全球变化知识资源中心

Large ensemble climate modelling experiments demonstrate the large role natural variability plays in local climate on a multi-decadal timescale. Variability in local weather and climate influences individual beliefs about climate change. To the extent that support for climate mitigation policies is determined by citizens' local experiences, natural variability will strongly influence the timescale for implementation of such policies. Under a number of illustrative threshold criteria for both national and international climate action, we show that variability-driven uncertainty about local change, even in the face of a well-constrained estimate of global change, can potentially delay the time to policy implementation by decades. Because several decades of greenhouse gas emissions can have a large impact on long-term climate outcomes, there is substantial risk associated with climate policies driven by consensus among individuals who are strongly influenced by local weather conditions.

Extreme weather events, including hurricanes Sandy and Katrina and the heatwaves experienced by much of Europe in 2010 or in the United States in 2012, spur public discussion about reducing greenhouse gas emissions — regardless of the extent to which these events can actually be attributed to past emissions. In contrast, the fact that most places have experienced relatively little warming over the past decade¹ has caused some to doubt the reality of human-induced global warming and to argue against actions to reduce greenhouse gas emissions. Natural variability and weather extremes, in particular local warming, influences people's belief in climate change and their willingness to support potentially costly climate mitigation measures^{2, 3, 4, 5, 6}. However, climate variability and extreme events involve a fundamentally chaotic component^{7, 8}. Society's response to the threats posed by climate change partly depends on unpredictable meteorological events, adding another layer of uncertainty and unpredictability to policy outcomes.

Despite overwhelming scientific evidence for the impending damages caused by anthropogenic climate change, climate policy leading to substantial emissions reduction has been slow to materialize⁹. A number of recently published works in the psychological literature has shown that citizens' belief in climate change and their willingness to pay for emissions abatement are influenced by their experiences with heatwaves and other extreme local conditions^{2, 3, 4, 5, 6}. Risk perceptions and willingness to pay for risk mitigation are influenced by local extreme events or accidents. From heatwaves and floods, to hurricanes, to nuclear power or natural gas pipeline accidents, people exhibit an availability bias^{2, 10} when making economic decisions impacted by environmental or disaster-related risks^{11, 12, 13, 14, 15, 16}. These changes in perception and willingness to pay are influenced by proximity¹⁴ and event severity¹³, and decay with time after the event^{16, 17}.

An extreme event may therefore open a 'policy window'¹⁸ — a limited opportunity to implement climate change policies that would lack sufficient political support in the absence of such a focusing event. While an extreme event alone is unlikely to bring about policy change¹⁹, when aligned with hospitable political and institutional conditions it may provide a critical impetus for local policy adoption^{20, 21}.

As global temperatures continue to rise, a confluence of natural variability and the forced response of the climate system to greenhouse gases will increase the frequency of unprecedented events, such as the 2010 heatwave in eastern Europe²². Although the forcing-driven component of increased extreme event frequency is to some extent predictable, the natural variability-driven component — and thus the timing and location of these events — is largely stochastic.

Large ensemble climate modelling experiments have demonstrated the important role natural variability plays in the range of regional climate change predicted for coming decades. Indeed, in many regions, the multi-decadal component of natural variability may be as large as the forced response to rising greenhouse gas concentrations over the next several decades^{7, 27}. Natural variability influences local meteorological conditions on timescales from days to decades^{24, 25, 26, 27}. Thus, natural variability may mask the forced response in some countries while exacerbating the forced response in others, with differential consequences for public support for climate policy in each country. Because support for climate policies is likely to be affected by the local experiences of citizens, natural variability can be expected to significantly influence the timescale for action to mitigate climate change.

There is no shortage of factors that make it difficult to predict the future evolution of complex social systems. To illustrate the influence that unpredictable extreme weather events may have on the time to reach a global agreement on climate policy, we present an analysis in which weather is the only unpredictable factor. In our simple illustrative model, deterministic citizens are confronted with a stochastic world represented by climate model projections. We analysed the output from a 40-member coupled climate model ensemble²³ to illustrate how local experiences might affect the timing of acceptance of strong climate policy measures. Each of these ensemble members are subject to identical climate forcings, yet experience different weather (see Supplementary Information for details about our analytical methods). Our analysis illustrates how timescales for domestic and international policy action will be influenced by natural variability in local weather if a nation's decision to take strong actions to abate emissions is contingent solely on the local experiences of its citizens.

