globalchange  > 气候变化事实与影响
DOI: doi:10.1038/nclimate2483
论文题名:
Influence of internal variability on Arctic sea-ice trends
作者: Neil C. Swart
刊名: Nature Climate Change
ISSN: 1758-1033X
EISSN: 1758-7153
出版年: 2015-01-28
卷: Volume:5, 页码:Pages:86;89 (2015)
语种: 英语
英文关键词: Cryospheric science
英文摘要:

Internal climate variability can mask or enhance human-induced sea-ice loss on timescales ranging from years to decades. It must be properly accounted for when considering observations, understanding projections and evaluating models.

A broad range of evidence shows with high confidence that human-induced climate warming has driven a decline in Arctic sea-ice extent over the past few decades1. However, the rate of sea-ice decline has not been uniform. Arctic sea-ice extent was lost at a considerably higher rate from 2001–2007 than in the preceding decades (Fig. 1), which caught the attention of scientists and the public alike2. In contrast, from 2007–2013 there was a near-zero trend in observed Arctic September sea-ice extent, in large part due to a strong uptick of the ice-pack in 2013, which has continued into 2014. By deliberately cherry-picking these periods we will demonstrate how using short-term trends can be misleading about longer-term changes, when such trends show either rapid or slow ice loss.

Figure 1: Arctic September sea-ice extent anomalies.
Arctic September sea-ice extent anomalies.

Sea-ice extent anomaly relative to 1980–2000 from observations (red) and 102 realizations from 31 CMIP5 models (grey), along with the CMIP5 ensemble mean (black). Linear trends are fitted to the observations over 2001–2007 (green) and 2007–2013 (blue). The CMIP5 ensemble mean is calculated such that each model has a weight of 1. Observations extend to 2014.

How likely is a 7-year period of near-zero trend in September Arctic sea-ice extent, as observed between 2007 and 2013? To answer this question we examine trends in Arctic sea-ice extent for all 7-year periods between 1979 and 2013 in the observations and in 102 realizations from 31 Coupled Model Intercomparison Phase 5 (CMIP5) global climate models (see Supplementary Information). If there was no long-term background trend in sea-ice extent we would expect random variability to lead to positive 7-year trends about 50% of the time (or with a probability p = 0.50). Alternatively, if internal variability was small compared to the background trend we would expect 7-year positive trends to be rare. In the model simulations of the past 35 years a 7-year period where a September extent trend was greater than or equal to zero occurs with a probability of p = 0.34 on average across the models (Fig. 2a). Thus, according to the models there is about a one in three chance of a 7-year period with a positive sea-ice trend, despite strong anthropogenic forcing.

Figure 2: Arctic September sea-ice extent trends.
Arctic September sea-ice extent trends.

a, Distribution of all possible 7-year trends between 1979–2013 for observations (red), the CMIP5 realizations (grey) and the 30 CESM1 LE realizations (cyan). b, As in a but for 14 year trends. c, As in a but for 35 year trends. The solid green and blue lines in a are the observed linear trends from 2001 to 2007 and 2007 to 2013, respectively. Density implies that the area under each distribution equals one.

Previous studies have indicated that the multi-model mean long-term September sea-ice extent trend is smaller than observed1, 15. The observed trend reflects both the forced response of the system to anthropogenic influence as well as internal variability, even when considering multidecadal timescales1. In the model ensemble mean the influence of internal variability has effectively been averaged out and the trend mostly reflects the true model response to external forcing. Thus, there is no reason to expect the observed trend to fall on the model mean, even given perfect models. Similarly, the fact that the observed trend falls outside the given confidence interval of many individual model realizations is often taken as evidence that the models underestimate the observed trend11, 15. However, trends in the observations and individual model realizations may differ significantly due to random differences in internal variability, rather than differences in the true response to external forcing16, 17.

Here we compare the observed long-term trend over 1979–2013 (35 years) with the full distribution of trends simulated by the many individual model realizations which include internal variability (Fig. 2c). This shows that the observed trend falls well within the distribution of simulated trends, lying at the thiteenth percentile of the CESM1 LE distribution and at the twentieth percentile of the CMIP5 distribution. For CMIP5 the spread of trends arises due to internal variability and inter-model spread. To properly account for both of these sources of uncertainty, we apply a carefully designed statistical test with the null hypothesis that the observed and simulated trends are equal (Supplementary Information). With p-values of 0.075 for the CMIP5 ensemble and 0.19 for CESM1 LE we cannot reject the null hypothesis at the 5% level for either set of models. Therefore, when accounting for internal variability, long-term trends in September Arctic sea-ice extent do not support the conclusion that the models, as a group, systematically underestimate the response to anthropogenic forcing.

The probability of a pause in September sea-ice extent changes in the future, largely because of changes in the background anthropogenic trend. This also means that future probabilities depend on the emissions scenario. The strong mitigation of emissions under Representative Concentration Pathway (RCP) 2.6 reduces the long-term trend in September sea-ice extent (Fig. 3a). As the long-term trend reduces, the probability of a 7-year pause increases, and approaches p = 0.5 after about 2050 when the average long-term trend becomes near zero (Fig. 3b). Similarly, the probability of a pause increases towards the end of the century under RCP4.5. In contrast, under increasing emissions in RCP8.5 the long-term decline in Arctic sea-ice extent accelerates over the next few decades and then begins to level off towards the end of the century (Fig. 3a). As a result the probability of a pause under RCP8.5 reduces sharply until about 2075, after which it holds roughly constant near p = 0.16 (Fig. 3b).

