For a personal project, i wanted to develop a global atmospheric database, to compile data like wind speed and direction, temperature, etc. from different sources.

My question is this : Does such a database exist? either in a free or paying format.

  • 1
    $\begingroup$ Do model reanalyses count? $\endgroup$ Commented Jul 24, 2017 at 21:03
  • 1
    $\begingroup$ @BarocliniCplusplus I'm looking for a compilation of original data, but i'm curious anyway. $\endgroup$
    – Sclrx
    Commented Jul 24, 2017 at 22:25
  • 3
    $\begingroup$ opendata.stackexchange.com/questions/10154/… is our "canonical" answer to this question. $\endgroup$
    – user967
    Commented Jul 25, 2017 at 2:02
  • 1
    $\begingroup$ @Sclrx What do you mean by "original data"? What are you trying to do? Reanalysis uses petabytes of "original data". Can you please clarify how reanalysis does not satisfy your need? $\endgroup$
    – gerrit
    Commented Jul 25, 2017 at 8:43
  • 1
    $\begingroup$ @userLTK He may not there are standardized networks worked by single groups. But either way, in the US we have ASOS stations, co-op stations, mesonet stations, home weather stations, etc. They aren't typically compiled together anywhere, so I'd think that's what he means. ASOS are the standard of typical official reports, but other stations can still be useful as well $\endgroup$ Commented Jul 25, 2017 at 20:25

2 Answers 2


By your edit, it seems you are more interested in the "big data" side of the science. Model reanalyses contain large amounts of historical weather data, so that is a safe bet for ample amounts of data. NOMADS, which is a service provided by the National Weather Service is a good resource for model data. There is another version of NOMADS for operational data; I'd recommend that. For generic data, RDA has data. NASA has a server that also has a lot of data.

You can also generate your own weather data for prediction and the like, by using models like WRF or SPEEDY.

Departing from model data, you have observations, you can access ground-based observational datasets via NCDC. NCDC also has satellite data, radar data, and so on down the line (you don't need me to explore the site for you).

Other countries have other datasets, though many come with a price. ECMWF, for example, has reanalysis products, operational datasets, etc. The same is true for JMA, and other countries with their own weather service.


Wind speed, direction, temperature (highs and lows only or time of day?)

And do you want this for all over the Earth or just regional, like the continental US?

If this answer is correct, there are roughly 30,000 weather stations across the globe, 365 days a year that's about 11 million rows of data per year. That's a large enough database where you'd need special software to handle it.

And if you want hourly wind and weather (1 wind speed and direction per day is pretty limited), now you're over 250 million rows of data per year, unless you want to limit your database to highs and lows - which might be the way to go.

And if you want this going back 100 years, well, that's a monster amount of data. There's also the problem that weather stations don't always stay in the same place. In the 1990s they opened up a whole lot of new weather stations, a percentage of which have since been closed. That makes for some fun reading, especially on the Fox-News side.

This might be the best source, at least, the best I found doing a quick search.


and daily: https://www.ncdc.noaa.gov/cdo-web/datatools/records

They have a search function and a variety of datasets. I haven't looked at them in detail.

This might also be of interest: https://en.wikipedia.org/wiki/Citizen_Weather_Observer_Program

I also just noticed, most of this info is in the source data question posted in comments. (oops).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.