How does a weather forecast decide if it will rain?
A weather forecast decides if it will rain by giving a chance from 0 to 100 percent. Forecasters measure the air today and run computer models of how it will change. Deciding "yes, bring an umbrella" means picking a line on that chance, and no line is right every day.
In one line: Computer models that use measurements of the air today to work out the weather ahead, giving a chance of rain from 0 to 100 percent.
Last reviewed: 25 October 2026 3 min read
What decision does it make?
For each place and time, the forecast works out how likely rain is. A 40 percent chance of rain means a 40 percent chance that at least a tiny measurable amount of rain falls at that spot. Turning that chance into "yes" or "no" is a second decision, and you can make it yourself.
| Decision | Possible answers |
|---|---|
| Will it rain? | A chance from 0 to 100 percent. You pick the line: at or above it, take an umbrella. |
What does it look at?
- Weather stations. Instruments on the ground measure the air where they stand.
- Weather balloons. Balloons carry instruments high into the sky and measure temperature, moisture and air pressure.
- Satellites. They watch huge areas from space, including remote places far from any station.
- The rules of physics. Equations describe how air, heat and water move, and computer models use them to step the weather forward in time.
How does it learn the rule?
- Measure the air now. Stations, balloons and satellites send in what the air is like right now.
- Run the model. A computer model uses physics equations to work out how that air will change over the next hours and days.
- Run it many times. The measurements are never perfect, so the model runs again from slightly different starts. The share of runs that end in rain helps set the chance of rain.
- Check against what happened. Forecasts are compared with the real weather, so forecasters can see where the models go wrong and improve them.
Where can it go wrong?
- You take an umbrella, and it stays dry. A high chance is not a promise of rain.
- You leave the umbrella, and it rains. A low chance is not a promise of a dry day either.
- Small errors grow. Tiny mistakes in today's measurements get bigger over time, so a forecast for next week is less sure than one for tomorrow.
- No line is perfect. Move your line down and you carry the umbrella on more dry days. Move it up and you get caught in the rain more often.
Try it
Toy example: ten days, each with the chance of rain from the forecast. Move your umbrella line until as few days as possible go wrong. Can you get zero?
Toy example: ten made-up days with the forecast chance of rain and whether it really rained.
| Item | Values | Right bin |
|---|---|---|
| Monday: dark clouds rolling in | Chance of rain (%): 95 | Umbrella |
| Tuesday: storm in the afternoon | Chance of rain (%): 80 | Umbrella |
| Wednesday: showers on and off | Chance of rain (%): 70 | Umbrella |
| Thursday: clouds, but they passed by | Chance of rain (%): 60 | No umbrella |
| Friday: one quick shower | Chance of rain (%): 50 | Umbrella |
| Saturday: gray all day, stayed dry | Chance of rain (%): 40 | No umbrella |
| Sunday: surprise drizzle at dinner | Chance of rain (%): 30 | Umbrella |
| Next Monday: a few clouds | Chance of rain (%): 20 | No umbrella |
| Next Tuesday: sunny and windy | Chance of rain (%): 10 | No umbrella |
| Next Wednesday: clear blue sky | Chance of rain (%): 5 | No umbrella |
Check it
What does a forecast give you about rain?
Answer: A chance from 0 to 100 percent. It gives a chance. You decide where your yes-or-no line goes.
The forecast said 30 percent and it rained. Was the forecast broken?
Answer: Not necessarily, 30 percent means rain can happen. A 30 percent chance means rain is possible, just less likely than not.
Why is a forecast for next week less sure than one for tomorrow?
Answer: Small measuring errors grow over time. Tiny differences at the start grow bigger the further ahead the model looks.
For older readers
This is numerical weather prediction: observations are blended into the model's starting state (data assimilation), then equations of the atmosphere are solved forward in time. Because the atmosphere is chaotic, small errors in the starting state grow, so forecasters run an ensemble of forecasts from slightly different starts. The chance of rain is a probability of precipitation: in the US it is the chance of at least 0.01 inch at a point. Forecasts are checked with verification tools such as reliability diagrams and the Brier score.
Questions
It means there is a 40 percent chance that at least a tiny measurable amount of rain falls at that spot. It does not mean it will rain for 40 percent of the day.
A chance is not a promise. Also, measurements are never perfect and small errors grow, so a shower can form where the model did not expect it.
Forecasters use computer models fed with measurements from stations, balloons and satellites, and they check the results against the real weather.
Sources
- National Weather Service: what is probability of precipitation? (opens another website)
- NASA Earth Observatory: weather forecasting through the ages (opens another website)
- NOAA National Centers for Environmental Prediction: ensemble forecast background (opens another website)
- University atmosphere course: prediction and predictability (lecture notes) (opens another website)