Task 2.6 Added Value of Global Multi-Model-Predictions for User-relevant variables

The advantages of using a multi-model ensemble for decadal prediction has been established in the
literature (e.g. Smith et al., 2012; van Oldenburgh et al., 2012). In recent years, the number of
institutions providing decadal prediction has increased and the quality of the predictions has
improved (Borchert et al., 2020). Previous studies of the decadal prediction ensembles show that
larger ensembles improve the skill of initialised predictions and also the spatial extent with a
predictive variance (e.g. Sospedra-Alfonso and Boer, 2020; Sienz et al., 2016). Thus, a large or at
best a very large ensemble is required to extract the predictable signal and to overcome the partly
small signal-to-noise ratio in climate models (e.g. Smith et al., 2019). Thus, data from decadal prediction from multiple modelling centres – as collected by the WMO Lead Centre of Annual-toDecadal Climate Prediction (ADCP1) will be utilised as a reference to test the robustness and skill of the Coming Decade ICON-based ensemble system, and to provide robust basic (temperature, precipitation, atmospheric circulation) and user-oriented climate information (e.g. Mömken et al.,
2021) on multi-year timescales. In particular, the associated reliability and uncertainty of the Coming Decade prediction system can be assessed, thus overcoming some of the limitations of the singlemodel approach. Given the nature of the data, the resolution will be lower than for the ICON ensemble, which starts at R2B5 (80 km). To achieve the goals of the project, this task will cooperate
with Task 2.2 on the recalibration and verification, and with Task 2.3 on the development of user-specific products.

Contributors

Karlsruhe Institute for Technology
Prof. Dr. Joaquim Pinto
Dr. Patrick Ludwig
Hendrik Feldmann
Dr. Woon Mi Kim

Max Planck Institute for Meteorology

Freie Universität Berlin

Deutscher Wetterdienst
Dr. Andreas Paxian