What we do

Predictability

Current position: Index -> What we do -> Task Teams -> Predictability
Science and methods

Predictability Task Team

The Predictability Task Team focuses on forecast skill, uncertainty, and the sources of predictability in impact-relevant sub-seasonal to seasonal forecasts. Its work helps identify when, where, and under what conditions forecast information can support decision-making.

Regime-dependent skill

Understanding how forecast skill varies with large-scale or local flow conditions.

Forecast uncertainty

Quantifying uncertainty so that it can be communicated and used in decision-making.

Reforecast datasets

Supporting analysis through extended sub-seasonal reforecasts and experiments.

Calibration methods

Developing post-processing approaches that account for regime-dependent skill.

Scientific focus

01

Identify sources of predictability

S2S forecast skill can arise from multiple interacting drivers, including large-scale modes of climate variability, weather regimes, and local flow characteristics. Understanding these sources is essential for interpreting when forecasts are likely to be informative.

02

Quantify regime-dependent skill and uncertainty

Forecast skill and uncertainty can vary across different regimes. Quantifying these variations can help users understand the confidence that should be placed in forecast information under different conditions.

03

Use dynamical and data-driven approaches

The team considers both traditional dynamical forecasting systems and newer data-driven methods, including AI and machine-learning approaches for sub-seasonal timescales.

04

Support post-processing and calibration

Better understanding of forecast skill and uncertainty can inform calibration and post-processing methods, helping forecasts become more relevant for decision contexts.

Connection with applications

From predictability research to decision support

The Predictability Task Team provides scientific understanding and methods that can support the application-sector Task Teams. Its work helps connect forecast reliability, uncertainty, and calibration with real-world decision needs.

Evidence for forecast use

Understanding when forecasts are more or less skilful helps identify where forecast information can be used responsibly in applications.

Methods for uncertainty

Quantifying uncertainty supports clearer communication of forecast confidence and helps users incorporate probabilistic information into decisions.

Support for sector-focused work

Predictability research can support agriculture, health, disaster risk reduction, energy, and other application areas by improving how skill, uncertainty, and calibration are handled.