islands, or coastal towns adjacent to mountainous regions).

Note: The station height on the meteogram is defined for:

  • named locations, towns, cities as the value of the model altitude at that point.
  • locations defined by latitude and longitude as the geographic altitude at that point.

Users should use meteogram output with caution - the data should not be taken as definitive but should be assessed and possibly adjusted.   ENS forecast values should not be taken at face value but there should always be consideration of the ways that temperature and other values are derived.  The effects of local influences are most important.  Disentangling coastal effects from altitude effects can be difficult. 

In particular users should:

  • critically assess forecast values in the light of experience regarding differences between previous forecast values and actual observed observations.
  • consider the representativeness of the meteogram in coastal, island or mountainous regions and take into account consequent differences in height between altitude of the grid point and that of the desired town or location.
  • consider the variation in temperature (and possibly precipitation) that might be expected in different parts of the town or city.  Some cities spread from sea level to a few hundred metres in altitude.
  • consider the structure of the lower atmosphere as IFS temperature adjustments make assumptions of a uniform lapse rate (6.5K/km).  In the case of temperature inversions the forecast of 2m temperature needs to be used with great care; in such situations, depending on the inversion level, the standard lapse rate assumptions can be very inappropriate.
  • consider meteograms for nearby offshore locations which can add useful information for adjacent coastal locations.
  • consider if the same ENS grid point has been selected by IFS for both inland and coastal locations.  Meteograms may not indicate correctly the differences between the locations.  
  • note influences of any adjacent sea areas on coastal areas may be over- or under-represented by the ENS meteograms.  
  • assess the effect of the forecast winds (e.g. if the wind blows from land to sea then the influence of a nearby sea point may not be relevant).
  • assess whether a local effect might be relevant (e.g. onset of a sea breeze), or the local prevalence of persistent cloud (e.g.sea fog and low cloud drifting onshore), or the influence of turbulent mixing with stronger winds. 
  • consider the land-sea mask value(s) at the grid point and the consequent impact of heat, moisture and momentum fluxes by HTESSEL and FLake on the forecast parameters.
  • note some areas well inland from coasts can be governed by fluxes derived using FLake.  Sea grid point (defined as a grid point surrounded by >50% water surface) can be over land and have an altitude defined by ENS orography.  ENS sea grid points do not always have an altitude of 0m.
  • note the spectral representation of orography in the IFS, can:
    • smooth model orography and local effects can be under-identified.
    • lead to "topographic ripples" over adjacent sea/large lakes, which decay with offshore distance, and which are most prominent where there are steep-sided high mountains nearby.
  • note wind speeds cannot be relied upon in mountainous areas as winds are strongly modified by orography and local effects.


  It is for the user to assess critically the representativeness of the meteogram displayed and to make adjustments in the light of local knowledge and experience.