A wide range of climate risk-reduction policy options are likely to be considered simultaneously, involving a wide range of decisions, but for simplicity our illustration involves a single policy decision. Because quantitative data about the magnitude and timescale of the influence of extreme weather events on willingness to take action to mitigate climate change is limited, we draw on the broader literature about risk decision making following natural disasters (including weather-related ones) to formulate our model. In our illustration, we assume that in any given climate model grid cell²⁸ the fraction of the population that is convinced of the need for strong climate change risk mitigation policies is increased by unusually hot months. The fraction of the unconvinced population that becomes convinced increases with the extremity of the event. In addition, the fraction of the population previously convinced (and thus willing to pay for policy changes) decreases each month with a timescale similar to those observed in the literature about risk mitigation behaviour following natural disasters^{16, 17}, which is consistent with timescales suggested in the more limited quantitative assessments of changes in climate risk perceptions following extreme weather events^{3, 5}.

In this model, the fraction of the population (f) convinced to pay for mitigation policy is represented by the following equation:

where N represents the extremity of the event (as in a 1-in-N hot month) relative to a baseline period. N is calculated in terms of the month's temperature as measured in standard deviations (σ) from an early twenty-first century baseline: N = 1/(1 – erf(σ/√2)). The parameter k represents the sensitivity of the unconvinced population to extremes, and τ is the time constant according to which supporters of policy action lose their willingness to support those policies (Supplementary Methods and Supplementary Fig. 1).

In our model, a national tipping point for the policy action occurs when half of the country's population is convinced. We refer to the time that this social tipping point is reached for a particular country as the 'time-to-action' for that country. Because populations' sensitivities to extremes are uncertain, natural variability could delay action by more or less than indicated by our analysis. The point of this exercise is to illustrate the potential interplay of natural variability and the forced climate response in influencing public perceptions and climate policies within the confines of the available dataset. Therefore, other aspects of our model are kept as transparent and simple as possible.

To the extent that climate policies are driven by the weather experienced by a country's citizens, variability in weather will result in significantly disparate times-to-action. The spread of the model results for each country is entirely due to internal climate variability — our modelled social system is deterministic and the only differences between simulation ensemble members are small perturbations in the initial conditions of the atmosphere. For the top six global CO₂ emitters, natural climate variability results in a range of times-to-action spanning several decades (Fig. 1). Furthermore, the times-to-action for these large emitters are not strongly correlated (Supplementary Table 1). The potential range of outcomes is narrower and the median times-to-action are nearer-term for the European Union, India and Japan than for China and, in particular, the United States and Russia.

Time-to-action is the time it takes a country to support an international agreement. Box-and-whiskers show the median, 25–75% confidence, and maximum–minimum limits for the 40 ensemble members for each nation/economic entity.

Climatologically effective emissions abatement will be contingent on widespread international adoption of abatement policies. Like domestic policy tipping points, policies at the international scale also have a documented, but highly uncertain, tendency to propagate after reaching certain thresholds. Once large-scale changes to economic and social norms (such as the reduction of trade barriers, universal suffrage, women's rights and international human rights standards) are adopted by an unpredictable 'critical mass' of states or actors, adoption by many more follows in a relatively short period of time^{30, 31}. Diffusion of policies can take place in national³², international³³ or even sectoral contexts³⁴. Such policies may be eventually codified in formal international legal agreements, but such binding agreements are not necessarily required for these norms to be effective^{35, 36}. Therefore, large-scale international action to mitigate climate change could potentially occur if enough powerful international actors adopt new norms, which would ultimately induce near-global adoption. Such norm adoption need not necessarily be codified through the UNFCCC process, although it could be.

We consider the timescales for effective international action under two different institutional constraints on reaching an agreement, which reflect some of the potential differences in international policy processes. Under the first, stricter constraint ('simultaneous consensus'), a window for effective international action will only open when a quorum of nations is simultaneously convinced to take action. Under the second, looser constraint ('incremental consensus'), a nation permanently buys into an international agreement at the time its own population is convinced to take action, and the international agreement then enters into force when a quorum is reached. Under this constraint, national policies are passed when a national-scale policy window opens, and an international policy window is also partially propped open. In effect, this incremental approach ratchets an international policy window open state by state.

