This is the public monthly scorecard for The Purple Report September 2026 | Disaster Forecasting and Predictions.
September 2026 is the baseline.
We do not rewrite it.
Every future month has to answer to what was actually written at the beginning.
The Purple Forecast is not a collection of predictions. It is a long-running test of whether forecasts improve when later outcomes are recorded and used to correct future assessments.
Baseline — 1 September 2026
| Total hotspot systems | 50 |
|---|---|
| Earthquake / tsunami | 15 |
| Volcano / lahar | 10 |
| Glacier / GLOF / landslide | 8 |
| Flood / cyclone / coastal | 10 |
| Heat / drought / wildfire | 7 |
| HIGH-CONCERN | 8 |
| ELEVATED | 14 |
| WATCH | 28 |
These three watch labels are not probabilities and are not official emergency alerts.
They describe the Purple Report’s public assessment of the evidence available at the time, including the physical hazard, changing conditions, possible consequences, preparedness and uncertainty. They are not official alert levels or event probabilities.
The September 2026 HIGH-CONCERN Set
- Campi Flegrei / Naples — live bradyseismic unrest and enormous urban exposure.
- Svartsengi–Sundhnúkur / Grindavík — continuing uplift/magma accumulation and short possible warning windows.
- Mayon — PHIVOLCS Alert Level 2 plus tropical rainfall/lahar coupling.
- Kanlaon — PHIVOLCS Alert Level 2 inside a densely inhabited island.
- Popocatépetl — Yellow Phase 2 with ongoing volcanic activity.
- Gyirong–Rasuwa / Shigatse–Nepal corridor — immediate post-catastrophe secondary hazard and reconstruction.
- Nepal Himalayan hydropower corridors — recent failure demonstrates tunnel, worker, power and access cascade.
- Iberia and southwest France — exceptional 2026 heat/dryness and wildfire conditions entering the baseline.
The first October scorecard must show whether each remains HIGH-CONCERN, falls, rises for a different reason or records an event.
Monthly State Changes
| Status | Definition |
|---|---|
| STRONGER | New evidence raises concern or expected consequence. |
| WEAKER | Unrest or loading falls, or new evidence lowers concern. |
| UNCHANGED | No material state change. |
| RESOLVED | A temporary elevated state returns to its structural background. |
| EVENT OCCURRED | A material event occurs inside the defined hazard system. |
| FALSE POSITIVE | A defined time-bounded elevated watch expires without the expected event. |
| MISSED EVENT | A material event occurs without the forecast having identified the relevant system appropriately. |
| WRONG MECHANISM | Place or system relevance was correct but the trigger pathway differed. |
| DATA GAP | Critical observation was missing or inaccessible. |
| SUCCESSFUL PREVENTION | Hazard or warning was justified but preparation materially reduced the expected cascade. |
Four Different Things Must Be Scored Separately
A disaster forecast can be right in one layer and wrong in another.
1. Hazard State
Did the physical event or defined hazard state occur?
2. Timing / Lead Window
Was the concern raised early enough to be useful and within the declared forecast window?
3. Mechanism
Did the forecast correctly identify why the event became dangerous?
4. Consequence / Cascade
Did the expected human, infrastructure and systemic effects actually occur?
This prevents one lucky location match from being called a perfect prediction.
A Worked Example: Volcano Unrest Without Eruption
Suppose a volcano remains HIGH-CONCERN for three months because deformation, seismicity and official alert status remain elevated.
No eruption follows.
Was that forecast wrong?
Not necessarily.
If the public statement was “eruption will occur”, yes.
If the statement was “live unrest justifies closer observation and readiness”, the observed unrest may validate the state even if it later resolves without eruption.
The scorecard must therefore compare the outcome against the actual claim, not the dramatic event readers may have imagined.
A Worked Example: Seasonal Cyclone Outlook
Suppose a basin is forecast to have a below-normal cyclone season.
One severe hurricane still strikes a city.
