The last session of the workshop included a presentation of key messages developed over the course of the 2 days. Many of the messages have been discussed in depth previously in the report (see Box 1), and therefore are not repeated here. This section focuses on the discussions related to challenges and opportunities in observing and modeling decadal variability, as well as key knowledge gaps.
Metrics of Decadal Climate Variability
Many participants discussed ways to measure and detect decadal variability. In particular, they questioned whether global mean surface temperature (GMST) should continue to be used as the prominent metric for change. There are large uncertainties in calculating GMST, largely because of the lack of station data coverage in the Arctic, Antarctic, and African regions, according to some participants. Although remote sensing can help to address these data gaps, ground-truthing would be necessary to verify these measurements. Another concern is that GMST only measures part of the global energy budget, and a relatively small component at that. As such, variations in GMST do not fully reflect the effects of human-caused emissions on the climate system.
Despite these shortcomings, most workshop participants agreed that it is important to monitor and understand surface temperature, given that people live on Earth’s surface and there are long records of this metric relative to others. Shang-Ping Xie also pointed out that regional affects and impacts are much more important than a global average for informing decision makers. Xie argued that use of GMST as a metric causes information loss. He said that “unpacking the data” reveals seasonal and regional information that may hold the key to identifying important mechanisms.
In response to these limitations of GMST, some participants suggested alternative metrics that might be more accurate measures for global change. One possible metric is ocean heat content, particularly if expanded observations can help reduce current uncertainties. New observations would be most useful if the focus is on regions where heat uptake is thought to be the largest, including deep water formation regions in the high latitudes and the upper ocean. These observations would need to be more uniform, according to some participants, because Argo does not evenly sample the global ocean and thus is subject to sampling errors in large-scale averages or balances. Some participants noted that it would also be important to focus on where the change in heat uptake is thought to be largest (e. g., the top 300 m) and to include deep Argo measurements for multi-decadal timescales. Other participants noted that available ocean observational data (ocean temperature and heat content estimates) are within acceptable uncertainty, particularly after 2005 (Nieves et al., 2015).
Other possible metrics include the top of atmosphere (TOA) radiative balance, where there are still considerable inaccuracies; sea ice extent, which is well constrained by satellites but thickness and ice volume have only been possible to observe recently (via ICESAT and CRYOSAT); and global mean sea level (and sea level pressure), which would integrate the ocean and cryosphere response. Participants noted that adequate monitoring of climate change for studying decadal variability would truly require a combination of metrics, for example, the use of global sea level rise coupled with GMST.
Confronting Models with Observations
Regardless of the metric (or metrics) chosen to monitor global climate, the community’s understanding of drivers and mechanisms of decadal change is limited by the existing data record. Participants agreed that sustaining and enhancing observing networks to better monitor the global climate system is important, but synthesis of existing observations to better understand past variability and associated processes is generally lacking. Observations are required for the verification and testing of decadal predictions, but data coverage is inadequate and the length of records is short relative to what is needed to validate variability and dominant processes in existing models. Much of the workshop involved proposing potential mechanisms and drivers of change, which were analyzed in the context of a given time period. However, given the relative brevity of the instrumental era, very few samples exist to consider.
Many workshop participants also recognized the need to continuously confront models with observations. Verification of model performance from observations is an important step toward developing prediction capability. Confronting models with observations is also important to distinguish forced and internal change through fingerprinting (e. g., Coupled Model Intercomparison Project Phase 6 [CMIP6] pacemaker experiments). Providing real-time forcing datasets for better synthesis of the current state would also be beneficial in improving this capacity, according to some participants.
Observational Challenges, Needs, and Opportunities
Some individual participants identified additional observational challenges, needs, and opportunities:
- The challenges in combining local measurements over ocean or land are worsened by differences in platforms and temporal integrity of local observations. Therefore, in addition to making new observations, maintenance of current observational systems is required at a minimum.
- Paleo proxies offer many opportunities, although synthesis of existing records is currently under resourced and underutilized.
- Sources of observations other than temperature could be used to improve understanding of decadal climate variability. Paleo records would provide isotopes (to compare to rainfall on land). Argo would provide not only temperature but also salinity and gradients of salinity.
