Predictions of climate from seasonal to decadal time scales and their applications will be discussed in this session. With a time horizon from a few months up to thirty years, such predictions are of major importance to society, and improving them presents an interesting scientific challenge. This session aims to embrace advances in our understanding of the origins of seasonal to decadal predictability, as well as in improving the respective forecast skill and making the most of this information by building and testing new applications and climate services.

The session will cover dynamical as well as statistical predictions, and their combination. It will investigate predictions of various climate phenomena, including extremes, from global to regional scales, and from seasonal to multidecadal time scales ("seamless prediction"). Physical processes relevant to long-term predictability sources (e.g. ocean, cryosphere, or land) as well as predicting large-scale atmospheric circulation anomalies associated to teleconnections will be discussed. Also, the time-dependence of the predictive skill (hindcast period) will be investigated. Analysis of predictions in a multi-model framework, and ensemble forecast initialization and generation, including innovative ensemble approaches to minimize initialization shocks, will be another focus of the session. The session will pay particular attention to innovative methods of quality assessment and verification of climate predictions, including extreme-weather frequencies, post-processing of climate hindcasts and forecasts, and quantification and interpretation of model uncertainty. We particularly invite contributions presenting the use of seasonal-to-decadal predictions for risk assessment, adaptation and further applications.