The Deadliest Earthquakes of the 21st Century: Every Quake That Killed 1,000 or More (Mapped)

Twenty-three earthquakes since 2001 have killed at least 1,000 people. Here is every one, ranked, with the disputed tolls laid out and the shaking mapped.

What is the deadliest earthquake of the 21st century? The magnitude 7.0 Haiti earthquake of January 12, 2010. Haiti's government put the toll at 316,000, the figure carried by NOAA's National Centers for Environmental Information (NCEI) and the US Geological Survey (USGS). That toll is disputed: USGS notes that other estimates are "perhaps as low as fewer than 100,000". If the lower estimates are right, the deadliest is the December 26, 2004 Sumatra–Andaman earthquake, whose tsunami killed people in 10 countries around the Indian Ocean (USGS): 227,899 dead or missing according to NCEI, more than 283,100 killed according to USGS. Below are the ten deadliest, the full list of every earthquake since 2001 that killed 1,000 or more, why the counts differ, and what DisasterAWARE recorded for the ones since 2014.

How This List Is Ranked

• One source for the ranking. Every earthquake is ranked by NCEI's total deaths in its Significant Earthquake Database, pulled on October 9, 2026, covering 2001 to date. Using one source means every event is counted the same way. The USGS figure for each event, from its earthquake summary, sits alongside.
• Tsunami deaths are included. NCEI's total counts deaths from the earthquake and from its secondary effects, chiefly tsunamis (NCEI's definition: "the earthquake and secondary effects (eg Tsunami)"). That matters most for 2004 and 2011, where the tsunami did most of the killing.
• Missing people. NCEI's figure includes people missing and presumed dead where its sources do (Sichuan 2008, and the 2004 "dead or missing" total). Where NCEI lists missing people separately, they are not in the total, and we show them beside it: Boumerdès 2003 (800), Palu 2018 (667), Nippes 2021 (329), Myanmar 2025 (116) and Nepal 2015 (3).
• Magnitudes are USGS's preferred values. Dates are in UTC.

The 10 Deadliest Earthquakes of the 21st Century

Sources: NOAA NCEI Significant Earthquake Database (deaths); USGS ComCat earthquake summaries (other published tolls and magnitudes). Both queried October 9, 2026.

Two of the ten struck in the last four years, Türkiye–Syria in 2023 and Venezuela in 2026, and Myanmar 2025 ranks 13th. Two of the ten, 2004 and 2011, killed mainly through their tsunamis.

Where the Death Tolls Are Disputed

• Haiti 2010. The 316,000 is the Haitian government's figure, raised by the prime minister on the first anniversary of the quake. NCEI notes that "other estimates suggest substantially lower numbers of casualties", and USGS that they are "perhaps as low as fewer than 100,000". On the official figure Haiti ranks first. If the true toll was under 100,000, Haiti would fall below the 2004 tsunami, and below Sichuan too if it was under 87,652.
• Sumatra–Andaman 2004. NCEI's 227,899 is based on the Tsunami Evaluation Coalition's synthesis report (which gives 227,898) and counts the dead and missing together. USGS gives more than 283,100 killed plus 14,100 still missing in 10 countries. Either figure ranks it second.
• Türkiye–Syria 2023. NCEI has 56,697. USGS counts at least 50,399 dead in Türkiye and 8,476 in Syria, 58,875 in all, and Munich Re puts the toll at about 58,000. The ranking is the same on any of them.
• Kashmir 2005. NCEI has 76,213 for Pakistan, India and Afghanistan together. USGS reports at least 86,000 killed in Pakistan and 1,350 in India. On the USGS count it would still rank fourth, just below Sichuan's NCEI total.
• Tōhoku 2011. NCEI's 18,423 is the combined earthquake and tsunami toll. USGS counts 15,703 killed and 4,647 missing. Counting the missing, the USGS figure of 20,350 would move Tōhoku above Gujarat (20,005).
• Venezuela 2026. The 6,509 is the confirmed toll as of the last official update. Counts of the missing are far apart: USGS (as of August 24, 2026) lists an estimated 50,000 people missing; the official count of missing has stood at 157 since June 25; and NGO tallies put it at 41,000 (El Nacional, September 25, 2026). Munich Re's half-year review gave 5,500 deaths. If a large share of the missing are confirmed dead, Venezuela would move far up this list. We will update it when the counts are revised.

