With or Without a “Pause” Climate Models Still Project Too Much Warming

Global Science Report is a feature from the Center for the Study of Science, where we highlight one or two important new items in the scientific literature or the popular media. For broader and more technical perspectives, consult our monthly “Current Wisdom.”

A new paper just hit the scientific literature that argues that the apparent pause in the rise in global average surface temperatures during the past 16 years was really just a slowdown. 

As you may imagine, this paper, by Kevin Cowtan and Robert Way is being hotly discussed in the global warming blogs, with reaction ranging from a warm embrace by the global-warming-is-going-to-be-bad-for-us crowd to revulsion from the human-activities-have-no-effect-on-the-climate claque.

The lukewarmers (a school we take some credit for establishing) seem to be taking the results in stride.  After all, the “pause” as curious as it is/was, is not central to the primary argument that, yes, human activities are pressuring the planet to warm, but that the rate of warming is going to be much slower than is being projected by the collection of global climate models (upon which mainstream projections of future climate change—and the resulting climate alarm (i.e., calls for emission regulations, etc.)—are based).

Under the adjustments to the observed global temperature history put together by Cowtan and Way, the models fare a bit better than they do with the unadjusted temperature record. That is, the observed temperature trend over the past 34 years (the period of record analyzed by Cowtan and Way) is a tiny bit closer to the average trend from the collection of climate models used in the new report from the U.N.’s Intergovernmental Panel on Climate Change (IPCC) than is the old temperature record.

Specifically, while the trend in observed global temperatures from 1979-2012 as calculated by Cowtan and Way is 0.17°C/decade, it is 0.16°C/decade in the temperature record compiled by the U.K. Hadley Center (the record that Cowtan and Way adjusted).  Because of the sampling errors associated with trend estimation, these values are not significantly different from one another.  Whether the 0.17°C/decade is significantly different from the climate model average simulated trend during that period of 0.23°C/decade is discussed extensively below.

But, suffice it to say that an insignificant difference of 0.01°C/decade in the global trend measured over more than 30 years is pretty small beer and doesn’t give model apologists very much to get happy over.

Instead, the attention is being deflected to “The Pause”—the leveling off of global surface temperatures during the past 16 years (give or take). Here, the new results from Cowtan and Way show that during the period 1997-2012, instead of a statistically insignificant rise at a rate of 0.05°C/decade as is contained in the “old” temperature record, the rise becomes a statistically significant 0.12°C/decade. “The Pause” is transformed into “The Slowdown” and alarmists rejoice because global warming hasn’t stopped after all. (If the logic sounds backwards, it does to us as well, if you were worried about catastrophic global warming, wouldn’t you rejoice at findings that indicate that future climate change was going to be only modest, more so than results to the contrary?)

The science behind the new Cowtan and Way research is still being digested by the community of climate scientists and other interested parties alike. The main idea is that the existing compilations of the global average temperature are very data-sparse in the high latitudes. And since the Arctic (more so than the Antarctic) is warming faster than the global average, the lack of data there may mean that the global average temperature trend may be underestimated. Cowtan and Way developed a methodology which relied on other limited sources of temperature information from the Arctic (such as floating buoys and satellite observations) to try to make an estimate of how the surface temperature was behaving in regions lacking more traditional temperature observations (the authors released an informative video explaining their research which may better help you understand what they did). They found that the warming in the data-sparse regions was progressing faster than the global average (especially during the past couple of years) and that when they included the data that they derived for these regions in the computation of the global average temperature, they found the global trend was higher than previously reported—just how much higher depended on the period over which the trend was calculated. As we showed, the trend more than doubled over the period from 1997-2012, but barely increased at all over the longer period 1979-2012.

Figure 1 shows the impact on the global average temperature trend for all trend lengths between 10 and 35 years (incorporating  our educated guess as to what the 2013 temperature anomaly will be), and compares that to the distribution of climate model simulations of the same period. Statistically speaking, instead of there being a clear inconsistency (i.e., the observed trend value falls outside of the range which encompasses 95% of all modeled trends) between the observations and the climate mode simulations for lengths ranging generally from 11 to 28 years and a marginal inconsistency (i.e., the observed trend value falls outside of the range which encompasses 90% of all modeled trends)  for most of the other lengths, now the observations track closely the marginal inconsistency line, although trends of length 17, 19, 20, 21 remain clearly inconsistent with the collection of modeled trends. Still, throughout the entirely of the 35-yr period (ending in 2013), the observed trend lies far below the model average simulated trend (additional information on the impact of the new Cowtan and Way adjustments on modeled/observed temperature comparison can be found here).

 

Figure 1. Temperature trends ranging in length from 10 to 35 years (ending in a preliminary 2013) calculated using the data from the U.K. Hadley Center (blue dots), the adjustments to the U.K. Hadley Center data made by Cowtan and Way (red dots) extrapolated through 2013, and the average of climate model simulations (black dots). The range that encompasses 90% (light grey lines) and 95% (dotted black lines) of climate model trends is also included.

The Cowtan and Way analysis is an attempt at using additional types of temperature information, or extracting “information” from records that have already told their stories, to fill in the missing data in the Arctic.  There are concerns about the appropriateness of both the data sources and the methodologies applied to them.  

A major one is in the applicability of satellite data at such high latitudes.   The nature of the satellite’s orbit forces it to look “sideways” in order to sample polar regions.  In fact, the orbit is such that the highest latitude areas cannot be seen at all.  This is compounded by the fact that cold regions can develop substantial “inversions” of near-ground temperature, in which temperature actually rises with height such that there is not a straightforward relationship between the surface temperature and the temperature of the lower atmosphere where the satellites measure the temperature. If the nature of this complex relationship is not constant in time, an error is introduced into the Cowtan and Way analysis.

Another unresolved problem comes up when extrapolating land-based weather station data far into the Arctic Ocean.  While land temperatures can bounce around a lot, the fact that much of the ocean is partially ice-covered for many months.  Under “well-mixed” conditions, this forces the near-surface temperature to be constrained to values near the freezing point of salt water, whether or not the associated land station is much warmer or colder.

You can run this experiment yourself by filling a glass with a mix of ice and water and then making sure it is well mixed.  The water surface temperature must hover around 33°F until all the ice melts.  Given that the near-surface temperature is close to the water temperature, the limitations of land data become obvious.

Considering all of the above, we advise caution with regard to Cowtan and Way’s findings.  While adding high arctic data should increase the observed trend, the nature of the data means that the amount of additional rise is subject to further revision.  As they themselves note, there’s quite a bit more work to be done this area.

In the meantime, their results have tentatively breathed a small hint of life back into the climate models, basically buying them a bit more time—time for either the observed temperatures to start rising rapidly as current models expect, or, time for the modelers to try to fix/improve cloud processes, oceanic processes, and other process of variability (both natural and anthropogenic) that lie behind what would be the clearly overheated projections. 

We’ve also taken a look at how “sensitive” the results are to the length of the ongoing pause/slowdown.  Our educated guess is that the “bit” of time that the Cowtan and Way findings bought the models is only a few years long, and it is a fact, not a guess, that each additional year at the current rate of lukewarming increases the disconnection between the models and reality.

 

Reference:

Cowtan, K., and R. G. Way, 2013. Coverage bias in the HadCRUT4 temperature series and its impact on recent temperature trends. Quarterly Journal of the Royal Meteorological Society, doi: 10.1002/qj.2297.