Technical Remarks

1. What do the time steps of forecast data refer to ?

The weather models of ERA40 and ERA Interim are restarted every couple of hours (tm = model start time). For all analysis and forecast data the time given in the grib file header is this model start. So the, e.g., step 06 forecast data (ts=6), stored with time 00h for a certain day, is the 6h-forecast (made by an initialization with the 00h values) valid for 06h UTC that day (= verification time = tv).

In ERA5 these things are different - see this chapter in the ERA5 Documentation at ECMWF: https://confluence.ecmwf.int/display/CKB/ERA5+data+documentation#ERA5datadocumentation-Meanratesandaccumulations

2. What about the accumulated values ?

Some parameters (e.g., precipitation) are only available as forecast data, not as analysis (sometimes called step 00 forecast). This mainly is accumulated data. They express the time integral of a certain physical quantity, starting from the start of the programme (ERA40 and ERA-Interim).

So in the step 18 forecast, a 12 o'clock value of precipitation contains the rainfall during 18 hours from 12:00 UTC to 06:00 UTC the next day (made by an initialization with the 12h values).

MIND: It is relevant that there are three time axes: tm = start time of model run; ts = forecast step with (after tm); tv = verification time = tm + ts. Grib headers of ERA40 and ERA Interim contain tm and ts. The tool wgrib lists this out. The cdo tools (cdo info) add tm and ts and simply give out tv (as of 2018-01). They do not give out the real file metadata content.

3. Gauss grids vs. spherical harmonics

During runtime the model switches between calculations on spherical harmonics (for the atmospheric dynamics) and calculations on Gauss grids. Within the ERA40 frame, the conversion method between both data types yields so called linear Gauss grids, whereas formerly quadratic grids were used with this weather model.

Gauss grids are called Nxx, where xx is the number of grid points between equator and pole. The high density of grid points close to the poles leads to some redundancy. So in addition to regular Gauss grids, sometimes so called "thinned" and "reduced" Gauss grids are used.

4. Quadratic vs. linear Gauss grids

The following table gives quantitative details on different Gauss grids:

resolution
(sph har)
pointsGauss gridapprox lon x lat
(deg)
T 42 128 x 64 N 32 ~ 2.810 x 2.810
TL 63 128 x 64 N 32 ~ 2.810 x 2.810
T 63 192 x 96 N 48 ~ 1.875 x 1.875
TL 95 192 x 96 N 48 ~ 1.875 x 1.875
T 106 320 x 160 N 80 ~ 1.125 x 1.125
TL 159 320 x 160 N 80 ~ 1.125 x 1.125
TL 319 640 x 320 N160 ~ 0.560 x 0.560
TL 511 1024 x 512 N256 ~ 0.350 x 0.350
TL 799 1600 x 800 N400 ~ 0.225 x 0.225
TL 1200 2408 x 1204 N602 ~ 0.150 x 0.150

Here T refers to the quadratic conversion scheme between Gauss and spherical harmonics, TL to the linear one (which was used for ERA40).

 

So a Gauss grid of N80 resolution is sometimes called T106 when it origins from a programme version that uses quadratic conversion between Gauss grid values and T106 resolution in the spherical harmonics.

On the other hand, a Gauss grid of N80 resolution is sometimes called TL159 when it origins from a programme version that uses linear conversion between Gauss grid values and T159 resolution in the spherical harmonics (as for ERA40 data).

 

5. Types of monthly means

As data is output every 6 hours, there are two types of monthly means datasets avalable:

  • Four values every month: means of the (about 30) 00, 06, 12, and 18 UTC data ("monthly separate means").
  • One value every month: means of the (about 120) values of this month ("monthly daily means").

 

Document Actions