- Other specific areas that would benefit from improved observations include
- geographical distribution of aerosols below 15 km to determine contribution of external forcing, and
- ocean isotope geochemistry in the equatorial Pacific to determine El Niño/La Niña occurrences during the past 1,000 years.
Modeling Challenges, Needs, and Opportunities
Some individual participants identified additional modeling challenges, needs, and opportunities:
- Regional patterns and cross-timescale interactions are important, but not all models can capture the full collection of processes and phenomena that have been deemed relevant to regional (or basin-scale) variability, which limits understanding.
- It is important to focus efforts on improving model representation of the modes of variability that have the potential for predictability, although the questions may still
- remain: why are some modes more or less predictable? What are the mechanisms leading to this predictability?
- It is also important to employ a hierarchy of models—process-based, linear inverse models, and global climate models—to better explore the limits of predictability.
- Models could be used to inform observational needs.
- Other general areas for improvement identified include
- model initialization (e. g., coupled assimilation),
- reduced model uncertainties and bias (i. e., we do not yet know which biases affect variability on decadal timescales), and
- better incorporation of known forcing and known uncertainties in forcing.
Although much progress is being made toward understanding decadal variability, as presented at the workshop, important questions remain, in particular in separating the contributions of each proposed driver. Many of the mechanisms examined might be driving decadal variability, but what is driving the mechanisms themselves? For example,
- If Pacific Decadal Variability (PDV) is a combination of different modes, how can they be parsed out? What role does each play? What is the mechanism for each mode?
- Although the North Atlantic Oscillation (NAO) seems to drive Atlantic Multidecadal Variability (AMV), what drives NAO multi-decadal variability?
Other knowledge gaps include the following:
- The connection between Arctic sea ice loss and mid-latitude weather, and the consequential regional effects;
- The role of and quantitative data on stratification of the deep Southern Ocean;
- The relative importance of atmospheric vs. oceanic bridges in linking stochastic processes at mid-to-high latitudes (how does local atmospheric forcing produce remote response(s) on decadal timescales?); and
- How heat trapped in the ocean will be transported into the deeper layers in the one or two decades and how that might affect global temperatures in the future.
Some participants highlighted the importance of the community reaching agreement on how to quantify and communicate the concept of uncertainty to reduce confusion among the public, as well as among those studying climate variability. The participants emphasized the need for scientists to be clear and careful about their definitions and derivations of uncertainty, because differences can be easily misconstrued by the public as disagreement.
Many participants believe it is important to not associate “variability” with “oscillation,” and in discussions of variability to provide quantitative clarity. In addition, the scientific community should define a minimum time interval over which to label a GMST trend, and potentially define trends associated with the adjectives “small,” “moderate,” “strong,” and “extreme.”
Many participants suggested that the way forward includes improvement of the mechanistic understanding of the processes and drivers (both internal and external) that contribute to decadal climate variability, assessment of this understanding, followed by development of prediction and attribution capabilities.
The emphasis on the recent slowdown period has stimulated a very useful area of research in decadal variability and predictability more broadly. Examination of the questions related to recent GMST trends can offer many scientific insights about the physical climate system. A key focus moving forward should be to use these insights to predict these longer timescale variations in Earth’s climate. Enhanced understanding of the dynamics and underlying physics of variability in the climate system would lead to higher quality information that could inform model development and validation, which the community can then use to make and verify predictions.
Some participants reiterated the importance of developing predictive capability for decadal variability of seasonal-to-interannual coupled ocean-atmosphere phenomena, including weather, for selected geographical areas for specific phenomena:
- El Niño/La Niña
- Tropical Atlantic
- Arctic sea ice
- Southern Ocean/Antarctic sea ice (surface winds, ocean stratification)
- Bottom- and intermediate-water formation
They noted, however, that some studies of predictability of this most recent, as well as other slowdown periods, have met with some success (see Toward Predictability).
In addition to improving prediction capabilities, addressing the gaps in knowledge of decadal climate variability could lead to better-informed climate change attribution studies, that is, the ability to detect the signals of anthropogenic climate change and internal variability distinctively for certain events with much greater accuracy. Both the prediction of decadal climate variability and attribution of specific climatic events and trends can be used to better inform decision makers.