What a 70% Chance and a Hurricane Cone Actually Mean — Reading Forecast Numbers Without Fooling Yourself

Direct answer: forecast numbers describe uncertainty, and almost every common misreading comes from treating them as descriptions of a thing instead. The hurricane cone shows where the centre of the storm is likely to go, not how wide the storm is — serious impacts routinely occur far outside it. A 70% chance of rain is the chance that measurable rain falls at a given point, not how much, how long, or over how much of the map. And an event that was forecast at 30% and did not happen is not a failed forecast; it is what 30% means.

The cone is the single most dangerous graphic in weather

A tropical cyclone forecast cone is built from the historical accuracy of past forecasts. It is drawn so that the storm's centre stays inside it roughly two thirds of the time. Two things follow, and both are routinely missed:

• It says nothing about the size of the storm. The cone is about a point — the centre. A storm's damaging winds, rain and surge extend far beyond that point, often for hundreds of kilometres. Being outside the cone is not being outside the storm.
• Roughly one time in three, the centre goes outside it anyway. That is not a flaw. A cone drawn to contain the centre every time would be so wide it would tell you nothing.

The most consequential impacts frequently arrive well away from the track, and for coastal areas the killer is usually not the wind at the centre at all — it is storm surge, which depends on the shape of the coast and the angle of approach rather than on how close the line passes.

Category is a separate axis again: it describes intensity, not size and not reach, which is covered in why a Category 1 still kills people.

What a percentage chance of rain is actually saying

A probability of precipitation is a statement about one point in the forecast area over the forecast period: the chance that a measurable amount of rain falls there. It is not a claim about intensity and not a claim about duration.

The three usual misreadings:

• “It will rain over 70% of the area.” No. Coverage and probability are different things, and services that mean coverage usually say so in words like “scattered”.
• “It will rain for 70% of the time.” No. Duration is not in the number at all.
• “70% means heavy.” No. A high-confidence forecast of light drizzle is a high percentage. Amount is a separate forecast.

Definitions differ slightly between national services, which is worth knowing before comparing two apps that disagree — a different flavour of the problem described in why two agencies report the same disaster differently.

Where the uncertainty comes from

Modern forecasting does not produce one future. It runs the atmosphere forward many times from slightly different starting conditions, because the starting conditions are themselves measurements with error, and small differences grow. What comes back is a spread of outcomes.

When those runs agree, confidence is high. When they disagree, it is genuinely uncertain — and a forecast that reports the disagreement is doing its job better than one that hides it behind a single confident-looking line. The single line on most weather apps is a summary of that spread, and the spread is the part that was thrown away.

The hardest idea: a forecast can be right and still not happen

If a forecaster says 30% and the thing does not happen, most people file it as a wrong forecast. But 30% is a claim that in roughly three cases out of ten it will happen — so it not happening is the expected outcome, and it is not evidence of anything on its own.

Probabilistic forecasts can only be judged over many events: if everything forecast at 30% happens about three times in ten, the forecasts are well calibrated and working exactly as intended. Judging one forecast by one outcome is like judging a die by one roll.

This matters far beyond pedantry, because it is how trust in warnings is lost. A community evacuated for a storm that turned is not a community that was misled — it is a community for which the low-probability branch did not occur that time. Treating it as a false alarm makes the next evacuation harder, and the next one may be the branch that does occur.

Low probability is not the same as ignorable

A 10% chance of something catastrophic is not a small matter, because the right way to weigh a forecast is probability against consequence. Ten percent chance of light rain is a reason to bring a jacket. Ten percent chance of two metres of water in your street is a reason to move.

This is the single most useful habit in reading any forecast: never read the number alone. Read it against what happens if it comes true.

