This article is part of The Purple Report September 2026 | Disaster Forecasting and Predictions.
A dangerous place with excellent sensors can look more alarming than a dangerous place with almost no sensors.
That is one of the most important biases in global disaster forecasting.
At Campi Flegrei, scientists can watch earthquakes, ground deformation, gas and bradyseism in detail.
At Reykjanes, satellite and ground observations can show magma accumulation.
Japan, Taiwan and the United States publish sophisticated seismic-hazard products.
Elsewhere, a glacier, slope, river or coastal community may have much thinner observation coverage.
Silence in the data is not the same thing as safety in the world.
Quick Read
The Purple Forecast treats observability as its own risk dimension.
When good monitoring exists, the system can ask whether the hazard is strengthening or weakening.
When monitoring is weak, the correct output is not automatically a lower watch state.
It may be:
DATA GAP — PHYSICAL SUSCEPTIBILITY IS IMPORTANT, BUT CURRENT STATE IS NOT OBSERVED WELL ENOUGH.
An Early-Warning System Is More Than a Forecast
The UN Early Warnings for All framework makes this explicit.
A working warning system needs several connected capabilities:
- Risk knowledge: know what can happen and who is exposed.
- Detection and forecasting: observe the environment and estimate how the hazard may evolve.
- Warning dissemination: get a clear message to the right people.
- Preparedness and response: make sure people can act on the warning.
A brilliant forecast trapped inside a technical centre is not a complete early-warning system.
A message received by a household with no evacuation route is also incomplete.
WMO/UNDRR — Global Status of Multi-Hazard Early Warning Systems 2025
Global Progress Is Real—and Uneven
The 2025 global status assessment reports that 119 countries, around 60% of countries, report having multi-hazard early-warning systems.
That is substantial progress.
But small-island developing states remain under-covered, and regional capability varies strongly.
WMO’s 2026 Africa climate assessment says only about 40% of African countries report multi-hazard early-warning systems, with preparedness and response gaps particularly important.
This should change the Purple Forecast.
A highly instrumented volcano and an uninstrumented floodplain should not compete on the number of signals available.
The data-poor location needs a higher uncertainty flag.
WMO — State of the Climate in Africa 2025
The Five Observation Gaps
1. Measurement Gap
The physical process is not measured at adequate resolution.
A glacier has no velocity monitoring.
A slope has no deformation monitoring.
A river lacks upstream gauges.
A coast lacks reliable tide and land-motion observations.
2. Temporal Gap
Data exist, but not frequently enough.
A satellite image from two weeks ago may be excellent for structural mapping and useless for a failure evolving over hours.
3. Spatial Gap
A national forecast can hide the one valley, neighbourhood or coast where local conditions are much worse.
Regional haze is not local Singapore air quality.
National rainfall is not the rainfall on one unstable mountain catchment.
4. Institutional Gap
Different organisations hold different parts of the observation.
The weather service sees rainfall.
The geological agency sees the slope.
The road operator sees damage.
The hospital sees injured people.
If those signals cannot be combined quickly enough, the world is observed in pieces.
5. Action and Access Gap
The forecast is accurate, the warning is issued—and the person still cannot act.
No transport.
No mobile signal.
No accessible evacuation route.
No language match.
No safe shelter.
This is not a forecast-verification success merely because the warning reached a server.
Gyirong: A Cross-Border Information Problem
The 2026 Gyirong–Rasuwa disaster demonstrated a specific type of observability problem.
The physical system crosses an international boundary.
Ice, water, sediment and debris do not stop at customs.
But detailed glacier, water-level, weather and hazard information can sit inside different national institutions.
Nepal and China had already discussed glacier/weather hazards before the catastrophe, and the post-event debate has increased attention on formalised cross-border information sharing.
That does not prove a data-sharing failure caused the disaster.
It shows that the warning system itself is a legitimate risk factor.
A transboundary hazard needs a transboundary observation and warning path.
Observation Can Fail During the Disaster
There is another problem.
The same event being observed can destroy the observing system.
Flood removes gauges.
Landslide cuts fibre.
Earthquake interrupts power.
Cyclone destroys mobile towers.
Wildfire damages transmission and monitoring equipment.
The system therefore needs resilient observation:
- satellite observations;
- backup power;
- redundant communication;
- multiple sensor types;
- local human reporting;
- independent data routes.
The Purple Data-Gap Card
When a hotspot cannot be evaluated properly, the report should publish the gap rather than hide it.
A reader-facing card should answer:
- What hazard mechanism is credible?
- What would we ideally observe?
- Which observations exist?
- Which are stale, missing or inaccessible?
- How much uncertainty does the gap create?
- What low-cost observation could reduce uncertainty most?
This prevents the forecast from confusing evidence abundance with hazard severity.
Some of the Best Observations Come from People
High-technology monitoring is powerful.
It should not make local observation invisible.
A road worker sees new cracks.
A farmer notices a spring changing.
A community observes a river suddenly becoming muddy or unusually low.
A volcano observatory receives reports of ash or smell.
Human observations are not automatically reliable.
Neither are instruments automatically meaningful without interpretation.
The strongest system can combine both while preserving provenance and verification.
Early Warning Is a Chain
The warning chain can fail at every join:
physical change → monitoring → data network → forecast → authority → message → person → action → survival
A failure anywhere can break the chain.
The practical lesson is simple: a warning must survive the whole journey from observation to action.
A warning is successful only when useful information survives the whole journey to the person who must act.
What This Project Will Treat as a Forecasting Failure
- a hotspot is omitted because data were sparse;
- high-quality monitoring existed but the relevant data were not checked;
- an upstream signal was available but not linked to downstream exposure;
- a warning existed but could not reach the people who needed it;
- the observation network failed during the event with no redundancy;
- a cross-border hazard was modelled independently on each side with no useful information-sharing path;
- the report interpreted “no data” as “no hazard”.
The 2026–2036 Goal
The forecasting project should improve not only by discovering better models.
It should also discover where the world needs better observation.
That means a useful forecast output can sometimes be:
We cannot responsibly raise or lower this hotspot because the decisive observation does not yet exist. Here is the observation, dataset or communication step that would reduce the uncertainty.
That is not forecasting failure.
That is an honest map of where forecasting capability itself is incomplete.