Figure 3: Probability of a pause in September Arctic sea-ice extent.
Probability of a pause in September Arctic sea-ice extent.

a, September Arctic sea-ice extent anomaly relative to the 1980–2010 period for the CMIP5 models historical and three RCP experiments. b, Probability of a 7-year pause over a 21 year rolling window. c, Probability of a pause as a function of pause length in the Historical-RCP4.5 experiment over 1979–2013 (black), and in the future over 2066–2100 under the RCP2.6 (blue), RCP4.5 (cyan) and RCP8.5 (red) experiments. The horizontal dashed line represents a probability of p = 0.05. A pause is a period with a trend ≥0. Only ice extents ≥1 × 106 km2 are considered.

How important is internal variability in model projections of future sea-ice extent, relative to other uncertainties? To visualize this we use an 'uncertainty cascade'18 (Fig. 4) to show the mean sea-ice extent during four different 20-year periods and for the 6 CMIP5 climate models that have multiple realizations for each scenario and at least one pentad overlapping the observations (see Supplementary Information). The cascade represents the uncertainty due to the choice of emissions scenario, model, realization, and pentad (5-year mean) in descending order from the top, for each period.

Figure 4: Cascade of uncertainty in CMIP5.
Cascade of uncertainty in CMIP5.

September sea-ice extent is shown at four different levels of averaging (top left). i, The multi-model mean from three experiments (RCP2.6, 4.5 and 8.5), representing emissions scenario uncertainty. ii, The multi-realization mean from each of six models, representing model uncertainty. iii, The time-mean for each realization available. iv, For pentads (5-year means) from each realization, which along with iii represents internal climate variability. See Supplementary Information for a list of models used. Note that the range of extents in the full CMIP5 ensemble is considerably larger than for the six models shown here.

When accounting for internal climate variability, observed and simulated September Arctic sea-ice extent trends over 1979–2013 are not inconsistent. Internal variability can also either mask or enhance human-induced changes for decades at a time. Thus, pauses in sea-ice loss, such as seen over the past eight years, are not surprising and are fully expected to occur from time to time. Additional single model large ensembles that capture this variability would be valuable for advancing our understanding. Further evaluating the physical processes responsible for decadal variability in sea-ice extent in both observations and simulations will also improve our ability to understand how sea-ice is likely to evolve in the next few years and decades.

Corrected online 16 April 2015
In the Commentary ‘Influence of internal variability on Arctic sea-ice trends' (Nature Clim. Change 5, 86–89; 2015), in Fig. 3c, the x-axis label for pause length of 20 years was incorrectly repeated. Corrected after print 16 April 2015.
  1. IPCC Climate Change 2013: The Physical Science Basis (eds Stocker, T. et al.) (Cambridge Univ. Press, 2013).
  2. Holland, M. Nature Geosci. 6, 10–11 (2013).
  3. Holland, M. M., Bitz, C. M. & Tremblay, B. Geophys. Res. Lett. 33, L23503 (2006).
  4. Kay, J. E., Holland, M. M. & Jahn, A. Geophys. Res. Lett. 38, L15708 (2011).
  5. Day, J. J., Hargreaves, J. C., Annan, J. D. & Abe-Ouchi, A. Environ. Res. Lett. 7, 034011 (2012).
  6. Hawkins, E. et al. Nature 511, E3–E5 (2014).
  7. Hawkins, E., Edwards, T. & McNeall, D. Nature Clim. Change 4, 154–156 (2014).
  8. Deser, C., Knutti, R., Solomon, S. & Phillips A. S. Nature Clim. Change 2, 775–779 (2012).
  9. Taylor, J. Remember all those breathy predictions about an ice free Arctic By 2015? Nevermind... Forbes (12 September 2013).
  10. Revkin, A. C. Skeptics on human climate impact seize on cold spell. New York Times (2 March 2008).
  11. Overland, J. E. & Wang, M. Geophys. Res. Lett. 40, 2097–2101 (2013).
  12. Wadhams, P. in Sustainable Humanity Sustainable Nature our Responsibility (The Pontifical Academy of Sciences, 2014); http://go.nature.com/doRX4w
  13. Kay, J. E. et al. Bull. Am. Meteorol. Soc. http://dx.doi.org/10.1175/BAMS-D-13-002551 (in the press).
  14. Stroeve, J. C. et al. Clim. Change 110, 1005–1027 (2012).
  15. Stroeve, J. C. et al. Geophys. Res. Lett. 39, L16502 (2012).
  16. Santer, B. D. et al. Int. J. Climatol. 28, 1703–1722 (2008). URL:
http://www.nature.com/nclimate/journal/v5/n2/full/nclimate2483.html
Citation statistics:
资源类型: 期刊论文
标识符: http://119.78.100.158/handle/2HF3EXSE/4861
Appears in Collections:气候变化事实与影响
科学计划与规划
气候变化与战略

Files in This Item:
File Name/ File Size Content Type Version Access License
nclimate2483.pdf（480KB）期刊论文作者接受稿开放获取
CC BY-NC-SA
View Download

Recommended Citation:
Neil C. Swart. Influence of internal variability on Arctic sea-ice trends[J]. Nature Climate Change,2015-01-28,Volume:5:Pages:86;89 (2015).
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[Neil C. Swart]'s Articles
百度学术
Similar articles in Baidu Scholar
[Neil C. Swart]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[Neil C. Swart]‘s Articles
Related Copyright Policies
Null
收藏/分享
文件名： nclimate2483.pdf
格式： Adobe PDF
此文件暂不支持浏览
所有评论 (0)
暂无评论
 

Items in IR are protected by copyright, with all rights reserved, unless otherwise indicated.