Methodology

How the index works

Every number on this site is reassessed every 24 hours: seven threats read official live data feeds, and all 27 baselines are re-scored each sync by an AI analyst briefed with the fresh sensor readings plus headline digests from official sources (CISA, WHO, USGS). This page documents the exact formulas, sources and assumptions. It is an awareness tool, not a forecast.

Structure

Threat scores, step by step

  1. 01Each of the 27 tracked threats starts with a baseline severity from 0–100, set from expert assessment (see sources below). Every 24 hours an AI analyst pass re-scores all 27 baselines against current world conditions and the freshly synced sensor readings — nothing on the board is left stale.
  2. 02Where a public live feed exists — 7 of the 27 threats — the server converts the raw measurement into a 0–100 live score using the formulas documented on this page.
  3. 03The two are blended: 45% baseline, 55% live signal. Threats without a live feed show their baseline unchanged.
  4. 04Scores are bucketed into four risk bands: Watch, Elevated, Severe, Critical.
  5. 05The Doomsday Clock maps the average of all 27 threat scores onto minutes remaining before midnight.

Formula

The blending calculation

score = baseline × 0.45 + liveScore × 0.55 (when a live feed exists) score = baseline (all other threats) overall index = mean(all 27 threat scores) minutes to midnight = max(0.2, (100 − overall) × 0.15)

The 45/55 weighting deliberately lets the live data lead: when the ground shakes or the magnetometer spikes, the score moves with it, while the baseline keeps a quiet feed from ever reading as zero risk. The clock mapping takes the average index — an index of 100 is midnight — and spreads the remainder over 15 minutes, the same convention as the Bulletin of the Atomic Scientists' Doomsday Clock.

Metaphor

What 'minutes to midnight' means

The clock is a shorthand, not a timer. Midnight represents the end of the world as we know it — the point where civilisation's basic systems stop functioning for most people. The minutes shown are derived from the current average threat score: the higher the index, the closer the hands move to midnight. An index of 0 would read as 15 minutes to midnight; an index of 100 is midnight itself. It does not predict a date — it summarises how stressed the global and local risk posture looks right now.

Bands

What the labels mean

Critical75 – 100
Severe55 – 74
Elevated35 – 54
Watch0 – 34

Bands are descriptive, not predictive. "Severe" means the threat posture is historically elevated — not that an event is imminent on any given day.

Live feeds

Each live signal, exactly as calculated

Six threats read public instrument data. Every formula below is the literal calculation in the code; "clamp" pins the result between 0 and 100.

Global Seismicity

USGS Earthquake Hazards Program
Measures
All M4.5+ earthquakes worldwide, trailing 24h
Formula
score = clamp(20 + count×3 + (maxMag − 4.5)×12)
In plain terms
Each significant quake adds 3 points; the day's largest event adds 12 points per magnitude unit above 4.5.

NZ Earthquakes

GeoNet (GNS Science / MBIE)
Measures
Felt quakes (MMI ≥ 3) in New Zealand, trailing 24h
Formula
score = clamp(38 + count×5 + (maxMag − 3)×8)
In plain terms
Baseline of 38 reflecting NZ's permanent seismic exposure; each felt quake adds 5, and the largest event adds 8 per magnitude unit above 3.

NZ Volcanic Activity

GeoNet Volcanic Alert Level (VAL) system
Measures
Highest VAL across 12 monitored volcanic systems
Formula
score = [20, 42, 58, 78, 90, 100][VAL]
In plain terms
A direct lookup from the official alert level: VAL 0 (quiet) scores 20, a major eruption (VAL 4–5) scores 90–100.

Coronal Mass Ejection

NOAA Space Weather Prediction Center
Measures
Planetary K-index (Kp), latest 3-hour reading
Formula
score = clamp(24 + Kp×9)
In plain terms
Kp 0–4 (quiet to unsettled) keeps the score near baseline; a G4 storm at Kp 8 scores ~96. Labels follow NOAA's G-scale.

Asteroid Impact

NASA Near-Earth Object Web Service (NeoWs)
Measures
Close approaches today, hazardous flags, closest miss distance
Formula
score = clamp(12 + hazardous×14 + (closest < 5 lunar distances ? 25 : 0))
In plain terms
Each potentially hazardous asteroid passing today adds 14; anything threading inside 5 lunar distances adds a further 25.

Solar Flare Activity

NOAA GOES primary X-ray sensor
Measures
Peak 0.1–0.8nm X-ray flux, trailing 24h
Formula
score = 24 if B-class · 40 if C · 62 if M · 88 if X
In plain terms
Stepped by flare class: each decade of flux (C → M → X) roughly doubles severity. X-class means HF radio and GPS blackouts.

Severe Weather

NASA EONET (Earth Observatory Natural Event Tracker)
Measures
Active severe storm systems tracked worldwide, right now
Formula
score = clamp(34 + activeStorms×6)
In plain terms
Every currently active severe storm system adds 6 points on top of a climate-loaded baseline of 34.

Baselines

The other 20 threats

Most existential risks have no instrument to read — there is no sensor for "how close is nuclear war". These threats hold a baseline informed by current expert assessment, reassessed every 24 hours by an AI analyst pass on the server (see Sync cadence below). If that pass is ever unavailable the scores fall back to the reference baselines below:

Baselines are informed by public expert sources including the Bulletin of the Atomic Scientists, the Global Challenges Foundation's Global Catastrophic Risks report, the Global Preparedness Monitoring Board, NZ's National Emergency Management Agency, and GNS Science fault studies (see references).

Refresh

Sync cadence and cost

Once per 24-hour UTC window (the same fixed schedule for every visitor) the server does a full reassessment: every live feed is re-synced, then a single AI analyst call re-scores all 27 threat baselines given the current date, world conditions, the fresh sensor readings and headline digests pulled from official sources at the same moment (CISA cybersecurity advisories, WHO disease outbreak news, and USGS significant-event reports) — so every threat, sensor-backed or not, is reassessed against something authoritative at each sync. That same pass writes a short situation note per threat, shown at the top of each full briefing, so the briefings update with the sync when something has actually changed. The result is served from memory for every visitor after the first — at most two upstream pulls per feed and two AI call per day. Browsers cache the result for the same 24 hours and never refetch on tab focus. If the AI pass fails, baselines fall back to the static expert values, so the board never goes blank. The whole site is static apart from those reads, so running costs are near zero.

References

Sources and further reading

Get prepared

Want to do something useful with this information?

Preparing.nz is a practical guide for New Zealanders who want to be ready for the unexpected — earthquakes, storms, power cuts, pandemics and more.

Visit preparing.nz