Forecast uncertainty is the range of ways the weather could reasonably turn out, given what forecasters know now. It comes from imperfect starting data, imperfect models and the atmosphere's sensitivity to small differences, and it generally grows the further ahead a forecast reaches.
Where it comes from
Observations never capture the whole atmosphere, so every model starts from a state that contains error. Model equations also leave out or simplify some processes.
The atmosphere is a chaotic system, so small errors in the starting state grow as a forecast runs. NOAA's ensemble documentation notes this would happen even with a perfect model.
How fast those errors grow depends on the circulation pattern, the season and the region. Some patterns stay predictable for many days while others lose detail much sooner.
How forecasters measure it
Ensembles run a model many times from slightly different starting points. How far the members spread apart shows where and when the forecast is less certain, and counting members lets any forecast quantity be expressed as a probability.
On spaghetti plots, members typically overlap early and diverge later, which is the growth of uncertainty with time made visible.
How it is communicated
The most familiar example is the chance of precipitation. A 40 percent chance means a 40 percent probability of at least 0.01 inch of rain at the forecast point.
Forecasters also use ensembles to describe a range of scenarios and the most likely one. That lets them still give a specific forecast while being honest about what could change.
At storm scale
For thunderstorms, much of the uncertainty is about exactly where and when storms form, not only whether they do. A forecast can be confident in the environment and still uncertain about which county gets the strongest storm.
That is why storm-day planning leans on probabilities and short-range updates, then on observations and radar as the event begins.
