- The Science
- CLIVAR Research Foci
- CLIVAR Imperatives
- CLIVAR Endorsed Projects & Activities
- CLIVAR Objectives
- CLIVAR Successes
- WCRP Science
- Panels and Working Groups
- CCl/CLIVAR/JCOMM Expert Team on Climate Change Detection and Indices (ETCCDI)
- Global Synthesis and Observations Panel (GSOP)
- WGCM/CLIVAR Working Group on Ocean Model Development (WGOMD)
- Asian-Australian Monsoon Panel (AAMP)
- Atlantic Implementation Panel (AIP)
- CLIVAR-GEWEX Africa Climate Panel (ACP)
- CLIVAR/IOC-GOOS Indian Ocean Panel (IOP)
- Pacific Panel (PP)
- CLIVAR/CliC/SCAR Southern Ocean Panel (SOP)
- Variability of the American Monsoon Systems (VAMOS)
- National Programmes
- Job opportunities
- Early Career Scientists
Frontier 2: Decadal variability, predictability and prediction
Identify and understand phenomena that offer some degree of decadal predictability and to skilfully predict these climate fluctuations and trends
Prepared by: S. Griffies, G. Danabasoglu, H. Drange, W. Hazeleger, B. Kirtman and A. Timmermann
Decadal climate variability and the growing impacts from anthropogenic climate forcing present scientists with an increasingly important challenge: to skilfully predict climate fluctuations and trends for the coming decades. Scientifically meeting this challenge requires identifying and mechanistically understanding phenomena that offer some degree of predictability, and in turn to develop accurate dynamical prediction systems able to realistically capture all forms of climate predictability. This challenge of decadal climate prediction is at the heart of CLIVAR’s mandate. More precisely, a key aim of CLIVAR is to quantify sources of climate predictability on interannual to decadal timescales, and to provide probabilistic regional forecasts with skill sufficient for planning and decision making purposes. Fully “cracking” the climate prediction problem involves an unprecedented multi-disciplinary collaborative effort that includes many of the varied communities focused on earth system science. Indeed, the earth science community has been building its capabilities over multiple generations towards this aim. Faced with a radically changing planet, society demands nothing less from its investments in the science.
Scientific Background and Major Challenges
Scientific elements of decadal variability and prediction: The Earth’s climate possesses a broad range of space-time variations, and those variations extending from the inter-annual to longer timescales involve ocean and/or coupled ocean-atmosphere-ice dynamics. Furthermore, the space-time scale of natural decadal climate variations overlaps with expected trends and patterns associated with anthropogenic climate warming (see figure below). This overlap presents a difficult signal-to-noise problem for attributing observed trends over the next few decades: is the observed fluctuation natural or anthropogenically induced? Understanding the mechanisms of decadal climate variations, quantifying predictability of these variations, coordinating the required observing system, and developing prediction systems, are goals fundamental to CLIVAR’s mission.
Multi-model and multi-scenario ensemble 30-year surface air-temperature trend derived from the CMIP3 database. Blank areas exhibit a higher level of natural variability as compared with the anthropogenic trend
Observed decadal variations in the Atlantic and Pacific: The Atlantic Multi-decadal Variability (AMV) is a basin-wide fluctuation evidenced by 50-70 year swings in sea surface temperature (see figure below). While anthropogenic radiative forcing (i.e. greenhouse gases and aerosols) may have contributed to shaping the observed AMV during the 20th Century, model simulations suggest that similar variability can be generated naturally as a result of variations in the strength of the Atlantic Meridional Overturning Circulation (AMOC). The AMV is known to influence the position of the Intertropical Convergence Zone over northern Africa, causing multi-year droughts (Zhang and Delworth, 2006); it has been demonstrated to affect the strength of the Indian Monsoon, climate over Europe (Collins et al. 2006; Pohlmann et al. 2006), and potentially impacts tropical and mid-latitude cyclones.
Normalised and detrended time series of the Atlantic Multidecadal Variability (red) and the 10-year running mean of the All Indian Rainfall Index
On time scales from years to decades, the formation, propagation, and decay of temperature and salinity anomalies in the North Atlantic subpolar gyre (SPG) have received considerable attention because of the SPG's strong and rapid variability, its importance for the marine climate and ecosystems, its relation to variations in the AMOC, and its potential predictability. The abrupt weakening of the SPG starting in the mid-1990s has motivated extensive studies of its dynamics (Häkkinen and Rhines, 2004, 2009; Hatun et al., 2005, 2009a; Böning et al., 2006; Lohmann et al., 2009). Recent shifts in the North Atlantic SPG have led to an extensive warming and saliniﬁcation of the region, first in the eastern and thereafter throughout the northern North Atlantic (Hatun et al.; 2005, Holland et al., 2008; Sarafanov et al., 2008). The rapid melting of some outlet glaciers on the western coast of Greenland (Holland et al., 2008), as well as large shifts in the marine ecosystem in the northern North Atlantic (Hatun et al., 2009a,b), have also been attributed to the recent weakening of the North Atlantic SPG.
