Forecasting more than one future.

GenCast used machine learning to generate ensembles of possible weather outcomes up to 15 days ahead.

Demonstrated · source published 2024-12-04

The human problem

A single forecast does not communicate the range of weather that could occur.

Ensemble forecasting uses multiple simulations to represent uncertainty.

A diffusion model generated probabilistic global forecasts.

The research reported stronger skill than the comparison operational ensemble on the evaluated targets.

Better uncertainty estimates could support weather-dependent decisions.

Retrospective evaluation is not proof of operational benefit in every region or extreme event.

Test operational reliability and local decision usefulness.

Source history & evidence assessment

Legacy source check; no named human reviewer is recorded in this projection.

Maturity describes the tested or operational setting. Confidence describes support for the particular claim; one does not determine the other.

Institutions: Google DeepMind

Editorial connections between distinct settings and results; these links do not imply replication.

No corrections recorded.