
Robust global ocean cooling trend for the pre-industrial Common Era
McGregor HV, Evans MN, Goosse H, Leduc G, Martrat B, Addison JA, Mortyn PG, Oppo DW, Seidenkrantz M-S, Sicre M-A, Phipps SJ, Selvaraj K, Thirumalai K, Filipsson HL & Ersek V
Nature Geoscience, vol. 8(9), 671-677, 2015https://doi.org/10.1038/ngeo2510
A contribution to the Ocean2k Working Group - Low Resolution team.
Additional information to accompany:
“Robust global ocean cooling trend for the pre-industrial Common Era”
McGregor et al.
Nature Geoscience
Published online 17 August 2015
DOI: 10.1038/NGEO2510
Access the article here
Access the supplementary information here
Access the PAGES press release here
The subset of metadata and data used in this analysis, the criteria by which they were selected, and the compositing code that produce Fig 2a and Supplementary Fig 1a, are available at the NCEI/WDCA repository: http://www.ncdc.noaa.gov/paleo/study/18718
Frequently asked questions
1. What's new about the observation of a 2,000 year cooling in a global average sea surface temperature estimate?
This is the first study that synthesizes only surface ocean reconstructions. The world's oceans absorb vast amounts of heat and as such regulate the climate on long timescales. Understanding how this major component varies, thanks to the ocean synthesis, means we can compare different parts of the climate system to see how they interact. Furthermore, by comparing the ocean synthesis to climate model simulations we learn more about what drives global climate variations (e.g. volcanic forcing, greenhouse gases, solar variations).
Volcanoes
2. How do volcanic eruptions cause the climate to change?
If volcanic aerosols are injected into the stratosphere, they may spread rapidly around the globe. The primary effect of these aerosols is cooling, as they reflect solar radiation to space before it becomes part of Earth's radiation balance. The cooling lasts as long as the aerosols are not removed from the stratosphere, which is generally in the order of a few years (Alan Robock's website for many citations; Hansen et al. 1992 link).
3. Were there lower rates of volcanism in the last 100-200 years?
See Figure 4c of the paper. It appears that the last 200 years were not unusual in terms of rates of volcanic activity than previous 200-year intervals in the past millennium. Below is the timing of the 40 largest eruptions since 800 CE according to Sigl et al. (in his Fig. 3, 2015, link).
5 in the 1801-2000 CE interval
4-5 in the 1601-1800 CE interval
3-4 in the 1401-1600 CE interval
5 in the 1201-1400 CE interval
3 in the 1001-1200 CE interval
3 in the 801-1000 CE interval
The 1801-1815 period appears to include 2 of the larger events, although definitely not the largest events over the 801-2000 period.
4. Could there still be a cooling signal from the volcanic forcing even with greenhouse gases causing warming?
Yes; to first order, the response to volcanic forcing and to greenhouse forcings are independent, meaning they can superimpose on one another, reinforcing or cancelling out the response of one by the other. For the 20th Century greenhouse gas warming swamps any volcanic cooling signal.
5. Does this study mean that if there are a lot of volcanic eruptions we can stop global warming?
Volcanic eruptions counteract the warming effects of greenhouse gases but we can't predict when volcanic eruptions will occur. Alan Robock (Rutgers University, USA) and others have shown that large volcanic eruptions can cool the climate (link), but in their report to the US National Academies of Science (link) they concluded that mitigation was cheaper than geoengineering solutions. Aerosols from volcanic eruptions contribute to natural climate variability, but global warming is occurring on top of natural climate variability, and the associated natural forcings do not appear to explain the observations for the 20th century (Stott et al. 2006; Bindoff et al. 2013).
6. Can volcanic eruptions be triggered to stop global warming?
This is well beyond our expertise, and not only likely to be unfeasible, but also unpredictable and uncontrollable.
7. So all these volcanic eruptions that stop flights are good for the planet?
Depending on the amount of aerosols released and to what altitude they are injected, they can counteract the warming effect of greenhouse gases for regions.
In the troposphere the aerosols may create health hazards, as for instance in the aftermath of the Mt St Helens eruption in the USA in 1980. If the aerosols get into the stratosphere and mix globally, they can cool the climate for the subsequent year (e.g. Hansen et al. 1992). However, the greenhouse gases stay in the atmosphere for decades to centuries, even if we stop emitting them today (Ciais et al. 2013 (link); Collins et al. 2013 (link), and references therein).
Furthermore, the global temperature is just one aspect of the problem. Volcanic eruptions may reduce the temperature and partly compensate for global warming at the global scale for a few years but the effects of anthropogenic and volcanic forcings can be very different at local scale and for precipitation, leading to no compensation or even adding to the effects in some cases.
8. How is this study different from previous work?
A millennial cooling was observed by Mann et al (1999 link) in their analysis and reconstruction of Northern Hemisphere mean surface air temperature anomalies for 1000-2000 CE from terrestrial data sources (their Fig 3).