Tsunami Deaths: 2004 and 2011

Two of the ten killed mainly by water, not shaking.

• 2004. NCEI's source, the Tsunami Evaluation Coalition, says no separate death toll is available for the shaking, because the tsunami followed within 20 minutes, and estimates that the shaking alone probably killed fewer than 1,000. NCEI's tsunami database carries the same 227,899 deaths for the tsunami.
• 2011. NCEI's earthquake record attributes 1,474 deaths to the earthquake itself and 18,423 to the earthquake and tsunami together. The USGS summary says the majority of casualties and damage in Iwate, Miyagi and Fukushima came from the tsunami.

ShakeMap, USGS's estimate of how hard the ground shook, makes the same point. The peak intensity for 2004 is MMI 7.9 and for Tōhoku MMI 8.2, lower than Haiti (9.0), Gujarat (9.4) or Türkiye (9.5). Tsunami deaths also run through the rest of the list: at Palu in 2018, USGS attributes the toll to "the combined effects of the earthquake, tsunami, liquefaction and landslides".

Every 21st-Century Earthquake That Killed 1,000 or More: Numbers 11 to 23

Sources: NOAA NCEI Significant Earthquake Database; USGS ComCat earthquake summaries; queried October 9, 2026. One NCEI entry is left out: a magnitude 4.7 shock on January 17, 2002, centred in DR Congo, that damaged Gisenyi, Rwanda. Its total of 1,001 does not match NCEI's own description, which says "several people killed" by the quake, one of a series that accompanied the eruption of Nyiragongo volcano in Congo.

Below 1,000, the next deadliest since 2001 are El Salvador in January 2001 (844), Awaran, Pakistan, in 2013 (825) and the 2006 Pangandaran earthquake off Java (802), whose deaths USGS attributes entirely to its tsunami.

Map: Where the Deadliest Earthquakes Struck

The map shows all 23 earthquakes with 1,000 or more deaths, each at its USGS epicentre and sized by death class (four sizes, by NCEI deaths). For the ten deadliest, it adds USGS ShakeMap intensity contours at MMI VI (strong), VII (very strong), VIII (severe) and IX (violent), where each ShakeMap reaches them. Click a dot for magnitude, depth, deaths and what DisasterAWARE recorded. Zoom in to see the contours. Shaded areas, on at load, are ShakeMap intensity (MMI IV and up) for all 23; where two overlap, the stronger shaking shows.

Open map full screen ↗

Explore the map: switch on more layers with the layers icon at the bottom right of the map: USGS "Did You Feel It?" felt reports for 17 quakes; aftershocks of magnitude 4.5 or more in the first week, for the five deadliest (1,100); NOAA tsunami height and runup measurements for 2004 and 2011 (8,489); and USGS finite-fault rupture models for 15 quakes, with fault slip at zoom 6 and closer.

What the map can show, by era. Every earthquake on the list has a USGS epicentre and a ShakeMap, from 2001 to 2026. For the older events, and for some recent ones such as Nepal 2015 and Palu 2018, the ShakeMap in USGS's catalogue is a reconstruction from its ShakeMap Atlas. For the others, it is the operational ShakeMap issued and revised after the quake: Türkiye 2023 is on version 19 and Myanmar 2025 on version 23. Felt reports exist for 17 of the 23 (none for Gujarat, Nahrin, Bam, Boumerdès or Nias, and none for the first Herat shock, the one mapped), finite-fault models for 15, and USGS PAGER alerts, shown in the popups, for the ten since 2015.

