How Shockradar's global earthquake forecast works
Shockradar is an experimental research map of estimated earthquake probabilities for 5° × 5° geographic cells worldwide. Choose a magnitude threshold and a forecast horizon to explore the map.
Important safety notice: Shockradar is experimental. It does not predict the exact time or epicenter of earthquakes, is not an earthquake early-warning system, and must not be used to make public-safety decisions. Follow official earthquake agencies and local authorities.
Understanding the map
| Selection | Meaning |
|---|---|
| 1 day | Estimated probability of at least one earthquake during the 24 hours beginning at the stated forecast issue time. |
| 7 days | Estimated probability during the seven days beginning at the stated issue time. |
| M4.5+ | At least one earthquake of magnitude 4.5 or greater within the cell. |
| M5.0+ | At least one earthquake of magnitude 5.0 or greater within the cell. |
| M6.0+ | At least one earthquake of magnitude 6.0 or greater within the cell. |
The blue-to-red legend indicates increasing forecast probability. A low-probability blue cell can still experience an earthquake. Geographic cells are not equally sized in square kilometers, especially near the poles.
How are forecasts produced?
The EFAIM research model uses historical seismic activity and recent earthquake observations. Its input catalogs include earthquakes of magnitude 4.5+, additional magnitude 3.5–4.5 context, and magnitude 2.5–3.5 context when available. Smaller-event observations are model inputs, not extra forecast targets.
Catalog sources can include the USGS Earthquake Catalog and the International Seismological Centre (ISC). Reporting delays, geographic detection differences, and subsequent catalog revisions may change the available observations. The date shown on the map identifies the forecast issue window.
What does a 20% probability mean?
A seven-day M5.0+ probability of 20% in a cell means the model estimates a one-in-five chance of at least one M5.0-or-larger earthquake in that region during that period. It does not identify the earthquake's exact location, depth, or time.
Is a blue region safe?
No. Low probabilities do not guarantee safety. For earthquake preparedness, use official seismic hazard guidance and local emergency-management information.
How is forecasting accuracy evaluated?
Forecasts should be saved before their observation windows and compared with earthquakes that occur afterward. Useful measures include Poisson negative log-likelihood, Brier scores, probability calibration, top-ranked-cell event coverage, and comparisons against historical-rate baselines. A visually concentrated map alone does not prove predictive skill.
Open the global 1-day and 7-day earthquake forecast
Project updates: @henrikvaughnt on X. Coastline attribution: GSHHG/NOAA.