A novel approach to large-ensemble modelling: the time-slice Large Ensemble

Large-ensemble modelling has become an increasingly popular approach to study the climatic response to external forcing. The idea of a large ensemble is to generate different realizations of a forced climate to explicitly reproduce the system’s internal variability. With these large datasets it is not only possible to quantify and statistically test changes in the mean climate, but also changes in climate variability and subsequent changes in extremes. Typically, the approach to generate a large ensemble set is to force the model with a transient forcing and start the different simulations from slightly different initial conditions. However, this is expensive due to the high computational demand of full-complexity GCMs or ESMs.

Here we propose a large-ensemble design that generates a multitude of years to describe the climate states of interest, while being more economical regarding computational resources: a time-slice Large Ensemble. The core of the concept is to generate multiple time slices rather than long transient simulations. The time slices represent the present-day climate and a future warmer climate. These are segments of, for example, 10-years; too short to show significant climate change. Using stochastic physics, we add a randomizing component to the simulations. This allows us to branch multiple simulations from one set of initial conditions.

We present the advantages and limitations of this design and we quantify the underlying assumptions. Further, we demonstrate examples of analyses from earlier work for which this type of large ensemble is well (or better) suited, in particular for studying future extreme events and finding analogues of observed extreme events. Finally, we present ongoing work on the generation and analysis of a new time-slice large-ensemble dataset with EC-Earth v3. The experimental set-up is to branch off from 16 full historical and SSP2-4.5 simulations to represent the present-day climate and a future +2K climate.