Quality and availability of tornado data varies greatly across the world, and it’s important to know how much you can trust what you’re seeing on the Explorer. This page provides descriptions of domain, quality, and availability for each dataset on Tornado Archive (you will see the names of these in the “Source” field when a tornado is selected on the explorer). In the future, we aim to reduce the limitations they have, and extend our coverage into more countries, by manually compiling online information for individual events. At some point we might even allow (sourced) contributions from the public.
Some Tornado Archive members have already added data to our site. When certain tornadoes are selected, you will see links to sources other than the datasets described here, simply labeled “Source 1”, “Source 2”, and so on. We ensure that all information we display has a source.
Please contact us here or on Twitter regarding questions, concerns, errors, or potential enhancements to our data.
North America
SPC: Storm Prediction Center. https://spc.noaa.gov/wcm/#data
Domain: United States, 1950-2019.
Status: Public Domain, Available for Download.
Description: Storm prediction center data improves over time. Ratings and tracks especially prior to 1970 contain many errors.
NCEI:
National Centers for Environmental Information – Storm Events Database
Domain: United States, 1950-2021.
Status: Public Domain, Available for Download.
Description:
SPC data is derived from NCEI data, so the two datasets are very similar. However, by combining the two we are able to access additional path information as well as event descriptions from NCEI.
DAT:
NOAA Damage Assessment Toolkit
Domain: United States, ~2013-2022 (currently using up to 2019).
Status: Public Domain, Available for Download.
Description:
Highly detailed paths and damage polygons. The first record is from 1933 but coverage is very limited until around 2013 onward.
TPG:
Thomas P. Grazulis – Significant Tornadoes 1880-1950 and Significant Tornadoes 1680-1991
Domain: United States, 1680-1991 (we only use up to 1949).
Status: Public Domain, Available for Download (future).
Description:
Grazulis’ data is “reverse engineered” using descriptions in his book and additional sources, and may contain errors. Most original data is just start and end coordinates, which may mean some tracks do not go through specific towns or areas where they should. This data also only includes F2+ tornadoes as well as deadly F1s.
ECCC:
Environment and Climate Change Canada
Domain: Canada, 1980-2009. ~1,840 events.
Status: Public Domain.
Description:
Data quality is quite good. We are aware of further datasets that will be published in the future, and we will use them once they are available.
For tornadoes where this dataset provided start and end coordinates for the tornado, but no path length, we calculated it ourselves.
NTP:
Northern Tornadoes Project
Domain: Canada, 2017-2021. ~370 events.
Status: Public Domain.
Description:
Data quality is very good, but coverage is not as full for 2017-2018.
We use a combination of NTP’s point event data and damage track data.
León-Cruz et al 2022:
Research dataset
Domain: Mexico, 2007-2020. ~350 events.
Status: Obtained from author.
Description:
All tornadoes are single points and a small portion have ratings. Quality is good.
Europe, Central Asia and North Africa
ESWD:
©ESSL European Severe Storms Laboratory – European Service Weather Database
Domain: Europe, Russia, Central Asia, Mediterranean coast, 0066-2021. ~17,800 events.
Status: Used with permission. No commercial use of this data is allowed without written permission of the ESWD or direct purchase of data: https://eswd.eu
Description: ESWD data is aggregated from (interpreted) legend, historical texts, government data, newspapers, news websites and modern social media. Quality of data varies by country and reliability of data improves by 2000+ CE. Some tracks are multiple points rather than true start and end coordinates. Some dates or times are approximate and should be treated with caution. “Possible” tornadoes have been removed.
If maximum path width is not provided for a tornado, but mean path width is, we use that instead.
South America
Balbi and Barbieri 2017:
Research dataset
Domain: Argentina, Uruguay, 1889-2016. ~330 events.
Status: Public domain.
Description: This dataset has fairly consistent data between 1930 and the start of 2016. The data itself is quite limited: most events have ratings, but there are no paths (just points), and no fatality or injury counts. The locations are probably approximate, since most events are simply placed in town centers.
One tornado from Brazil is included in this dataset.
Bertoni 2013:
Research dataset
Domain: Brazil, 1923-2009. ~85 events.
Status: Public domain.
Description: Locations are generally not more accurate than town centers.
Vicencio:
Research dataset
Domain: Chile, 1880-2021. ~50 events.
Status: Obtained from author.
Description: Data is of good quality. A couple tornadoes have detailed paths; most are points.
Oceania
BOM:
Australia Government Bureau of Meteorology: http://www.bom.gov.au/australia/stormarchive/
Domain: Australia, 1795-2019. ~1,320 events.
Status: Public domain.
Description: This dataset is relatively low-quality – rating and location information are unreliable, and there are many overlapping points. Please use with caution. We hope to obtain a better Australian dataset later this year, once it is published.
For tornadoes where this dataset provided start coordinates, path length, and direction, but no end coordinates, we calculated them ourselves.
Unfortunately, unrated tornadoes are listed as F0 in the original dataset.
NIWA:
National Institute for Weather and Atmospheric research: New Zealand Historic Weather Events Catalogue
Domain: New Zealand, 1908-2019. ~120 events.
Status: Creative Commons 3.0 License. Free to use, share, adapt, transform, with attribution.
Description: Events are almost entirely single points; the ones that are paths are approximate, and most events are not rated. None of the information seems egregiously inaccurate. We may acquire a better dataset later this year.
East Asia
JMA:
Japan Meteorological Agency: https://www.data.jma.go.jp/obd/stats/data/bosai/tornado/index.html
Domain: Japan, 1961-2016. ~980 events.
Status: Public Domain.
Description: Tornadoes with ambiguous time or location information were not included. Data quality seems quite good.
When more than one rating is given in the source, the lowest is selected (so F1-F2 becomes F1).
For tornadoes where this dataset provided start and end coordinates for the tornado, but no path length, we calculated it ourselves.
Zhou et al 2020:
Research dataset
Domain: China, 2007-2016. ~980 events (some overlap with below).
Status: Public Domain.
Description: Data quality is uncertain. Data is very limited – only date and location information available. We are aware of a much, much more comprehensive dataset (https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.5369) but thus far the author has not responded to our request. If you have connections with Jiayi Chen, Xuhui Cai, any other authors of the paper, or Peking University, please let us know.
Zhang et al 2022:
Research dataset
Domain: China, 2003-2019. ~700 events (some overlap with above).
Status: Public Domain.
Description: Data quality is uncertain. Data is very limited – only rating, date, and location information available. We are aware of a much, much more comprehensive dataset (https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.5369) but thus far the author has not responded to our request. If you have connections with Jiayi Chen, Xuhui Cai, any other authors of the paper, or Peking University, please let us know.
Records from this dataset were automatically matched with records from the above dataset if they were close in time and space. There were some ambiguous cases but this process removed most duplicates.
CTDEL:
China Tornado Disaster Event List
Domain: China, 1966-1990. ~90 events.
Status: Public Domain.
Description: Data quality seems OK.
South Asia
Finch, Hosen: Hosen et al. 2016 and Jonathan Finch
Domain: Bangladesh and eastern India, 1838-2013. ~100 events.
Status: Public Domain.
Description: Data is very limited; locations are very approximate.
South Africa
Muller 2022: Research dataset
Domain: South Africa, 1905-2022. ~340 events.
Status: Public Domain.
Description: Locations are generally not more accurate than town centers.