by Judith Curry
A new study finds that human-caused warming in the west tropical Pacific was not responsible for a series of frigid North American winters experienced over the early 2000s.
A number of papers have been published in recent years arguing that global warming — via declining Arctic sea ice — is causing increased snowfall and colder winters in the Northern Hemisphere.
A 2014 paper published in Science by Tim Palmer entitled Record breaking winters and global climate change argued that rising greenhouse gas emissions may have played a role in the severe 2013-2014 winter in the U.S. midwest via a mechanism whereby global warming has caused changes in the location of the jet stream tied to warming of the surface waters in the tropical west Pacific.
A new paper by Sigmond and Fyfe contradicts Tim Palmer’s notion the notion that tropical Pacific changes have increased the probability of the unusually cold winters observed in recent years.
Tropical Pacific impacts on cooling North American winters
Michael Sigmond & John C. Fyfe
Abstract. The North American continent generally experienced a cooling trend in winter over the early 2000s. This cooling trend represented a significant deviation from expected anthropogenic warming and so requires explanation. Previous studies indicate that climate variations in the tropical Pacific contributed to many mid-latitude climate variations over the early 21st Century. Here we show using large ensembles of fully-coupled, partially-coupled and uncoupled model simulations that in northwest North America the winter cooling was primarily a remote response to climate fluctuations in the tropical Pacific. By contrast, in central North America the winter cooling appears to have resulted from a relatively rare fluctuation in mid-latitude circulation that was unrelated to the tropical Pacific. Our results highlight how decadal climate signals – both remote and local in origin – can together offset anthropogenic warming to produce continental scale cooling.
Published in Nature Climate Change [link]. The paper can be read on readcube [link]
The methods used in the paper combine global observations, and a series of climate model experiments that include fully coupled, partially coupled (constrained by observations) and uncoupled (atmosphere only) simulations. Excerpts:
Having established that the models used in our study reproduce the observed linkages between the tropical Pacific and North America, we now investigate their simulated trends in the early 21st Century. First we employ the ensemble of fully-coupled CanESM2 simulations to explore potential links between variations in the simulated ΔSAT (surface air temperature) trend and variations in the simulated SAT trend averaged over NWNA (northwest North America). [There is] a positive trend in ΔSAT (corresponding to a weakening zonal temperature gradient) and a relatively small cooling trend over NWNA. This indicates that had the simulated tropical Pacific variability in the model been aligned in time with that observed that it would have likely simulated the observed cooling over NWNA. This is our first line of evidence that the observed winter cooling over NWNA from 2001-2002 to 2013-2014 was a remote response to decadal changes in the tropical Pacific. Our second line of evidence is obtained from an ensemble of partially-coupled CanESM2 simulations where surface wind stress in the tropics is constrained to follow its observed monthly evolution from January 1979. In all of these so-called “pacemaker” experiments the observed trade wind intensification and associated ΔSAT increase is associated with cooling surface temperatures over NWNA. These results are consistent with relationships that exist between winter-to-winter fluctuations and 14-year trends in ΔSAT and NWNA SAT.
Indeed, it appears that the Aleutian Low weakening observed over the early 21st Century, arguably the most pronounced feature in recent Northern Hemisphere SLP (sea level pressure) trends, can be attributed to tropical Pacific climate variations. Moreover, the associated cold air temperature advection was sufficient to overcome the externally-forced warming to cause cooling over NWNA. By contrast, tropical Pacific related SLP trends over the east coast of North America drove southwesterly winds to produce warming, rather than cooling, over CNA (central North America). Hence, tropical Pacific variability cannot explain the observed winter cooling over CNA.
It is clear that tropical Pacific SAT trends played an important role in early 21st Century mid-latitude climate trends, particularly over NWNA. We ask if this is due to cooling in the central to east Pacific, or due to warming in the west tropical Pacific? To address this question we employ uncoupled simulations using CanESM2. We perform control runs with prescribed SST and sea-ice averaged over the period between 1997 and 2007, and simulations that use the control SST and sea-ice except over the key regions.
In response to warming in only the west tropical Pacific a slight warming of CNA is found, contradicting the notion that west tropical Pacific warming was responsible for the observed CNA winter cooling. Our analysis indicates that tropical Pacific changes during the early 21st Century produced a warming impact over CNA.
Using our uncoupled simulations, we investigate the speculation [of Palmer] that the probability of these unusually cold winters may have increased due to the tropical Pacific changes. To arrive at statistically robust conclusions, we extended the uncoupled control simulation and the simulation with SST changes in the tropical Pacific region, where the SST and sea-ice fields were taken from observations. We find that it is less likely to find a colder winter in the perturbed simulation than in the control simulation. Hence, our results contradict the notion that tropical Pacific changes have increased the probability of the unusually cold winters observed in recent years.
If the recent CNA winter cooling cannot be attributed to tropical variability, then what was its cause? To address this question we return to the large ensemble of coupled CanESM2 simulations and note that in three of 100 ensemble members a stronger than observed CNA winter cooling was simulated. Figure 6 shows the average simulated climate trends over the five ensemble members with the largest CNA winter cooling. As in observations this cooling is the result of northerly winds associated with a ridge of increased SLP over the west coast of North America and CNA. The composite shows that outside the mid-latitudes there are no climate trends that are substantially different from the forced response. This indicates that the observed CNA winter cooling over the early 21st Century was not a response to decadal changes in the tropical Pacific, but was instead the result of a local internally-generated fluctuation in circulation that was unrelated to the tropical Pacific.
Our analysis shows how remotely and locally generated decadal climate variations have offset anthropogenic warming to produce North American winter cooling over the early 21st Century. Similar to the slowdown in the rise of global-mean surface temperature this cooling is very likely to be a temporary feature, as future decadal variations could be opposite in sign and amplify anthropogenic warming of North American winters.
JC reflections
I featured this paper for several reasons. The first reason is that I really like the way the climate model experiments were designed, particularly in the use of simulations that are constrained by observations in key regions. This methodology is similar to that used by Kosaka and Xie discussed in a previous post Pause tied to equatorial Pacific surface cooling.
The second reason is that the U.S. winter temperatures drive energy demand, determining the price of natural gas (and profits for energy traders and regional power providers) and also the amount of carbon emitted in keeping the nation warm. My company’s energy sector clients are already asking about next winter’s temperatures, and whether we can expect a typical La Nina pattern. Depending on the situation, we can start to get a read on next winter’s ENSO sometime during the period May-July. For this year, all the signs are for ENSO to be in negative territory for next winter — whether this means La Nina or neutral (with negative shades) can’t be decisively determined at this point.
Global seasonal climate forecast models are getting more sophisticated, with an improving track record, but their skill is not very high at making a forecast 6 months out (for winter, initialized in early summer). Hence statistical seasonal forecasts of ENSO and long range weather forecast patterns are also made by forecast providers. Untangling the relevant signals from past observations often provides tantalizing relationships that seem to work for a few seasons, then stop working. Which of course implies that the complete climate dynamics in play weren’t adequately accounted for.
The key challenge is to integrate our understandings of the mechanisms and transitions of the modes of multi-decadal, interannual and seasonal climate variability so that we can develop better seasonal climate forecasts. Studies such as this new paper by Sigmond and Fyfe are providing key insights into the predictable versus unpredictable parts of seasonal climate variability.Filed under: Attribution