They speculated this cooling was related to orbital forcing, and they suggested that centennial variations arose from solar irradiance forcing (Mann et al 1999 link). Crowley (2000 link) suggested that volcanic forcing might be associated with the Little Ice Age cooling in northern hemisphere mean temperature anomaly, but showed (in his Fig 5) that the residual pre-anthropogenic trend (1000-1850 CE) was not different from zero.
Distinct from these interpretations and that of Liu et al (2014 link), which also attributed the last millennium cooling to solar and volcanic forcing, our single and cumulative forcing analysis (Fig 4) suggests that a change in the frequency of volcanic activity, and possibly also land use forcing are the primary forcing factors associated with the cooling trend.
9. What was the influence of volcanism on global temperatures during the Medieval Climate Anomaly and in the 20th Century?
If the Medieval Climate Anomaly (MCA, also known as the Medieval Warm Period) is defined as a warmer period within the interval 900-1200 CE, then this appears to have been a period without much volcanic activity (Fig 4c) and shows no anomaly with respect to the 0-2000 CE period (Supp Section 7, Table S13).
This is partly an artifact of its timing in the middle of our study period and the data standardization procedure. The scaled results in Supp Fig S6 for the 801-1800 CE period also suggest the MCA was slightly warmer (~0.2° C) than the remainder of the pre-anthropogenic interval 1201-1800). It's not inconsistent to expect a lack of volcanic events to result in an increase in warming due to a reduction in the reflectivity of the atmosphere.
The 1801-2000 CE period has volcanic activity not very different from that of the 1201-1800 CE period, but shows a small but significant warming from the previous century (Supp Info, Section 7, Table S13).
Large-scale climate variability
10. Why doesn't the Medieval Climate Anomaly appear to occur globally in your results, and why does the Little Ice Age appear to occur globally?
Several studies have suggested that the so-called Medieval Climate Anomaly (a period of pervasive warming in the Northern Hemisphere from around 900-1200 CE) was actually not homogenously warm at the global scale (PAGES 2k Consortium 2013 (link)). Maximum temperature appeared to occur at different times in different regions, with some regions not displaying any particular temperature increase compared to the long trend during that period. Furthermore, when averaging over wide regions and 200-year time scales, as we did in this study, the maximum temperatures appear less clearly. Our findings are in line with the PAGES 2k Consortium's analysis of terrestrial temperature reconstructions, which also found the Medieval Climate Anomaly did not have a global warming signal. This further shows the value of ocean records which can provide a global temperature signal, which supports previous analyses.
The cooling is more homogenous for the Little Ice Age and thus clearer in our results. Our study confirms the role of volcanism in the cold conditions during the Little Ice Age. However, the results of single forcing experiments (Fig 4b) suggest that land use forcing may also be a cause of the LIA cooling we observe. A reduction in solar irradiance has been suggested to play a role too but, in accordance with many other studies, we were not able to detect its influence and the role of solar forcing was not required to explain the reconstructed trends.
The data, the analysis, and the uncertainties
11. Where are the data? I want to re-examine this work.
All results are derived from publicly archived datasets associated with peer-reviewed publications, and were found by searching the metadata of the NCEI and PANGAEA repositories.
The subset of metadata and data used in this analysis, the criteria by which they were selected, and the compositing code that produce Fig 2a and Supplementary Fig 1a, are available at the NCEI/WDCA repository: http://www.ncdc.noaa.gov/paleo/study/18718.
12. What about the proxies used in the reconstructions - what time periods did they show, what was the global distribution, and what can and can’t they tell us?
The reconstructions we gathered were themselves derived from measurements of the C37 alkenone unsaturation index (UK37) in sediments, Mg/Ca ratio in foraminifera, the TEX86 organic index, Sr/Ca in coral, relative abundance of Dinoflagellate cysts in marine sediments, and planktonic foraminiferal assemblages in marine sediments.
Depending on the nature of the observation, the paleoclimatic archive, and the choices made in sampling, they can tell us about local-to-regional SST on seasonal to bicentennial time resolution, with a response that is seasonal to annual in nature (Supp Info, Section 6). They cover at least some time period within the 0-2000 CE period (Supp Info Section 1), their spatial distribution is shown in Fig 1 (white circles) and in Supp Info Section 6 by data type and categorization (see Fig 2b and S9).
Because the observing network is sparse and includes multiple data types, resolutions, chronology resolution and accuracy, they are best considered, to first order, in the whole rather than regionally, and at coarse rather than fine time resolution.
13. Why was the data combined into 200-year brackets?
The dataset was composed of individual records that had time resolutions of 1 to 2 measurements per 200-year period, to hundreds per 200-year period (Supp Figure 1a). Each record was averaged into the consecutive 200-year brackets, which gave the same number of data points for each record (one for each 200-year window for the past 2000 years = 10).
To allow for the most inclusive (largest) dataset and focus on variations on century/longer timescales, and to avoid biasing the results toward records with higher resolution, we averaged all records to 200-year resolution. In addition, the 200-year brackets also accounted for how well we know the age of the variations.
14. Isn't the network of reconstructions used for the synthesis biased toward the North Atlantic and the periphery of the ocean basins? How could it be representative of a global change?