In the DisasterAWARE Archive Since 2014

Every one of the ten deadliest earthquakes since 2014 is in the DisasterAWARE historical hazard archive, pulled fresh on October 9, 2026. The archive is alert-grain: one earthquake can have several records (mainshock, big aftershocks, separate shocks of a doublet). We matched each earthquake to its records by time, place and USGS event ID, and counted it once. NCEI's death toll sits on exactly one record per earthquake, in the record's impact block.

Source: DisasterAWARE Historical Hazards API, queried 2026-10-09. "Record created" is the time between the earthquake and the creation of DisasterAWARE's record, our closest measure of when the alert went out. ShakeMap peak MMI is USGS's modeled estimate of the strongest shaking. The alert footprint is the population inside the shaking area of the alert, not the number of people affected.

What the records show:

• Türkiye–Syria 2023. DisasterAWARE's record of the magnitude 7.8 was created at 01:30 UTC, 13 minutes after the quake. It carries ShakeMap version 19, with peak MMI 9.54. The magnitude 7.5 that struck about nine hours later, at 10:24 UTC, has its own record; NCEI counts all the deaths under the first shock, so the event is counted once here.
• Myanmar 2025. The record was created about 20 minutes after the quake, with ShakeMap peak MMI 9.95, the highest of any earthquake on this list. Its alert footprint held 48.46 million people in shaking of MMI V or more, 6.82 million of them in severe (VIII) or stronger shaking.
• Venezuela 2026. Two large shocks struck 32 seconds apart: a magnitude 7.2 at 22:04 UTC and a magnitude 7.5 at 22:05. DisasterAWARE's archive record is for the first shock and was created 13 minutes after it; the M7.5's own record is still an active hazard on the DisasterAWARE platform, so it is not yet in the historical archive. The record's alert footprint held 7.70 million people in MMI V or more, 257,000 of them in VIII or stronger. Our deep dive on the doublet explains why it was so deadly.
• Herat 2023. Two magnitude 6.3 shocks struck 32 minutes apart on October 7, with a 5.4 between them and a 5.9 about half an hour after the second. The archive holds four records; NCEI's 1,482 deaths sit on the first.
• Nepal, Palu, Haiti 2021, Paktika. The records carry the USGS ShakeMap and PAGER alert, but no exposure block, so no alert-footprint figure is shown.

For Insurers and Reinsurers: The Deadliest Quakes Are Not the Best Insured

The deadliest earthquakes strike where buildings are weakest and insurance is thinnest, so the death toll and the insured loss often point in opposite directions. Munich Re's figures show the gap:

• Türkiye–Syria 2023: overall losses of about US$50 billion, insured losses of US$5.5 billion, about 11%, even though earthquake insurance is compulsory for homes in Türkiye through the Turkish Catastrophe Insurance Pool (Munich Re, January 9, 2024).
• Myanmar 2025: overall losses of about US$12 billion, of which US$1.5 billion was insured (Munich Re, January 13, 2026).
• Venezuela 2026: overall losses of about US$30 billion, with insured losses under US$0.5 billion, less than 2% (Munich Re half-year 2026 factsheet, July 30, 2026).
• Tōhoku 2011: overall losses of US$210 billion, of which insured losses may have reached as much as US$40 billion (Munich Re, January 4, 2012).
• For contrast, Christchurch, New Zealand, February 2011: 185 deaths (NCEI), about US$16 billion in losses, of which about US$13 billion was insured (Munich Re, January 4, 2012).

Parametric cover closes part of the gap, quickly. Haiti held an earthquake policy with CCRIF SPC, the Caribbean catastrophe risk insurance facility. CCRIF's policies pay when the modelled loss from an event equals or exceeds the attachment point the country chose, without waiting for claims. After the August 14, 2021 earthquake, CCRIF's loss model put Haiti's government losses above that point, and Haiti was due about US$40 million (US$39,953,272 in CCRIF's final event briefing), the full limit of its earthquake policy for 2021/22. CCRIF said it would pay US$15 million within a week and the rest within the 14-day window. That was CCRIF's largest single payout to that date. After the 2010 earthquake, Haiti had received US$7.8 million. Set against USGS's preliminary damage estimate of more than US$1 billion for 2021, even a full payout covers a small share.