The landfall does not automatically falsify the basin-season forecast because basin activity and local landfall are different forecast objects.
But a local preparedness system that interpreted “below normal” as “no landfall risk” has made an interpretation error.
The scorecard should record both.
A Worked Example: Earthquake Structural Watch
Nankai, Cascadia and Alpine Fault can remain on the structural watch for years without an earthquake.
That is not a stream of false positives because the claim is not “earthquake this month”.
The claim is that these systems justify long-term preparedness based on official hazard/probability evidence.
The scorecard keeps each authority’s published probability horizon intact and reviews the Purple Report’s public communication separately:
- Did we preserve the correct probability horizon?
- Did we avoid false countdown language?
- Did new science update the watch?
- Did preparedness and exposure materially change?
Forecast Verification Metrics
Where a hazard produces enough comparable forecasts and observations, the project can use standard verification concepts such as:
- Probability of detection: what fraction of defined events were caught?
- False-alarm ratio: what fraction of declared events did not occur?
- Bias: does the system systematically over- or under-forecast?
- Reliability: do forecast probabilities match observed frequencies?
- Resolution: can the system distinguish genuinely different risk states?
- Lead time: how early did useful information appear?
These measures come from operational forecast-verification practice.
They must not be applied mechanically across incomparable hazard classes.
WMO — Guidelines for Streamflow Forecast Verification
The Baseline Must Be Immutable
This is the central integrity rule.
If an event occurs in 2028, we do not edit the September 2026 article so that the old wording magically becomes more accurate.
We attach a new dated record:
- the original published state;
- the later evidence;
- the observed outcome;
- the public review conclusion;
- any dated clarification needed for readers.
This turns the forecast into a ledger rather than a mutable essay.
Interim Updates vs Monthly Updates
The formal scorecard is monthly.
An interim update is justified when there is a material state change, such as:
- official volcano alert level changes;
- major cyclone or flood threat enters a defined hotspot;
- large measured slope/glacier deformation appears;
- significant earthquake changes aftershock/secondary hazard state;
- critical infrastructure is damaged;
- an official warning is issued or lifted;
- a major new data source materially changes uncertainty.
Interim updates should never erase the month’s starting state.
The October 2026 Template
| System | September state | October state | Change | New evidence | Outcome |
|---|---|---|---|---|---|
| Campi Flegrei / Naples | HIGH-CONCERN | — | — | — | — |
| Svartsengi–Sundhnúkur / Grindavík | HIGH-CONCERN | — | — | — | — |
| Mayon | HIGH-CONCERN | — | — | — | — |
| Kanlaon | HIGH-CONCERN | — | — | — | — |
| Popocatépetl | HIGH-CONCERN | — | — | — | — |
| Gyirong–Rasuwa / Shigatse–Nepal corridor | HIGH-CONCERN | — | — | — | — |
| Nepal Himalayan hydropower corridors | HIGH-CONCERN | — | — | — | — |
| Iberia and southwest France | HIGH-CONCERN | — | — | — | — |
The full monthly scorecard will retain all 50 systems, not only the eight shown here.
The Long-Horizon Scorecard
At the end of each year, the project should answer:
- How many material disasters occurred inside the 50 baseline systems?
- How many important disasters occurred outside the baseline?
- Which hotspot states were persistently over-sensitive?
- Which mechanisms produced the most false alarms?
- Which data gaps caused misses?
- Which resilience measures repeatedly reduced cascade severity?
- Did the system become better calibrated compared with the previous year?
By 2036, the September 2026 baseline should still be readable beside the final ten-year scorecard.
That is the proof that the project allowed the future to answer back.
The Final Score Is Not Accuracy
The project is ultimately trying to improve public decisions and preparedness.
A forecast that is statistically impressive but does not improve observation, preparedness, evacuation, redundancy or repair is incomplete.
So the scorecard keeps one final question beside all the numbers:
Did knowing earlier change anything useful before the next failure?
That is where forecasting becomes prevention.