The results are certainly biased toward these regions. But for century-scale averages, variations are smoothed in space, probably because of stirring by winds from day to day of surface ocean. We can see this from analysis of climate simulations at 200-year average resolution for the observing network compared to the "true" average from the same simulations (Fig 1, Supp Fig S7).
15. How robust are the results, what are the limitations and assumptions in the reconstructions, analyses and models?
Each individual sea surface temperature reconstruction is subject to uncertainty in its dating, in the precision and accuracy of the observed measure, and in the precision and accuracy of the conversion of the observation into an SST reconstruction. If these uncertainties are independent across records because the records were produced independently, then compiling the records together should mean there is less uncertainty in composite median measures than in the individual records. Because the ocean, and sedimentation processes tend to smooth and integrate changes over centennial timescales, we think the observed multicentennial cooling trend is robust to the uncertainties described by Fig 2 of the paper.
The individual model simulations are subject to uncertainty arising from imperfect knowledge of the forcing functions, limited resolution in space, time, and of some processes operating in the climate system. Although the models are all forced by similar external radiative forcing series, and are therefore all subject to uncertainties, we rely on a composite across several simulations that mimics the available observations in time and space to make a fair comparison with the observations.
The warming trend over the last 200 years
16. Does this composite agree with direct historical observations of sea surface temperature over the last 200 years?
This composite differs from many other paleoclimate reconstructions published recently (e.g. Mann et al. 2008, PAGES 2k Consortium, 2013 (link)) in that it is a composite of local sea surface temperature reconstructions at 200-year resolution, rather than a reconstruction of large-scale patterns or averages at annual resolution, and age uncertainties are therefore quite large. Because of this, the dataset and analysis in this paper are not designed for looking at detailed changes within the past two centuries.
In contrast to the terrestrial reconstructions (including the PAGES 2k Consortium (2013) paper, which show contrasting trends within its analysis), the global ocean records show the same trajectory across all of the records, i.e. the oceans smooth out short-term atmospheric variability and are arguably a more accurate way of providing a measure of long-term global temperature (hence the significance of this study)
But our best attempt at this suggests qualitative agreement between the Ocean2k Low Resolution synthesis and historical observations (more details can be found in the Supplement to this paper). We do find that the change from one 200-year period to the next is either a cooling or no change - with the exception of the change into the most recent 200-year period (see Fig 2 in the paper; or Table S13 in the Supplement).
17. Could the warming found in the last ~200 years just be the Earth's climate naturally correcting itself?
To the best of our knowledge volcanic eruptions are not predictable very far in advance, it is unlikely the Earth is actively modulating its climate in this way. But could the recently observed warming merely be a recovery from the volcanically induced cooling? Some part of the recent observed warming is possibly recovery from pre-anthropogenic volcanic activity, but it is unlikely to explain all of it. Here is why:
1. Our best estimates of the cooling trend from the Ocean2k observations (Supplementary Section 4; Table S6; Fig S6) are about -0.3 to -0.4° C per 1000 years for the 2000 year long interval and about -0.4 to -0.5° C per 1000 years for the pre-anthropogenic last millennium.
2. The time for mean surface temperature to recover from individual events is around a few years (e.g. Hansen et al. 1992 (link)). It appears that clustering of large events within a small amount of time is what created the cooling trend over the time that we observed.
3. From historical observations, realistically forced simulations, and detection and attribution studies (e.g. Stott et al. 2006 (link); Bindoff et al. 2013 (link), Ch 10, Fig. 10.1, pg 879), our best estimate is that since 1860, mean surface temperature has increased about 0.5 to 1.5° C over 150 years, or about 0.3 to 1° C per 100 years. This is despite some significant recent volcanic events (Agung, El Chichón and Pinatubo).
4. The recent warming is larger ([|~1° C / ~-0.5° C| = 2) and more rapid by a factor of roughly 20: (| [~ +1° C/100 years]/[~ -0.5/1000 years] | = 20) than the volcanically induced cooling trend over the last 1-2 preceding millennia.
5. We don't have a mechanism to plausibly amplify a 'volcanic rebound' as required by these results, but we do have plausible mechanisms for explaining the observations using realistically forced climate simulations for the 20th century. When we compare the observed global mean surface air temperature change to that simulated when there is no human-caused greenhouse gas forcings, the simulation doesn't accurately produce the level of warming we have observed in the 20th Century.
18. When exactly in the 1800s did the warming trend start?
Based on our results we are unable to answer this question for the reasons outlined in Question 16 above: We have compiled here a composite of local SST reconstructions at 200-year resolution, rather than a reconstruction of large-scale patterns or averages at annual resolution, and therefore the age uncertainties are quite large. Because of this, the dataset and analysis in this paper are not designed for looking at detailed changes within the past two centuries.
19. Why doesn't your study have a stronger global warming signal?
We averaged our data into 200-year windows. A sharp change, as induced by global warming, is 'averaged out' in this process.
There are other studies based on annual growth bands in corals that show the year-to-year changes in detail over the past 400 years and these studies show a distinct and persistent global warming trend. See the recent Ocean2k Paleoceanography paper by Tierney et al. 2015 (link).