Documentation

An introduction to decadal predictions

The evolution of the climate systems in the near future depends on changes in atmospheric composition and other external forcings as well as in the slow naturally generated internal climate variability. Until very recently, the only sources of future climate information that were available to interested users were seasonal predictions and climate projections. The former provide a future outlook of the earth’s climate system for a period ranging from 1 to 18 months into the future while the latter covers a continuous temporal range from the past century to the end of this century (or beyond) but with no relationship with the contemporaneous internal climate variability. At the seasonal timescale, the climate evolution mainly depends on the internally generated variabilities of the climate system and less on the changes in the externally forced components that occur over the period of forecast. On the other hand, climate projections are solely driven by changes in external forcings without constraints on the internal variability.

As an alternative to these types of climate information, recently developed decadal climate prediction systems attempt to fill the gap that exists between these two timescales (i.e. from a year up to a decade), where the evolution of the climate is impacted by both internally generated variability and externally forced components. Decadal prediction is then, in simple terms, the extension of seasonal forecasts wherein climate models are initialized by introducing observation-based data and run for a decade or so under the influence of contemporaneous changing external forcings (for instance, with rising greenhouse-gas concentration), as in climate projection. Predicting the variations in climate at this timescale is considered one of the most challenging problems faced by the climate forecasting community due to the relatively weak constraints that can be applied on the internal variability and the relatively weak anthropogenic external forcings at this timescale.

The first attempt at producing decadal climate predictions was made in the framework of the EU-funded ENSEMBLES project (2004-2009). Several studies investigated the capacity of different decadal prediction systems to accurately “predict” past climate variability in retrospective experiments called hindcasts (as opposed to the actual future predictions, called forecasts). Since then, the field of decadal prediction has grown significantly, in part due to the large socio-economic interest generated by these predictions. Clear examples of the growing interest in this field of research are the inclusion of decadal predictions in the recent phases of the Coupled Model Intercomparison Project (CMIP5 and CMIP6), the production and publication of real-time decadal predictions and a growing body of literature on potential applications of these forecasts.

Decadal prediction lies at the boundary between seasonal forecasting and climate change projections. While seasonal forecasting is considered an initial value problem (the evolution of the atmosphere-ocean system is largely determined by the initial condition) and climate projections a boundary value problem (the system evolution depends on the external forcing and formulation of boundary condition), decadal prediction is considered a joint initial-boundary value problem (Figure 1), with both internal processes and external forcings playing a role in decadal climate variations.

At the decadal timescale, the observed climate variability can be understood as the superimposition of an anthropogenically-driven trend on natural fluctuations. This simple view assumes that there is no interaction between the trend and the natural fluctuations, which might not be necessarily the case. While the trend is driven by changes in anthropogenic emissions, the natural fluctuations are generated internally by the interactions of the different components of the climate system (atmosphere, ocean and sea ice) or externally by other factors such as volcanic eruptions and solar activity. Provided that these different factors operate on a sufficiently long timescale (multiannual or longer) and their influence can be estimated with a sufficient level of accuracy, they can potentially be a source of skill in a decadal prediction context.

For a more simplified introduction to decadal predictions and the potential uses of such predictions in the agriculture sector, please see the infosheet "A brief introduction to decadal predictions".

BSC Operational Decadal Prediction System

BSC’s system is based on the Earth System Model EC-Earth version 3.3, jointly developed by a European consortium that brings together 27 research institutes from 10 European countries. EC-Earth3.3 is the version that was prepared to contribute to the 6th phase of the Climate Model Intercomparison Project (CMIP6), and has been used under different configurations (e.g. only the physical core, with interactive atmospheric chemistry, with dynamic vegetation, with interactive carbon cycle,...) and resolutions to perform historical simulations, climate change projections, decadal predictions and process-based studies.
 
The BSC-ES has been a member of the EC-Earth consortium since 2009, and is both a user and a developer of the coupled global climate model EC-Earth (http://www.ec-earth.org). The physical core of EC-Earth combines the atmospheric model IFS, the ocean component NEMO and the sea-ice model LIM, assembled via the OASIS coupler. 

The specific technical details on BSC’s operational decadal prediction system are provided in the table below: