全球变化知识资源中心

Recent sociological studies show that over short time periods the large day-to-day, month-to-month or year-to-year variations in weather at a specific location can influence and potentially bias our perception of climate change, a more long-term and global phenomenon. By weighting local temperature anomalies with the number of people that experience them and considering longer time periods, we illustrate that the share of the world population exposed to warmer-than-normal temperatures has steadily increased during the past few decades. Therefore, warming is experienced by an increasing number of individuals, counter to what might be simply inferred from global mean temperature anomalies. This behaviour is well-captured by current climate models, offering an opportunity to increase confidence in future projections of climate change irrespective of the personal local perception of weather.

Recent extreme climate anomalies in densely populated regions, such as the cold 2011/12 and 2013/14 winters in the eastern United States, ongoing drought in California, heat waves in Europe (2003), Russia (2010) and Australia (2013), or floods in Pakistan (2010), Colorado (2013) and the United Kingdom (2014), have received broad media attention and fuelled the discussion on the attribution of such events to climate change¹. From a purely physical point of view, the attribution of an individual extreme event solely to anthropogenic climate change is essentially impossible, as the synoptic, chaotic components will always dominate the genesis and evolution of an event. Attribution requires an increased number of events over time — hence enough data — so that a robust trend can be detected in the frequency of occurrence of extreme events. To tackle this issue, scientists have long used statistical and dynamical models to simulate such events multiple times, in order to increase the sample size or to conceptualize the genesis of these events and thus arrive at robust conclusions regarding the role of climate change in the story². Along the same lines, scientists have also debated how a slightly changed background state (such as increased sea surface temperatures or increased moisture in the air) may influence the likelihood or magnitude of an individual extreme event occurring³. It is worth noting that the few robust trends that have already emerged from the short and noisy observational records are mostly temperature-related and agree well with our physical understanding of how such extremes will change in a warming climate⁴.

Despite all the scientific evidence, local short-term variations in weather are more salient to an individual than a long-term trend and hence are critical for his or her perception of how weather and climate are interlinked^{5, 6, 7, 8}. By climate science standards, the studies in refs 5,6,7,8 focused on relatively short time periods and showed that seasonality and short-term trends in temperature can influence one's perception of whether it has actually become warmer or colder in a specific location⁶. They further emphasize how weather anomalies influence one's belief in the concept of climate change⁵ or, vice versa, how pre-existing belief in climate change or political orientation affects the perception of a given weather anomaly⁷.

Using monthly temperature from observations⁹ and climate model simulations, we illustrate how population-weighted climate data can help grasp the global scale of climate change, while retaining a close tie to the individual experience of short-term variations in temperature. The focus is on monthly temperature as it constitutes one of the longer and more reliable gridded climate records and is easily extracted from the climate model simulations on which the recent Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) based its future projections¹⁰.

As an illustrative example of spatial heterogeneity, Fig. 1a shows temperature deviations during the past year (November 2013 to October 2014) from the 1951–1980 average for all land areas. While the eastern United States saw colder-than-average temperatures, most other land areas experienced an above-average year, in line with 2014 being the warmest (or second-warmest: http://www.skepticalscience.com/cowtan_way_2014_roundup.html) year on record globally. Figure 1b shows the same time period, but expressed as standard deviations (σ) from the same reference period. The standard deviation offers a more tangible expression of temperature anomalies as it takes into account the natural range of temperature at a given location on the planet. Exceeding a certain local σ value therefore provides a good measure for how unusual a given temperature anomaly actually is for a person living there. Yet, people in the tropics might not notice small temperature changes, even if they are significant in light of the naturally small temperature variability there¹¹. At high latitudes, on the other hand, people might have experienced large but statistically insignificant changes in temperature over the past decades. Further, the reference climate for an individual person would depend on that person's age, but this is not considered here. Therefore, other metrics than the one used here could be thought of to characterize human temperature exposure¹².

a, Twelve-month mean temperature anomalies with reference to (wrt) the 1951–1980 period. b, As a, but expressed as standard deviations (σ).

To test the ability of climate models to capture the observed changes, we apply the same analysis to simulated temperature from the CMIP5 (Fifth Coupled Model Intercomparison Project) simulations (Fig. 4). The observed increase in 1σ and 3σ exceedances over the past decades is well encompassed by the multi-model range, suggesting that the models are skilful in this metric.

a, For the RCP 8.5 scenario; b, for the RCP 2.6 scenario. The shading gives the range of the CMIP5 models, thin black lines give the multi-model mean; observations are in thick black lines. The high scenario population projections are used here (see Supplementary Fig. S2). The time series are calculated with monthly mean temperature anomalies; however, for clarity, only annual means of these time series are shown here.

The results here imply that the debate on whether reduced decadal trends of global mean temperatures are undermining people's belief in climate change (or climate models) is essentially decoupled from the actual temperature perception based on population-weighted climate data over the past decades. It indicates that by focusing communication solely on global mean temperature changes over the past 15 years, objective information on the real temperature exposure of humans can be effectively obscured. Instead, the ability of climate models to capture recent trends in temperature threshold exceedances should enable the public and commentators to refer to model projections with more confidence in this quantity, which in addition provides a view on climate change that is more tailored to the human perception than global mean temperature.

Beyond individual perception of climate change, these results have a wider importance because humans primarily use ecosystems services close to their place of residence. Projections of how such services might evolve under climate change need to take into consideration how many people depend on them at a specific location.

For temperature observations we use the 2° × 2° Goddard Institute for Space Studies Surface Temperature Analysis (GISTEMP) as monthly anomalies to the reference period 1951–1980, a period of relatively little trend in global mean temperature and hence a good time frame for estimating natural variability⁹. Before estimating the standard deviation (σ) from this period, we linearly detrend temperature at each location over these 30 years. The CMIP5 models used for each scenario combination (historical + RCP 8.5 and historical + RCP 2.6) are listed in Supplementary Table S1. All simulations were bilinearly regridded to 2° × 2° and treated in the same way as the observations to estimate σ. Using an alternative dataset from the Hadley Centre and the Climate Research Unit (HadCRUT4¹⁷) did not alter the conclusions.

The population data stem from the History Database of the Global Environment (HYDE 3.2; ref. 13 and K. Klein Goldewijk and A. Beusen, manuscript in preparation), which incorporates census data from the UN World Population Prospects (http://www.un.org/en/development/desa/population/theme/trends/index.shtml) for 1950–2010 and bases on the Shared Socioeconomic Pathways (SSP)¹⁸ for 2010–2100. Its future projections are compatible with both the RCP 8.5 and RCP 2.6 scenarios in terms of socio-economic trajectory. The data have been regridded to 2° × 2°, conserving global population.

In case of σ exceedance for a given grid cell, the full population in that grid cell is counted towards the population experiencing the particular σ exceedance. Similar to ref. 2, we thereby aggregate both climate and population data to a spatial scale that may lead to an underestimation of the coupling between temperature anomaly and perception.

The multi-model range in Fig. 4b illustrates that for the metric presented here there are larger uncertainties associated with the mitigation scenario (RCP 2.6) than with the business-as-usual scenario (RCP 8.5), something that does not follow simply from the global or regional mean temperature, which show a comparable spread for RCP 2.6 and RCP 8.5 in the latest IPCC assessment¹⁹. Instead, it seems to be difficult for models to agree on whether temperatures exceed a given threshold, at the locations where the majority of the world population lives, in the presence of a weak climate change signal (RCP 2.6), while they agree better for a scenario with a strong climate change signal (RCP 8.5). Seemingly a signal-to-noise issue, the reasons for this scenario-dependence of model agreement are not easily diagnosed from the existing literature¹⁶ and may merit further investigation that is beyond the work presented here.

1. Herring, S. C., Hoerling, M. P., Peterson, T. C. & Stott, P. A. Explaining extreme events of 2013 from a climate perspective. Bull. Am. Meteorol. Soc. 95, Spec. Suppl. (2014).
2. Stott, P. A., Stone, D. A. & Allen, M. R. Human contribution to the European heatwave of 2003. Nature 432, 610–614 (2004).
   • CAS
   • ISI
   • PubMed
   • Article
3. Trenberth, K. E., Fasullo, J. T. & Shepherd, T. G. Attribution of climate extreme events. Nature Clim. Change 5, 725–730 (2015).
4. IPCC Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation (eds Field, C. B. et al.) (Cambridge Univ. Press, 2012).
5. McCright, A. M., Dunlap, R. E. & Xiao, C. The impacts of temperature anomalies and political orientation on perceived winter warming. Nature Clim. Change 4, 1077–1081 (2014).
   • Article
6. Howe, P. D., Markowitz, E. M., Lee, T. M., Ko, C-Y. & Leiserowitz, A. Global perceptions of local temperature change. Nature Clim. Change 3, 352–356 (2013).
   • ISI
   • Article
7. Howe, P. D. & Leiserowitz, A. Who remembers a hot summer or a cold winter? The asymmetric effect of beliefs about global warming on perceptions of local climate conditions in the U.S. Glob. Environ. Change 23, 1488–1500 (2013).
   • Article
8. Zaval, L., Keenan, E. A., Johnson, E. J. & Weber, E. U. How warm days increase belief in global warming. Nature Clim. Change 4, 143–147 (2014).
   • ISI
   • Article
9. Hansen, J., Ruedy, R., Sato, M. & Lo, K. Global surface temperature change. Rev. Geophys. 48, http://doi.org/ckmvrz (2010).
10. Collins, M. et al. in Climate Change 2013: The Physical Science Basis (eds Stocker, T. et al.) 1029–1136 (IPCC, Cambridge Univ. Press, 2013).
11. Mahlstein, I. & Knutti, R. Early onset of significant local warming in low latitude countries. Environ. Res. Lett. 6, 034009 (2011).
    • Article
12. Dunne, J. P., Stouff