In the Paciﬁc, two independent modes dominate climate variations on decadal timescales – the Paciﬁc Decadal Oscillation or Variability (PDV) (Mantua et al., 1997) and the North Pacific Gyre Oscillation (NPGO, Di Lorenzo et al., 2008). Both modes affect temperature, sea level, salinity, weather patterns, and marine ecosystems. Not all coupled general circulation models reproduce these modes realistically. It has been demonstrated (Qiu et al., 2007) that decadal sea-level variations in the northwestern Paciﬁc can be predicted 5-7 years ahead using observed wind-stress fields and simple ocean dynamical models. Integrating atmospheric noise associated with the North Pacific Oscillation (NPO) the NPGO forces decadal variations in the strength of the Kuroshio jet 3-5 years later. These relationships are expected to form the basis of future decadal prediction systems. In many cases it is the travelling time of Rossby waves at different latitudes that provides important sources of decadal predictability. Exploiting this robust process for the prediction of sea level anomalies, associated ocean transport anomalies, and nutrient anomalies years ahead will become an important component of the decadal prediction efforts in the Pacific region.
Towards a science of predictions: On timescales less than a few years, climate is dominated by internal variability. Predictions of seasonal to inter-annual fluctuations are thus largely initial value problems. On centennial time scales, climate is dominated by changes in external forcing, with anthropogenic changes in greenhouse gas concentrations having become the dominant external climate forcing in recent decades. A centennial projection thus represents a boundary value problem. The intermediate decadal to multi-decadal time scales are perhaps the most complex and least understood from a prediction perspective. It is here that anthropogenic forcing is superimposed on natural modes of internal climate variability; where the ocean's role, including the deep ocean, is fundamental, and where coupled interactions amongst the various components of the climate system play important roles in setting the space-time scales of fluctuations. The decadal prediction problem thus represents a mixed initial-boundary value problem of immense complexity. As for weather forecasting, a vigorous interaction between prediction system development, mechanistic studies with models and observations, and theoretical analysis is essential to build the scientific foundation for decadal prediction.
In recent years, there has been extensive focus on establishing a predictive capability for the Atlantic basin, aiming to examine whether the perfect model predictability examined by Griffies and Bryan (1997), Grötzner et al. (1999), and Collins et al. (2006) can be realised in practise. Predicting the natural low frequency and anthropogenically-forced variations of the AMOC poses an important challenge to climate models. Particular difficulties include: uncertainties in mechanisms of observed variations; biases in climate model simulations; difficulties assimilating the sparse network of observed ocean data into climate models in order to initialise the predictions; and a lack of reliable long-term data to evaluate prediction skill. Nonetheless, climate models initialied with observed ocean conditions appear to have skill in the Atlantic basin on decadal timescales (Zhang et al., 2007; Smith et al., 2007; Keenlyside et al., 2008). A concerted community effort is required to fully develop the scientific basis for such predictions, and more thoroughly examine the robustness of the predictive skill.
For the Pacific, attempts are underway to make long-term decadal predictions of sea level and thermocline depth, exploiting the role of ocean dynamics on these variables. It is at this stage unclear, how much potential predictive skill decadal PDV and NPGO-related SST and climate variations have. A thorough concerted multi-model assessment is needed to quantify the levels of potential decadal skill in the Pacific region for variables of societal interest.
Major Challenges: The decadal prediction problem will not be solved by one grand event. Instead, it is anticipated that the process of developing skilful prediction systems will occur in parallel to the continued research into mechanisms, and the furtherance of observational capabilities. Here, we offer a list of key topics for CLIVAR to consider over the course of the next few years. Addressing these topics involves an unprecedented multi-disciplinary effort that includes all elements of the earth science community, thus making it important for CLIVAR to support bridges to the many relevant scientific organisations.
- What physical and biogeochemical variables have the best signal-to-noise ratio, and which are of societal relevance for impacting predictability? Detailed studies need to be conducted to determine those physical and other processes that contribute to the long-term memory of the climate system and how this memory impacts the societally relevant climate processes. Given that patterns of decadal climate variability are large-scale, multi-basin teleconnections must be studied on these timescales. The ability of the classical basin-panel view to address these scales must be demonstrated.
- What are the best initialisation strategies for decadal predictions? Of particular importance include the implementation of an initialisation strategy that avoids the need for expensive spin-up and transient simulations, how to address model biases, the difficult challenges that coupled model data assimilation poses and the impact, presently unclear, of other components that can potentially affect the slow manifold such as soil moisture, sea-ice, and snow.