How to read a forecast without fooling yourself

1. For a cyclone, ignore the line and read the impact maps. Wind field, rainfall and surge products tell you what reaches you; the track tells you where a point is going.
2. Being outside the cone is not safety. If forecasters have anything to say about your area, it will be in a watch or warning for your area — which is what those words are for.
3. Weigh probability against consequence, every time.
4. Expect the forecast to change, and prefer the newest one. Changing is what a forecast being updated with new observations looks like.
5. Take your numbers from the agency responsible for your area. Every one of our country pages links its national meteorological service.
6. Do not judge a forecaster on one event. Calibration is a property of many forecasts, never of one.

What this site does with forecast information

• We do not issue forecasts. We aggregate what agencies publish and name the agency on every record.
• We do not turn a probability into a yes or a no. Compressing an uncertain forecast into a confident sentence is the failure mode this whole article is about.
• Where our app reports on forecasts, it reports agreement rather than values — whether the models concur about the band of outcomes, not a single number dressed up as certainty.
• We say nothing at all when there is nothing to say. A quiet forecast week produces silence, not a manufactured update.

The honest summary

Forecast graphics are compressions of uncertainty, and every misreading in this article comes from decompressing them into something too solid — the cone into a storm, a percentage into a promise, one outcome into a verdict on the forecaster. Read the cone as a line about a point, the percentage as a chance at a place, and any single result as one roll of a loaded die whose loading you were told about in advance.

This article explains how forecast products are constructed in general. This site does not issue forecasts; for the forecast where you live, and for what to do about it, use your national meteorological service.

Frequently asked questions

Does the hurricane cone show the size of the storm?

No, and this is the most consequential misreading in hazard communication. The cone shows where the centre of the storm is likely to travel, and it is drawn so the centre stays inside it about two thirds of the time. The storm itself — damaging wind, rain and surge — extends far beyond that centre, often for hundreds of kilometres. Being outside the cone is not being outside the storm.

Am I safe if I am outside the cone?

Not necessarily. The cone is about the track of a point, not the reach of the hazard, and roughly one time in three the centre goes outside it anyway. What matters for your area is whether a watch or warning has been issued for your area, and what the wind, rainfall and storm surge products say. For coasts the greatest danger is often surge, which depends on the shape of the coastline rather than how close the track passes.

What does a 70% chance of rain actually mean?

It is the chance that a measurable amount of rain falls at a given point in the forecast area during the forecast period. It is not the share of the area that will get rain, not the share of the time it will rain, and not a statement about how heavy it will be. A confident forecast of light drizzle produces a high percentage.

If a forecast said 30% and nothing happened, was it wrong?

No. A 30% forecast says the event happens in roughly three cases out of ten, so it not happening is the expected outcome. Probabilistic forecasts can only be judged across many events: if things forecast at 30% happen about three times in ten, the forecasts are well calibrated. Judging one forecast by one outcome is like judging a die by a single roll.

Why do forecasts keep changing?

Because they are updated as new observations arrive. Modern forecasting runs the atmosphere forward many times from slightly different starting conditions, since those conditions are measurements with error, and small differences grow. When the runs agree confidence is high; when they disagree the situation is genuinely uncertain. A changing forecast is the system working, and the newest one is the one to use.

Is a 10% chance small enough to ignore?

It depends entirely on the consequence. Probability has to be weighed against what happens if it comes true. A 10% chance of light rain is a reason to carry a jacket; a 10% chance of two metres of water in your street is a reason to move. Never read the number by itself.

Was an evacuation for a storm that turned a false alarm?

No, and treating it as one is how trust in warnings erodes. The evacuation was a response to a real probability at the time it was ordered; the low-probability branch simply did not occur on that occasion. Calling it a false alarm makes the next evacuation harder to achieve, and the next one may be the branch that does occur.

Sources

• NOAA National Hurricane Center — About the cone of uncertainty and track forecast graphics
• NOAA / NWS — Probability of precipitation explained
• ECMWF — Ensemble forecasting and forecast uncertainty
• WMO — Guidelines on communicating forecast uncertainty and impact-based forecasting
• NOAA National Hurricane Center — Storm surge and wind field products
• UK Met Office — How forecasts are verified
• UNDRR — Early warning and public understanding of risk information

Found an error? See our corrections policy.