Arguments
Human activity continues to warm the planet over the past 16 years
The skeptic argument...
No warming in 16 years
"...there has been no increase in the global average surface temperature for the past 16 years" (Judith Curry and David Rose)
What the science says...
| Select a level... |
Intermediate
|
Advanced
| |||
|
Once natural influences, in particular the impact of El Niño and La Niña, are removed from the recent termperature record, there is no evidence of a significant change in the human contribution to climate change. |
|||||
Update 21/02/2013: Troy Masters is doing some interesting analysis on the methods employed here and by Foster and Rahmstorf. On the basis of his results and my latest analysis I now think that the uncertainties presented here are significantly underestimated, and that the attribution of short term temperature trends is far from settled. There remains a lot of interesting work to be done on this subject.
Humans have continued to contribute to the greenhouse warming of the planet over the past 16 years at a rate statistically indistinguishable from preceding decades. The myth arises from two misconceptions. Firstly, it ignores the fact that short term temperature trends are strongly influenced by natural as well as human factors, and so are uninformative with respect to human contributions. Secondly it focuses on one small part of the climate system (the atmosphere) while ignoring the largest part (the oceans). We will address each of these errors in turn.
The human contribution to the 16 year trend
The instrumental temperature record shows continuing warming over the past 16 years, although at a reduced rate. In the case of GISTEMP at least the reduction is not statistically significant. The myth commonly claims that there has been no warming for 16 years. It is sometimes coupled with an explicit conclusion concerning the existance of anthropogenic global warming (e.g. in articles discussed here and here), however the conclusions are sometimes left to the reader without providing the contextual information to enable the reader to draw an informed conclusion.
There are multiple problems with the myth. The first is that testing for a statistically significant trend is the wrong approach - it can only falsify the claim that there has been no warming. A better test would be to check for a statistically significant deviation from a long term trend or projection.
However this is still a straw man argument. Climate science does not claim that all climate change is anthropogenic - there are both natural and human causes. The natural causes are particularly important over short time scales, and so any attempt to analyse short term trends must take them into account. The '16years' video demonstrates graphically how the natural and human factors can be separated, and thus what has been happening to the human contribution.
In order to address the audience of the media myths the language in the video has necessarily been heavily simplified. Additional information is provided in the video description and in the following sections.
The results of this analysis are consistent with a statement by WMO Secretary-General Michel Jarraud:
"Naturally occurring climate variability due to phenomena such as El Niño and La Niña impact on temperatures and precipitation on a seasonal to annual scale. But they do not alter the underlying long-term trend of rising temperatures due to climate change as a result of human activities"
Method
When using short time spans of less than two decades, natural variations can easily overwhelm greenhouse warming. In order to determine whether there has been a change in the human contribution to climate change over the past 16 years, it is necessary to isolate the human contribution from the natural variations. To do this, we must estimate the size of the natural variations in the temperature record, and subtract them from the temperature signal.
To do this we use the technique of multivariate regression, in which we construct timeseries for all the expected influences on climate. We then determine a value for the size of each contribution so that the sum of all the contributions gives the best approximation to the observed temperature data.
The method employed here is related to that used by Foster and Rahmstorf (2011), but with a couple of differences. Rypdal (2012) notes that the chance synchronization of several major volcanoes with declining solar activity, coupled with the longer term temperature impact of the volcanic cooling, may lead to the volcanic and solar terms being out of balance. To address this issue the volcanic and solar terms have been placed on the same scale (i.e. forcing in W/m2) and combined. Instead of time-shifting the combined term, an exponential lag is used to capture the longer term temperature impacts. The resulting model has two fewer parameters than that of Foster and Rahmstorf but still captures the bulk of the variation. As expected the volcanic contribution is somewhat increased and the solar contribution somewhat decreased in comparison to the Foster and Rahmstorf approach.
All calculations are performed using monthly data, however a 12-month moving average has been used for presentation of the graphs. Autocorrelation has been taken into account when calculating the statistical significance of trends.
Further tests for a change in trend
The calculation presented above follows the methodology of Foster and Rahmstorf in initially modelling the human contribution as a linear trend in order to determine the size of the natural influences. There is a risk that the method will therefore be biased towards producing a linear trend as a result, although given the large data/parameter ratio the risk is small. Nonetheless, three additional tests were designed to address this problem.
In the first test a second trend term was added covering the period before 1997 only, explicitly allowing for a difference in the trend pre- and post-1997 when fitting the natural influences. The resulting trend for the post-1997 period is again highly significant. In this case the trend difference term does show a marginally higher trend prior to 1997, however the magnitude of the difference falls short of even the 1σ threshold, i.e. the difference in trend is not distinguishable from noise.