We also examine timescales for agreement under three hypothetical international tipping points representing different international power schemes: (1) majority population, (2) majority economic output and (3) consensus among the US and China. We refer to the time to reaching effective international-scale action as the 'years-until-agreement', for which 'agreement' may refer to anything from a formal binding one to a substantive normative one.

Under the simultaneous consensus constraint, in which international agreements are not reached until all relevant parties have reached the tipping point simultaneously (dark-bordered bars in Fig. 3), the median years-until-agreement is considerably larger than under the incremental consensus constraint (light, solid bars in Fig. 3). Even if national-scale commitments are not effective at reducing emissions, but only effective at maintaining national support for eventual emissions reductions, such commitments serve to reduce long-term risk if they ratchet open an international policy window. The median 'ratcheting effect' of national policies is a decade or more reduction of the years-until-agreement under all international power schemes.

Histograms of years until a, countries representing more than 50% of world population have agreed to take action, b, countries representing more than 50% of world GDP have agreed to take action, and c, both China and the US have agreed to take action. Empty bars illustrate results under the simultaneous consensus constraint and light, solid bars show results for the incremental consensus constraint. For the ensemble members shown by the rightmost bar in panel c, the criterion for agreement is never met.

The more inequitable the world power scheme, the larger the influence of natural variability. If an agreement is reached when countries representing more than 50% of the world's population reach their time-to-action (Fig. 3a), then agreement is reached, on average five years sooner than when countries representing more than 50% of the world GDP reach their time-to-action (Fig. 3b). Few think that an international coalition will be both effective and viable without the participation of the United States and China. The timescale for an agreement in which both China and the US have passed the threshold is more uncertain, and more than a decade longer, on average, than other more inclusive schemes (Fig. 3c).

One might imagine that decisions based on global perspective, rather than on local conditions, would shorten timescales for policy action. When policy enactments are based on not only how convincing weather is locally, but how convincing it is if citizens adopted a global viewpoint (caring equally about the weather each person on Earth is experiencing) then timescales associated with all power schemes and both institutional constraints converge to the same distribution (that is, distributions in Fig. 3a–c all converge to that in Fig. 4a). If simultaneous consensus is necessary to achieve effective global emissions reductions, a global perspective will always reduce both uncertainty and the number of years until agreement (Fig. 4b, bordered box-and-whiskers). However, in a world with policy-window-ratcheting, an emphasis on global-scale impacts may have a mixed effect. Uncertainty is still always reduced, but not necessarily the timescales for international agreement.

a, Histogram of the years until an international agreement is reached if citizens weight the experiences of all mankind equally. b, Box-and-whiskers for all three power schemes and policy processes (colours as in Fig. 3) in comparison with the global result (black). Arrows illustrate the 'ratcheting effect' of an incremental international policy process.

Our model demonstrates risks presented by allowing local weather to drive policy. However, we counsel against reading too much into our precise quantitative analysis. As with many simple models of complex systems, the one we present omits many significant components of the system for the sake of mechanistic transparency. Our model represents naive views of human psychology and national and international politics. We aim to illustrate the large potential influence that natural variability has on policy when support for policies is contingent on local phenomena. Our analysis is not intended to provide quantitative predictions for the range of times to international agreement on emissions abatement. The social tipping points that we explore, while qualitatively supported by the behavioural science literature, are quantitatively arbitrary.

One could imagine any number of different events, trends or conditions that trigger public support for emissions abatement and other climate risk mitigation policies. While we use extremely hot weather in our illustrative example, extreme precipitation events undoubtedly play a role in people's perceptions as well. Our model assumes a common baseline for all nations, but contemporary public concern about climate change and public support for policy actions vary considerably from country to country³⁷. In reality, the population of each country has its own set of social and cultural conditions that would cause them to respond differentially to extreme weather events^{38, 39, 40}.

It is well documented that a 'focusing event', such as an extreme weather event, is