WCRP Decadal Prediction Cross-cutting Activity

At JSC-28, a proposal was made for a cross cutting activity to advance the science of decadal prediction based on internationally-coordinated multi-model experimentation to gauge the overall level of predictability arising from having both different initial conditions and different greenhouse gas forcings. This activity, which will cut across all of WCRP's core projects, now is being led by a joint WCRP/CLIVAR/WGCM/WGSIP sub group, the members of which are: Tim Stockdale, Gabi Hegerl, Jerry Meehl, James Murphy, Ron Stouffer, Marco Giorgetta, Masihide Kimoto, Tim Palmer, Wilco Hazeleger, Detlef Stammer, Ben Kirtman and George Boer.

A detailed proposal is being developed on numerical experimentation to assess the expected skill of decadal predictions. It is anticipated that the programme of experimentation initiated by this programme will contribute to any IPCC fifth assessment report. The present version of the proposal describes a framework for coordinating intended experimentation covering two different but related objectives:

Objective 1: Short-term prediction of climate for the next 30 years [to 2030 or 2035]
Objective 2: Developing the science of multi-decadal prediction in the context of a changing climate

Experimentation addressing Objective 1 was called for at the Aspen meeting of 2006 (Meehl and Hibbard, 2007) and is currently being planned by various groups, in many cases using high resolution models, and is expected by the wider international community. However, many questions remain about how best to initialize such forecasts, and how to assess the uncertainties in the resulting predictions. Thus, experiments to assess and develop the science of multi-decadal prediction are needed (Objective 2), and individual groups are already starting to work in this area. To address Objective 2, the current plan requests a coordinated set of low resolution experiments and a common framework in which individual groups explore ideas and sensitivities. By linking the framework of the scientific development (often at low resolution) with the requested "best guess" predictions to 2035 (often at high resolution), scientific return is expected to be maximized and comparisons between the various experiments facilitated.

A draft detailed paper outlining the experiments is currently under discussion. In particular, for the Objective 1 experiments:

The aims are:

To provide model integrations to allow estimation of the evolution of expected climate for the period 2005-2035, relative to the climate of recent decades.

To encourage use of higher resolution climate models, with the hope of better resolving synoptic processes associated with extremes, and assessing the benefits of higher resolution in general.

The final analysis of expected climate in 2005-2035 should:

Aim to give guidance on the changing risk of extremes

Aim to give guidance on the possibility of changes in the monsoons

Assess likely errors in the prediction, based on the spread of results from different models and different initialization techniques, and the errors seen in the ability of initialized climate forecasts to reproduce the observed changes of the last few decades - results from Objective 2 experimentation will be needed for this.

Be a step towards probabilistic forecasts of near-term future climate, using a variety of methods and sources, including multi-decadal model predictions, empirical predictions based on attribution of observed changes to date, and results of other modeling studies.

The details of the modeling strategy are still under discussion; currently, Objective 1 requests 30 year model integrations with initial dates 1st November 1960, 1980 and 2005. Each start date should be run with a 3-member ensemble, optionally to be increased to O(10). Ocean initial conditions must represent in some measure the observed anomalies for the start date. A set of proposed experiments is also available for Objective 2.

A key issue for decadal prediction is the determination of the initial conditions for the state of the climate system. So far, this initialization discussion appears to have focused exclusively on the ocean, and a number of approaches to ocean data assimilation have been employed to provide ocean initial conditions. The activities of CLIVAR's Global Synthesis and observations Panel (GSOP) in promoting and evaluating ocean synthesis activities provide an essential input to the decadal prediction activity therefore. The Panel recently held a 2nd Synthesis Evaluation Workshop (MIT from 24-25 September 2007); among its goals was coming to quantitative conclusions about which of the existing synthesis products can provide inputs useful for decadal studies. The report of the meeting is in preparation. GSOP's input to initialization of model runs for IPCC AR5 will be on the agenda of the next GSOP meeting in Southampton, UK, in March 2008.