In the second method the regression model was fitted to the pre-1997 data only. The resulting coefficients were used to remove the natural contributions from the post-1997 data, and a trend calculated using just these data. Again the difference in trends falls short of even the 1σ threshold.
In the third approach, a two-box response function model was used to fit forcings and El Nino to temperatures on the whole record from 1880 to 2010. This avoids the assumption of a linear underlying trend. This very different approach leads to an almost identical estimate of the El Niño term and similar overall conclusions.
Other versions of the temperature record
GISTEMP was used for this analysis because it is the only dataset with global coverage, which is critical when determining recent temperature trends - see for example these posts.
Foster and Rahmstorf applied their methodology to both the near-global record from NASA, and the substantially less complete HadCRUT and NCDC datasets. The adjusted series are remarkably similar, however there is a risk that the natural climate influences are being used to incorrectly adjust for the known lack of coverage. It may be more informative to derive near-global versions of the HadCRUT and NCDC data, using kriging for example. This approach will be explored in coming months.
The rest of the climate system
Focusing on surface air temperatures also misses more than 90% of the overall warming of the planet (Figure 2).
Figure 2: Components of global warming for the period 1993 to 2003 calculated from IPCC AR4 5.2.2.3.
Nuccitelli et al. (2012) considered the warming of the oceans (both shallow and deep), land, atmosphere, and ice, and showed that global warming has not slowed in recent years (Figure 3).
Figure 3: Land, atmosphere, and ice heating (red), 0-700 meter OHC increase (light blue), 700-2,000 meter OHC increase (dark blue). From Nuccitelli et al. (2012).
References
- Foster and Rahmstorf (2011), Global temperature evolution 1979–2010 doi:10.1088/1748-9326/6/4/044022
- Rypdal (2012) Global temperature response to radiative forcing: Solar cycle versus volcanic eruptions doi:10.1029/2011JD017283
- Nuccitelli et al. (2012) Comment on Ocean heat content and Earth's radiation imbalance. II. Relation to climate shifts doi:10.1016/j.physleta.2012.10.010
Credits: Calculations and video: Kevin C. Voiceover: Daniel Bailey. Advice: The SkS team.
Last updated on 21 February 2013 by Kevin C. View Archives
This is probably the right level of detail for a 2 minute video or a basic article. But the intermediate and advanced articles could mention that the remainder is the human contribution plus weather, plus decadal and longer natural variability.
Are most modes of decadal and longer natural variability internal to the climate system, or do they involve radiative forcings that haven't already been subtracted in the video?
If they're mostly internal, I think this point is already indirectly addressed via your graph of heat content. Internal climate variability should swap heat between the ocean and the surface, but all parts of the climate are warming. This could place a bound on the percentage of the surface warming trend which could be due to natural internal climate variability.
While limited to annual data and finishing at 2010, the model shows the same slowdown post 1998, and for the same reason as in the video - the trend in ENSO. In fact the ENSO term is almost identical (marginally larger) to the value used in the video.
This very simple model (20-30 lines of R or python) gives a very good fit of temperature from forcing and ENSO without invoking any multidecadal oscillations with an R2 or 92%. On the basis of this analysis at least there is no justification for invoking longer term climate cycles.
That would seem to settle the issue, however the case isn't completely closed. The result does depend on the choice of forcings. If instead you use the Potsdam forcings, the ENSO term is the same so the conclusions of the video are unaffected, but there is room for a small contribution from a multidecadal oscillation. I've been looking into the differences in forcings and understand some of the issues, but there are others I need to track down.
One other slight complication - there was a slight reduction in the forcing trend in the early '90s, I believe related to the phaseout of CFCs. That should also produce a slight change in temperature trend. But it's probably too small to detect over a 20 year period.
As I understand it, aerosols include particulate matter. Over the past couple of weeks we have seen news about air pollution in China as they close down factories and limit automobiles in the capital. Today, Japan is complaining about the air pollution coming over from China. How much of the aerosol load which is wafted up into the atmosphere is from this source and do we have any information on whether the load of aerosols in the upper atmosphere has been increasing along with China's increased manufacturing. I have heard an estimate that if we stoped the production of all aerosols, we might have as much as a 20C rise in temperature. A sobering thought if China (and the rest of us) cleaned up our act. Was the temporary flattening of the temperature record following the 40's due to American air pollution which they then cleaned up,
http://www.aip.org/history/climate/aerosol.htm
Yes, it seems probably that the aerosol cooling effect has been increasing. Unfortunately the effect is geographically dependent and not well measured.
The point of the video is that at this point I don't think we can detect that effect in the instrumental temperature record with any confidence. (There's an update coming which will show a small change, but still in the noise range.)