Major events — last 21 days drag the globe · click a ping or a row · auto-rotates when idle
⟳ Checking live feeds…
What qualifies as major (thresholds, not vibes): earthquakes M ≥ 5.9 — 193/yr globally from this site's own catalog, one every ~1.9 days (~11 in a 21-day window); anything smaller would bury the map in noise. Cyclones, floods, volcanoes, wildfires, droughts and landslides at GDACS orange or red alert — GDACS's alert score is calibrated to expected humanitarian impact (greens excluded) — plus tropical storms at hurricane strength (≥ 64 kt, NASA EONET) with precise positions. If GDACS is unreachable, ReliefWeb's current-disaster list fills in, pinned at the affected country's centroid. Duplicate reports of the same event across sources are merged. Ping size scales with severity — a Category-5 or an M8 dwarfs a threshold-grade event. See METHODOLOGY § 12b.
Seismologist's watchlist sub-threshold anomalies · cyan pulses on the globe · click a row to fly there
What this is: the globe above only shows events big enough to matter on their own. This card hunts for the quieter pattern a seismologist watches — a spot producing far more small earthquakes than that same spot's own recent baseline. Every ~15 minutes it scans USGS's rolling 30-day M 1+ catalog, grids the planet (~83 km cells), and flags cells whose last-7-day count is statistically incompatible with their prior 23 days (exact Poisson tail p < 10⁻⁴, ≥ 10 recent events; neighboring flagged cells merge). Sequences whose largest magnitudes are still climbing get an escalating tag; a mainshock followed by ordinary decaying aftershocks is labeled expected and listed last. What this is not: earthquake prediction — no reliable precursor is known to science. Worldwide, the chance that any earthquake is followed by a larger one nearby within a week is about 5% (USGS), and most swarms simply end. Comparing each cell against itself keeps uneven network sensitivity from minting fake anomalies, but well-instrumented regions (US, Japan, Iceland…) will still dominate, because only they report M 1–2 at all. See METHODOLOGY § 12c.
Live pulse
What this page is — and is not
- Situational awareness, not statistics. This page shows what open live feeds report right now. Nothing here feeds the probability model — the Overview statistics come only from the validated historical record (METHODOLOGY § 12b documents the separation).
- Sources: USGS FDSN (earthquakes, public domain; the watchlist scans the rolling 30-day M 1+ feed) and GDACS (EC/UN alert levels), with ReliefWeb as automatic fallback. Each feed degrades independently — an unreachable source shows a status line and nothing else changes.
- The watchlist is not prediction. It flags places statistically out of line with their own recent baseline. No reliable earthquake precursor is known to science; most flagged spots will quiet down on their own (METHODOLOGY § 12c).
- Globe geography: Natural Earth 110m coastlines (public domain).