The data question. A parametric trigger is only as good as its match to what happened on the ground. Basis risk shows up when the trigger's intensity measure, its footprint or its timing differs from the loss. DisasterAWARE's earthquake records put the pieces side by side for each event: the USGS ShakeMap peak MMI and peak ground acceleration with their contours, the PAGER alert level, NCEI's deaths, injuries, damage and houses destroyed, when the alert went out, and, where an exposure block exists, the population and capital in each shaking band. That is what you need to back-test an intensity-based trigger against past events, or to build a historical event set for a portfolio. Historical Event Intelligence licenses the archive worldwide; the sample datasets show the record format with no signup, and the earthquake data page covers the historical records and ShakeMap intensity data.

Frequently Asked Questions

What is the deadliest earthquake of the 21st century? The magnitude 7.0 Haiti earthquake of January 12, 2010, with an official toll of 316,000 (Haitian government, carried by NOAA NCEI and USGS). Other estimates are far lower, possibly under 100,000. If those are right, the deadliest is the 2004 Sumatra–Andaman earthquake and Indian Ocean tsunami: 227,899 dead or missing per NCEI, more than 283,100 killed per USGS.

What is the deadliest earthquake in modern history? On NCEI's figures, the deadliest since 1900 is Haiti 2010, at 316,000, followed by Tangshan, China, in 1976 (242,769), Gansu, China, in 1920 (235,502, often called the Haiyuan earthquake) and the 2004 Sumatra–Andaman earthquake and tsunami (227,899; USGS counts more than 283,100 killed). Because Haiti's official toll is disputed, the answer depends on which count you accept.

How many people died in the 2004 Indian Ocean earthquake and tsunami? NCEI records 227,899 dead or missing, based on the Tsunami Evaluation Coalition's report. USGS gives more than 283,100 killed and 14,100 missing in 10 countries. Nearly all of them were killed by the tsunami, not the shaking: NCEI's source estimates the shaking alone probably killed fewer than 1,000.

What was the deadliest earthquake of 2026? The Venezuela doublet of June 24, 2026: a magnitude 7.2 and a magnitude 7.5, 32 seconds apart, in northern Venezuela, west of Caracas. At least 6,509 deaths were confirmed by the last official update. Counts of the missing range from an official 157 to 41,000 in NGO tallies (El Nacional), and USGS (as of August 24, 2026) lists an estimated 50,000. It is the tenth deadliest earthquake of the century so far.

Why do earthquake death tolls differ between sources? Counts come from governments, relief agencies and surveys made weeks or months apart, and they differ in whether missing people, tsunami deaths and deaths in neighbouring countries are included. Haiti 2010 is the extreme case. That is why this list uses one source, NCEI, for the ranking and shows the USGS figure beside it.

Related Reading

• The Venezuela Earthquake Doublet: Why It Killed
• What Makes a Good Parametric Insurance Trigger
• Hazard Footprints and Reinsurance Exposure
• Earthquake Data: Historical Records and ShakeMap Intensity
• Earthquake Risk in Los Angeles, CA

Sources: NOAA NCEI Significant Earthquake Database and Global Historical Tsunami Database (hazel API, queried 2026-10-09); USGS ComCat earthquake summaries, ShakeMap and PAGER products (queried 2026-10-09); Munich Re press releases of January 4, 2012, January 9, 2024, January 13, 2026 and July 30, 2026, and its 2023 and 2025 full-year and 2026 half-year factsheets; CCRIF SPC release of August 18, 2021, Final Event Briefing for the August 14, 2021 Haiti earthquake (August 22, 2021) and Technical Paper 1 on its hurricane and earthquake policies (June 2012); El Nacional, September 25, 2026; DisasterAWARE Historical Hazards API (queried 2026-10-09).