Forecast uncertainty

A forecast is uncertain because future outcomes depend on imperfectly known parameters, initial conditions, future inputs, stochastic events and model structure.

Point and probabilistic forecasts

A point forecast gives one predicted value. A probabilistic forecast describes a distribution of possible outcomes and can report prediction intervals or probabilities of important events.

Forecast horizon

Uncertainty generally grows as the prediction horizon increases because more unknown future events and conditions can affect the system.

Calibration

A well-calibrated probabilistic forecasting procedure should produce uncertainty statements that agree with observed frequencies over repeated comparable predictions.