Comprehensive Global Nuclear Test and Detonation Dataset

Full report

On this page

Executive summary

I compiled an event-level CSV covering 2,058 nuclear test or detonation records. The backbone is the SIPRI/FOA report Nuclear Explosions 1945–1998, which explicitly says it lists all known nuclear explosions for the United States, Soviet Union, United Kingdom, France, China, India, and Pakistan and that it draws on official information from the United States, Russia, and France. I then appended the six DPRK declared nuclear tests using CTBTO detection records and added one low-confidence disputed row for the 1979 Vela incident using declassified U.S. intelligence records and later archival/scientific reanalysis.

The attached file is event-level, not device-level. That matters for clustered same-day detonations. SIPRI’s own later yearbook notes that India’s detonations on 11 and 13 May 1998 are counted as one test for each date, and Pakistan’s five detonations on 28 May 1998 are counted as one test. I therefore preserved event-level rows and populated the number_of_devices field where the grouping is confidently established.

I also included the Hiroshima and Nagasaki combat detonations because your request asked for tests and detonations. The U.S. DOE/NNSA official test volume explicitly excludes them because they were not “tests,” while the SIPRI/FOA global explosion list includes nuclear explosions broadly; I therefore retained them and marked their purpose separately.

A key completeness caveat remains: the structured SIPRI extraction maintained by the data-is-plural repository warns that the scanned source can contain transcription errors, and the SIPRI/FOA report itself cautions that some data may be incorrect because it was compiled from multiple sources. The most important consequence is a persistent Soviet total discrepancy: this compiled CSV preserves 714 Soviet rows from the extracted baseline, while later tallies often cite 715 Soviet tests. I flag that discrepancy rather than inventing a synthetic row with uncertain metadata.

Methodology and completeness

The source hierarchy was: official national publications and archives first, then SIPRI/FOA as the global harmonized backbone, then CTBTO for DPRK-era detections and verification notes, and finally declassified intelligence and peer-reviewed archival/scientific work for disputed undeclared events such as Vela. For the United States I prioritized DOE/NNSA’s United States Nuclear Tests July 1945 through September 1992; for France I used the official CIVEN list and CEA/DAM’s Polynesia history; for the Soviet Union I relied on SIPRI’s documented use of the Russian Ministry for Atomic Energy and Ministry of Defense compilation; for the UK I cross-checked against Ministry of Defence factsheets and GOV.UK archival summaries; and for DPRK I used CTBTO’s event pages and verification brochure.

Dates were normalized to the source’s UTC/GMT event chronology wherever the source gave a UTC/GMT origin time. That means some entries that are popularly remembered by local date can appear one day earlier in UTC, most visibly Hiroshima in the SIPRI chronology. I kept that choice because you explicitly requested UTC time if available, and the DOE/SIPRI source structure is time-zone aware.

I mapped source test-mode codes into your requested higher-level categories. In the SIPRI definitions, atmospheric codes include tower, airdrop, balloon, rocket, barge, surface, and water-surface modes; underground codes include shafts, tunnels, galleries, and mines; and a few entries are underwater. For the CSV, I collapsed them into atmospheric / underground / underwater / space / other, and I put ambiguous surface-water or ship/barge cases into other unless the source clearly indicated underwater.

I did not add alleged zero-yield “cold tests,” subcritical experiments, or later claims of possible post-moratorium supercritical activity unless there was a credible claimed or detected nuclear explosion. That is consistent with the CTBT’s focus on banning all nuclear explosions, and it keeps the file aligned with your request for tests and detonations rather than the broader universe of nuclear-weapons-related experiments.

CSV deliverable

Download the CSV

The CSV includes the requested analytical fields:

• unique ID
• test name or operation name
• date
• UTC time where available
• country or operator
• site/location
• test type
• delivery/platform
• reported yield
• weapon design/type
• purpose
• number of devices where confidently known
• casualties/incidents
• international monitoring notes
• primary sources
• notes/remarks
• provenance
• confidence level

I also added explicit latitude and longitude columns for convenience, while retaining the combined location field in the format you requested.

Summary tables

The following counts are derived from the attached 2,058-row compiled CSV described above. They therefore reflect the event-level methodology, the inclusion of Hiroshima and Nagasaki as detonations, the six DPRK tests detected by CTBTO, and the single disputed Vela row.

Counts by country/operator

Counts by decade

Counts by test type

The dataset’s peak year is 1962, with 178 event records, and the long-run shift is strongly toward underground testing, which matches the historical transition described in SIPRI’s background section after the Partial Test Ban era.

Timeline of testing

This timeline is computed directly from the compiled CSV. The sharp crest in the late 1950s and early 1960s, peaking in 1962, is followed by a steep reduction in atmospheric activity after the 1963 treaty era described by SIPRI; the only post-1998 confirmed additions are the six DPRK tests detected by CTBTO.

xychart-beta
    title "Nuclear test and detonation events per year"
    x-axis [1945, 1946, 1947, 1948, 1949, 1950, 1951, 1952, 1953, 1954, 1955, 1956, 1957, 1958, 1959, 1960, 1961, 1962, 1963, 1964, 1965, 1966, 1967, 1968, 1969, 1970, 1971, 1972, 1973, 1974, 1975, 1976, 1977, 1978, 1979, 1980, 1981, 1982, 1983, 1984, 1985, 1986, 1987, 1988, 1989, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017]
    y-axis "Events" 0 --> 178
    bar [3, 2, 0, 3, 1, 0, 18, 11, 18, 16, 24, 33, 55, 116, 0, 3, 71, 178, 50, 60, 58, 76, 64, 79, 67, 64, 53, 56, 48, 55, 44, 51, 54, 66, 59, 54, 50, 49, 55, 57, 36, 23, 47, 40, 28, 18, 14, 8, 1, 2, 7, 3, 0, 4, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 2, 1]

Major disputed and uncertain entries

The most important caveats are these. I kept them explicit in the CSV through the provenance, confidence_level, and notes_remarks fields.

• Vela incident, 1979-09-22. Declassified U.S. intelligence material includes an assessment giving the event a “90% plus” probability of being a nuclear test, while other contemporaneous U.S. review processes remained officially inconclusive. Later archival and scientific work has generally strengthened the pro-test interpretation, but attribution remains unresolved. I therefore included it as Undeclared / disputed with low confidence rather than omitting it or assigning it firmly to a state.

• Soviet total-count discrepancy. The structured SIPRI extraction used for row-level metadata yields 714 Soviet entries, but later tallies often cite 715 Soviet tests. Because the repository explicitly warns of possible transcription errors from the scanned source, and because the SIPRI report itself urges caution, I preserved the extracted rows and surfaced the discrepancy in this report instead of manufacturing a placeholder event with uncertain date, site, and yield.

• India and Pakistan in May 1998. Same-day grouped detonations are treated in SIPRI/SIPRI-yearbook convention as one test row per qualifying date, even though multiple devices were fired. That is why the file records India as 3 event rows and Pakistan as 2 event rows, while also carrying device-count metadata for Shakti and Chagai where confidently known. Yields for these entries remain heavily disputed in outside literature, so the CSV preserves conservative reporting and explanatory notes.

• DPRK design characterization, especially 2017. CTBTO data firmly establish the occurrence, timing, location area, and magnitude evolution of all six DPRK tests, but CTBTO’s role is to characterize the event and not to certify exact weapon design unless corresponding evidence exists. Because outside assessments of the 2017 device range from boosted-fission to staged thermonuclear interpretations, I left weapon_design_type as unspecified unless the underlying evidence was unusually clear.