- How can we disentangle the impact of anthropogenic radiative forcing and low frequency natural climate fluctuations? For instance, will performing ensemble integrations starting from different initial conditions of the atmosphere and ocean be sufficient. Alternatively, providing a skilful prediction of climate trends associated with climate warming may be of more use to society than predicting fluctuations associated with natural variability patterns.
- How can we address the issues of regional decadal climate predictions, at the scale where society is most impacted? Understanding and simulating the regional aspects of decadal climate change, both natural and anthropogenic, requires high-quality long-term observational records as well as climate model simulations that resolve and capture regional processes accurately. A committed effort to downscale global model control simulations to regional scales using dynamical as well as statistical techniques is required.
- Do we have enough ocean data for skilled initialisations? Ocean data are relatively sparse in space and time. Only since the last decade with Argo do we have a reasonable estimate of global ocean variations. However, data density, even with Argo, and deep-water coverage, for example, still remain limited.
- How do we best fill out the climate attractor with ensemble predictions? In weather prediction, techniques (e.g. singular vectors) have been developed for this purpose. Should such methods be used, or generalised, for decadal predictions?
- How can decadal predictions be evaluated for their potential skill? Forecasts are validated against an independent period and dataset. With very few recorded realisations of decadal variability in the observational record, this task will be a major challenge for decadal predictions but nonetheless essential for society to understand the relevance of the predictions. The paleo-community (PAGES) is presently collecting high-resolution datasets that might prove useful for the validation task requiring inclusion of this community in discussions on decadal predictability.
- How can CLIVAR usefully interface with the impacts community? By advancing understanding of the dynamics and predictability of decadal scale variability, CLIVAR will help to narrow the current gap between observed climate and IPCC-type climate projections. The decadal time window is of key interest and importance for a variety of regional as well as global sectors, like water availability, agriculture and fisheries, transportation, tourism and energy production. Particularly at regional scales climate is substantially governed by fundamental decadal scale modes. The necessary information can only be produced using realistic numerical climate models of the type used in CLIVAR.
Goals: To make progress towards CLIVAR’s decadal prediction goal, many scientific and engineering challenges need to be addressed, each requiring long-term commitment from CLIVAR, other scientific organisations, and various funding agencies. We identify here six goals that encapsulate many of the key tasks required to meet this challenge:
- To garner improved scientific insight into the statistical properties and dynamical mechanisms of low-frequency climate variability, including trends.
- To quantify the level of committed climate change due to past greenhouse gas emissions, and to understand how anthropogenic climate change interacts with natural variability modes on all relevant time scales.
- To identify those phenomena and trends possessing useful predictability, and to ensure that model systems possess accurate representations of all relevant components so to approach predictability limits with realistic prediction systems.
- To remedy sources of model biases at global and regional scales that render disagreements with observations, and which can lead to simulation features that vary widely between different model systems.
- To initialise dynamical predictions using data assimilative methods, and to quantify forecast skill by comparing to historical climate variations.
- To exploit high-quality and high-resolution paleo-climate records to help evaluate decadal climate simulations and predictions.
CMIP5 prediction experiments: A central element of decadal climate prediction research involves the planned prediction component of CMIP5. These initialised prediction experiments will help to identify areas of potential skill. CLIVAR should continue to play a leading role in the analysis of these experiments, and to sponsor a selection of workshops aimed at scientifically rationalising the results.
Biases: It is anticipated that the CMIP5 prediction experiments will highlight present limitations and biases of the various model prediction systems. CLIVAR should play a leading role in supporting the improvements of the systems, in particular by encouraging the reduction of model biases so that the models accurately represent the mechanisms active in the real world. One avenue for supporting the improvement of ocean model components includes the CORE-II hindcast experiments.
Workshops: In general, key workshops will be required to communicate developments, discuss and debate alternative priorities and research paths, and address how the forecasts might be used.
Experimental design: CLIVAR should foster WGSIP/WGCM decadal prediction experimental design activities. It should also continue to support workshops related to decadal variability, predictability, and prediction. Two examples of such workshops include the November 2009 KNMI workshop on initialisation strategies for decadal prediction and the September 2010 WGOMD-GSOP workshop on the role of the ocean on decadal variability.
Linkages to other groups: Given the multi-disciplinary features of the decadal prediction problem, CLIVAR meetings on decadal prediction should include selected representatives from the relevant scientific groups. In addition, there is a US CLIVAR working group on decadal prediction that is attempting to ensure that the community is well positioned to analyse the CMIP5 runs as they come available. International CLIVAR needs to identify where it ﬁts with this activity.