About & data
Every number on this map, where it came from, and what it cannot tell you. Written in the same register as the flood module's About page, and for the same reason: a hazard map that will not state its own limits is asking to be trusted further than it has earned.
Why two catalogues and not one
The obvious way to build this is to take the USGS feed and stop. That is what the site did first, and it was quietly wrong for the people most likely to read it. The USGS global catalogue is complete to roughly magnitude 4.5; below that its coverage is dense across the United States and thin nearly everywhere else. Queried for the box 21–30°N, 87–98°E across eleven days in August 2026, it returned zero events. EMSC returned 28 over the same box and the same window, including a M 3.0 in Assam, a M 2.8 in Meghalaya and a M 3.4 in Manipur.
So the map reads both. EMSC is a federation that ingests national catalogues — India’s NCS, Indonesia’s BMKG — which is how the moderate and small events a reader in Guwahati actually feels reach this page at all. GDACS is read as well for its alert levels and because it occasionally carries an event before USGS has reviewed it.
Where every number comes from
USGS — the all_week summary feed, filtered at M 2.5. Public domain. It is the primary for earthquakes and wins wherever it has the event, because it writes the most precise place (“68 km NNW of Ende, Indonesia” against a bare country name) and carries a reviewed depth.
EMSC — the seismicportal FDSN service, in two queries: worldwide at M 4.0 and above, and South and South-East Asia at M 2.5 and above. That asymmetry is deliberate and you should know about it: a M 3 in Assam is listed and a M 3 in Chile is not. This site is built and read in Assam, the flood module is Assam-only for the same reason, and a single global query at M 2.5 exceeds the service’s row cap. It is an editorial choice, not uniform global sensitivity.
GDACS — the EU/UN multi-hazard RSS, filtered to earthquakes. Basemap — CARTO dark tiles over OpenStreetMap data, © OpenStreetMap contributors under ODbL. No API key is used anywhere on this site, and no upstream service is ever called from your browser.
Where two feeds hold the same earthquake, one record is kept: matched across sources only, within 60 km and 150 seconds with magnitudes agreeing to 0.7 — or within 25 km with magnitudes agreeing to a tenth, which is not a coincidence. Two records from the same feed are never merged, because an aftershock sequence produces genuinely distinct quakes minutes and tens of kilometres apart.
How “critical” is decided
Three colours carry the ranking on the map, and until recently the page never said what earned each one. These are the thresholds the code actually applies — this list is generated from the same constants the feeds call, so it cannot drift from the map the way a hand-typed legend can:
They are this site’s own bands and not an official scale. There is no agreed magnitude-to-severity mapping to defer to. USGS publishes none. GDACS’s Green/Orange/Red is a different and in some ways better question — an impact model weighing population and expected shaking rather than magnitude alone — and where GDACS is the only source for an event, its alert level is kept rather than recomputed from the number. Everywhere else the cut is at M 4.5 and M 6.0, and that is a choice made here.
Why the magnitude carries two words after it
Every record now prints the scale that produced the number and, where the agency says so, whether a person has checked it — mww is a moment tensor, mb a body-wave reading, ml a local Richter-style estimate, and reviewed means a seismologist has signed it off where automatic means the software has not been checked yet and the figure may still move. Two scales can differ by half a unit on the same earthquake. A bare “M 5.7” hides that; both fields were in the feeds from the first day and this site printed neither until August 2026.
Where two or three agencies located the same event, all of them are named on the record. That agreement is the most useful thing three independent feeds can offer, and the deduplication step used to pick one and throw the others away.
What it cannot tell you
Nothing here predicts an earthquake. No one can. Everything on this map has already happened.
Absence is not safety. An empty area may be an area with no seismometers rather than no earthquakes, and below the magnitude floors above there is simply no coverage at all.
Marker size is not a measurement. Radius rises with magnitude so the map can be read at a glance, but magnitude is logarithmic and a marker scaled by true energy would be invisible at the low end. Read the number, not the circle.
“What may follow” is a rule of thumb, not a forecast. Every line says can or may, never will. The tsunami rule decides “coastal” from the wording of a place name and a list of archipelagic states, and the landslide rule decides “steep terrain” from bounding boxes drawn round the major mountain belts. Both are coarse and both will sometimes be wrong. Official tsunami warnings come from the regional warning centre, never from this page.
Times are origin times in UTC. Where a feed publishes only its own publication stamp, the origin time is recovered from the event title — GDACS runs about twenty minutes behind — and your local clock is shown beside it.