Copied to clipboard

Notes on Projection


The UK model forecast data are being made available on a 'Latitude-Longitude' (Lat-Long) grid, as well as the standard Lambert Azimuthal Equal Area (LAEA) grid. This has the advantage the users are more familiar with Lat-Long than LAEA, but the re-gridding process effects the quality of the data. The information below is to ensure that users are aware of the effects of the re-gridding process, potential weakness or strengths with the re-gridded dataset, to help ensure that the data are being used appropriately.

General points


The application of re-gridding involves some form of interpolation which always reduces the information, or the quality of that information, in the dataset. As bi-linear (or a variant on this) is used for almost all parameters, the effect is to smooth the values, reducing the extremes.

This will have the most significant effects where there is a lot of spatial variation on a field. For example:

  • Depending on their scale, urban heat islands may not be so well captured. Although an attempt to preserve the temperature variations at the coast is made, using a modified form of bi-linear interpolation that does not spread temperature values form land to sea or visa versa;
  • It is less likely that extreme maximum or minimum temperatures are captured as well, or extreme precipitation rates.
    In some cases this may slightly improve some aspects of the forecast, where the smoothing reduces an unjustifiable level of detail. For example:
  • In a showery situation where the model is unlikely to get the exact position of the showers, the 'smearing' out of the precipitation, might provide slightly better guidance on the chance of getting wet, but not on the intensity of the precipitation (although this is quite a small effect).


A Lambert Azimuthal Equal Area (LAEA) grid was chosen as this is area-preserving - the size of the grid squares are the same over the whole domain. For a latitude-longitude (or Equirectangular) grid, although the latitudinal (north-south) grid point spacing remains constant, the longitudinal (east-west) grid point spacing varies significantly between the north and south of the grid; the spacing at the extreme south is nearly 1.5 times that of the spacing at the extreme north.

This can introduce inaccuracies in some calculations, such as calculating area averages.

Non-rectangular domain

The grid of data is rectangular in the LAEA projection, but the projection difference means that the Lat-long data do not appear to be rectangular. This means that:

  • If all the data are required, they will be returned as a large grid, with quite a bit of missing data in the east and west of the grid, and some in the north and south;
  • If a dataset with no missing data is required, only a much smaller cut-out of the full grid is available.

Choice of dataset

Given the impacts of re-gridding from LAEA to Lat-long, if the points raised above are a concern, it is worth using the standard LAEA data, if possible. This is particularly true if further processing (e.g. interpolation to a set of points) will be carried out.