What are GRIB files?

About this page

What GRIB files are and a word of warning.

Preamble

GRIB files provide a low cost way of getting a great deal of weather forecast information. There is much misunderstanding of GRIBs, some of it fostered by those selling GRIB services and products. I hope that this and the associated pages will help users to know what they are getting, how to assess what is available at no cost and to avoid paying unnecessarily .


On other pages of this site see


On this page


What they are.

They are data from Numerical Weather Prediction models. Values are on a 3-D grid of points. Those of most interest to sailors are wind, rainfall, swell although there are many others.


Where do they come drom?

All National Met Services (NMS) use GRIB files to store and exchange forecast data. However, most do not release them for general use. Exceptions are

  • US National Weather Service (NWS) – a part of NOAA,
  • US Navy
  • Canadian Met Service
  • Météo France

These all make data available to anyone with the computer resources to handle massive data sets. Most free sources of GRIB data for anywhere in the world use the US GFS (General Forecast System) with the others as back-ups.

To see what GRIB files do in practice, see the GRIB examples page. This show forecasts for days 1 to 7 days ahead.


Large scale forecasts

Forecasting for more than a day or so ahead requires global models partly because weather can travel a long way in 24 hours and, more importantly, because there are long distance connections; weather occurring in one area can affect what happens far away.

National Met Services usually use grid lengths (spacing) of about 13 – 17 km. The smallest feature that they can represent – geographical and meteorological – is about five times the grid spacing ie over 60 km or 30 miles.

NOTE Although the GFS uses a slightly larger grid of about 1/8 degree lat/lon, 13 km, the data are only made available by NOAA on a 1/4 degree grid ie around 15 mile spacing.

This has implications for those who use the degraded data as a basis for high resolution (Limited Area or Meso-scale) models without using the mass of fine cale data available- see below.


Getting the information

There are three main routes by which we can get global forecasts (GFS) at no cost, except for communications.

1. By email.

These are not the easiest to use but the most flexible. For those with limited bandwidth it may be the only option. Data are saved automatically.

2. By FTP (File Transfer Protocol)

  • ZyGrib A comprehensive free service with many data options. For Windows, Mac OS-X, Linux. Data are saved automatically.
  • Tablet Apps. There are many offering a range of options for both Apple and Android devices. Data are saved automatically.

3. By browser.

Weatheronline, PassageWeather, Windfinder, and some others are included in a listing of GRIB services on another page.

These are easy to use, provided you have the bandwidth. Disadvantages are that data are not saved and the choice of areas, data, etc is limited.

Some services that seem to offer more are

WindGuru. (free. Interpolates the GFS to specific locations).

MovingWeather (on repayment. Zooms in from the GFS to whatever area you ask for.

XCWeather (interpolates forecasts to observing site locations).

These simply interpolate between the values on the 1/4 degree grid. There are no, and can be no, differences in the quality of the product. They give a false impression of precision.

GRIB files received by email or FTP can usually be viewed using a variety of viewers. each has its pros and cons. See my GRIB viewer page for some alternatives.


Small scale forecasts – Meso-scale or Limited Area Models (LAMs)

Meso-scale models set out to predict on scales smaller than can be seen on synoptic charts. Grid lengths can be down to around 1 km. For several reasons – such as the short lifetimes of small scale weather, the general variability of weather – these are not very useful to most sailors. I never recommend paying for them.


For more on Numerical weather Prediction see my pages on NWP, the importance of Grid Length and [[Forecast-Accuracy- Limitations. ting GRIB services.


 

File Sizes

If you are using a satellite phone then files sizes can be critical. A file of GRIB data at a 1 degree latitude/longitude spacing covering the area 30N to 45N, 10W to 15E, giving wind arrows for eight times ie 12, 24, 36, 48, 60, 72, 96, 120 hours ahead (12 hourly to 3 days and 24 hourly to 5 days) would be 15x25x3x8 ie 9000 values.

The file containing these data is less than 15 kb. Output for half the area would be half as much ie less than 8 kb. If only winds had been requested, then file sizes would be 2/3 of these values. An example of one of the eight charts received in this way is shown here using the Viewfax software.

The isobars shown here are at a rather fierce 2 hPa (mb) spacing. There are options to change colours, scale and isobar spacing.


Cautionary NOTE -

From whatever source, it is important to remember that GRIB files are computer generated forecast files from a National Weather Service computer. These are sent without review, and are offered on an as-is basis. There is no assurance that the data are available, accurate or correct. Systems providing information and the computer models are automated and subject to a variety of failures and errors. By using the data, users acknowledge and agree to these limitations.

It should be remembered that human forecasters can still add value to computer model output. Consequently, the prudent Sailor will only use GRIB products in the short term (say up to 24 or 36 hours) in the light of other sources of information eg text forecasts, forecast charts generated by man-machine-mix, from the UK Met Office for example, or satellite pictures. For longer term predictions - say over 48 hours to seven or eight days, the basic computer output is unlikely to be improved upon by a human to any great. However, when preparing forecast charts at 4 and 5 days ahead. forecasters at the UK Met Office do look at other forecasts from other centres and try to produce a "best" forecast from a consensus point of view.


Home Page