Live · updated every minute
Hazards near you and around the world, right now.
Earthquakes as they happen, every active tropical cyclone and tornado warnings — beside our forecasts of what is likely next, each traceable to its data, its model and its track record.
Near you · Columbus, US
From your approximate location, used for this page view only.
Official alerts
1
Frost Advisory in effect until 13:00 UTC: Frost Advisory issued October 18 at 1:22PM EDT until October 19 at 9:00AM EDT by NWS Wilmington OH
National Weather ServiceEarthquakes
Quiet
No earthquake of M2.5+ within 1,000 km in the last 24 hours.
Chance of an M6+ in your 2° cell, next 30 days: <0.1%
Earthquake forecastTropical cyclones
None nearby
No active tropical cyclone within 2,000 km. 3 active elsewhere.
All active stormsThunderstorms
None nearby
No thunderstorm tracked within 300 km (tracked ). Storm tracking covers the US and its edges.
Tornado forecastHazardPulse publishes research forecasts and relays official reports; it is not a warning service. In an emergency, follow your official warnings and local authorities. Disclaimer
Drag to turn the globe · select a marker for details
- Earthquake chance by cell (shaded)
- Highest earthquake cells
- Tropical cyclone
- Tracked thunderstorm
- Earthquake, last 24 h
- Tropical cyclone
- Tornado warning
- 30-day earthquake forecast
- You
- 35earthquakes M2.5+ in the last 24 hours
- 3tropical cyclones active worldwide
- 0tornado warnings in effect in the US
- 1thunderstorms tracked in the latest tornado forecast
Live counts and events update every minute · as of this page view
Happening now
What the agencies that observe hazards have just reported, newest first.
- M3.346 km WSW of Ivanof Bay, Alaska
- M2.925 km WNW of Puerto Real, Puerto Rico
- M2.7120 km SE of Ugashik, Alaska
- 70 ktTrudy (Hurricane, Category 1)
- M2.812 km SSW of Waikoloa, Hawaii
- M2.519 km SSW of Jal, New Mexico
- M2.718 km SW of Jal, New Mexico
- 135 ktChampi (Super typhoon, Category 4)
- M4.433 km SE of Sơn La, Vietnam
- M4.4Kermadec Islands region
- M4.5Fiji region
- M5.1129 km SSE of Kokopo, Papua New Guinea
- M2.515 km NE of Milford, Utah
- M4.7147 km SSW of Kokopo, Papua New Guinea
- M2.761 km NNE of Miches, Dominican Republic
- M4.4223 km NE of Lospalos, Timor Leste
Observations from the USGS, the National Hurricane Center and the National Weather Service, as they publish them. Storms outside NOAA’s areas come from the HazardPulse forecast feed.
What is likely next
Our forecasts, updated . The three windows differ — 30 days, 24 hours, 60 minutes — so the three numbers are not comparable with each other.
Highest chance of an M6+ earthquake in any 2° grid cell
Southeast of Loyalty Islands 21°S 169°E
−0.5 pts since the previous forecast (18 Oct 2026, 18:00 UTC)
Earthquake forecastHighest chance of rapid intensification: Champi
West Pacific · source: HazardPulse v8.3
+6.8 pts since the previous forecast (18 Oct 2026, 21:32 UTC)
Hurricane forecastHighest chance a tracked thunderstorm produces a tornado
6 mi NNW of Alamogordo, NM
no change since the previous forecast (19 Oct 2026, 03:17 UTC)
Tornado forecastIn an emergency, follow official warnings: the National Weather Service, the National Hurricane Center, the USGS, your national meteorological and geological services, and your local authorities. HazardPulse publishes research forecasts, not warnings.
How the forecasts have tested
Each figure is read from the results files of the model version that publishes the number, in tests written down before the data was scored; the earthquake figure is a later second look at its test years, and says so. Ranking accuracy (AUC) is 50% for a coin flip and 100% for a perfect ranking.
96.9%
of the time, a cell that went on to have an M6+ earthquake was ranked above one that did not (a long-term seismicity map alone: 96.7%; difference +0.19 points, 95% interval +0.09 to +0.29)
Every 2° cell, 2023–2025 (a second look at those years)
93%
ranking accuracy of NOAA DTOPS, the probability we publish for the Atlantic, East and Central Pacific; in the same test it beat NOAA SHIPS-RII, NOAA RIOC consensus and our previous model (v8.2). Elsewhere we publish our v8.3 model; it was tested on 2,967 2025–2026 cycles in the basins where it publishes, none used to fit it, 1,078 of them West Pacific
The 2025 season, scored once, 601 forecast cycles
97%
of the time, a storm that went on to produce a tornado was ranked above one that did not (NOAA’s ProbTor on the same storms: 94%, its whole-percent ties broken)
Every storm NOAA tracked in 2025, scored once
How a forecast is made
Public data in
Earthquake catalogs from the USGS, tropical cyclone guidance from NOAA’s National Hurricane Center, and NOAA’s live storm tracking and National Weather Service warnings.
A fixed, versioned model
Each model was chosen by a test written down before the data was scored, and is recorded in the registry with the hash of its file. Changing it means a new version and a new test.
Checked before it is published
Automatic quality checks run on every forecast: fresh inputs, a valid schema, sane values, a complete record. A failed blocking check stops it.
Frozen, then scored
Every forecast is saved with its inputs and hashes, chained into a ledger that shows any later edit, and scored against what actually happened.
Build on it
Open API
Every live forecast, its record and its track record as JSON, and the live feed itself. No key needed.
API referenceOpen data
The research datasets behind every model, in versioned releases with a hash for every file.
Data releasesModel registry
Every model version that has published a number, why it was promoted, and why it was retired.
See the models