Met Office Hadley Centre observations datasets
Home > HadISD > >

HadISD - global sub-daily station dataset - version

Note for version and later: The temperatures recorded in the recent western Canada & US heatwave in late June 2021 were sufficiently extreme and beyond climatology that a number of these have been flagged by our automated QC tests. We have updated some of the QC tests as detailed in our blog posts (1, 2, 3, 4). Therefore the minor version number has incremented by one (from 3.1.2 to 3.2.0). If you spot anything untoward as a result of these changes please do let us know. (07-Oct-2021)

Note for version and later: An issue has been discovered in how calm winds are encoded in the ISD after 2013. We describe the cause and impacts in Dunn et al, 2022, ERC. From this version onwards we have adjusted our routines to recover these calm periods. We have also updated the record values used by the World Record Check, and hence there has been an extra increment in the version number (08-Feb-2022, updated 08-Jul-2022)

Note for version and later: We have updated the calculation for the wet bulb temperature to reflect that now used for HadISDH. The original formula from Jenson et al (1990) was giving strange values at low RH at very high temperatures. The routines have been updated to use the formula from Stull (2011). For further details see forthcoming post on the HadISDH blog. (10-Jan-2023). Updated from version v3.3.1.202308p onwards to correct a bug.

HadISD is a global sub-daily dataset based on the ISD dataset from NOAA's NCEI. As well as station selection criteria, a suite of quality control tests has been run on the major climatological variables. The data span 1/1/1931 to end August 2023 and each station is available individually. The current version of HadISD is For previous versions please contact the dataset maintainers.

This is a new version of HadISD, version This version is the preliminary one for 2023 and hence has the final character "p". The station lists have been updated over v3.3.0.2022f, but will remain the same until February 2024 and v3.3.2.202401p. Further details are given in the blog and also a Hadley Centre Technical Note. For full details see the Changelog.

A station listing with IDs and location information is here (also with names).

Brief description of the data

Individual stations within the ISD were composited when it was appropriate to do so. Then stations were selected on the basis of their length of record and reporting frequency. There are 9673 stations in HadISD., and these were passed through a suite of automated quality control tests designed to remove bad data while keeping the extremes. None of the ISD flags were used. The QC tests focussed on the temperature, dewpoint temperature and sea-level pressure variables, although some were applied to the wind speed and direction and cloud data. The data files also contain other variables which were pulled through from the raw ISD record, but have had no QC applied. Note: These data have not yet been homogenised and so trend fitting should be undertaken with caution. Results from the homogeneity assessment can be found here.

Keep in touch

Follow us on twitter: @metofficeHadOBS for updates, news and announcements.

For more detailed information, follow our HadISD blog. Here we describe bug fixes, routine updates and other exploratory analysis.

HadISD sample image

Figure shows the distribution of HadISD stations on a 1x1degree grid. These are the subset from the ISD which report at least every 6 hours and have a long enough record. It was created when refreshing the station list for v3.3.1.202308p.

Go to download page Data are available from the download page without charge under the Non-Commercial Government Licence. Please read the terms and conditions.


When using the dataset in a paper, the following are the correct citations to use. Please also state the version number used and date of download

Dunn, R. J. H., (2019), HadISD version 3: monthly updates, Hadley Centre Technical Note

Dunn, R. J. H., et al. (2016), Expanding HadISD: quality-controlled, sub-daily station data from 1931, Geoscientific Instrumentation, Methods and Data Systems, 5, 473-491

Dunn, R. J. H., et al. (2014), Pairwise homogeneity assessment of HadISD, Climate of the Past, 10, 1501-1522

Dunn, R. J. H., et al. (2012), HadISD: A Quality Controlled global synoptic report database for selected variables at long-term stations from 1973-2011, Climate of the Past, 8, 1649-1679

Smith, A., et al. (2011): The Integrated Surface Database: Recent Developments and Partnerships. Bulletin of the American Meteorological Society, 92, 704-708

Product User Guide

An introduction to the data set is available as a pdf, which includes a "quick start" guide as well as more in depth material including.

Commercial and media enquiries

You can access the Met Office Customer Centre, any time of the day or night by phone, fax or e-mail. Trained staff will help you find the information or products that are right for you.
Contact the Met Office Customer Centre

Quality Control Code

The python code used for the station selection and quality control can be found on github.
We do not intend to provide in-depth support for this code. We emphasise that the choices we have made in the selection and QC one of a number of other equally reasonable choices, and encourage other interested parties to develop their own QC and filtering procedures to allow the structural uncertainties of this sub-daily dataset to be investigated.

Homogeneity Assessment

We have assessed the homogeneity of four of the observed meteorological variables present in HadISD: temperature, dew point temperature, sea-level pressure and wind speed. This has been performed on monthly averages of the sub-daily data, and is fully described in Dunn et al, 2014, Climate of the Past, 10, 1501-1522. The homogeneity assessment results for v3.3.0.2022f are available for download here.

Online Material

A number of summary and diagnostic outputs from the quality control code as well as station lists for a number of selection criteria are available here. Also available is the information from the homogeneity assessment of v3.3.0.2022f.

We have adapted our versioning system from CRUTEM4, and so the dataset numbering is of the form HadISD.X.Y.Z.1234i, and is outlined in more detail in the paper linked above. X is for a major change and would normally be accompanied by a peer-reviewed paper. Y is a more minor change, e.g. in one of the QC tests and would be described in a tech-note. Finally Z is a small change, for example addition or changes to data in the past. The last complete year of the dataset is given by 1234, and the final character shows if the dataset is f-final or p-preliminary. Therefore HadISD. is the final version of the dataset containing data up to the end of 2012, and there have been some changes to past data over HadISD. For the monthly updates, the final characters encode the final year and month of the dataset, e.g. 202001p for the version up to the end of January 2020.

A number of images and movies created from HadISD to show its capabilities are here.

Dataset produced in collaboration with: