Sunday, February 12, 2017

Stepladders and pausebusters

Last week, The Mail on Sunday published an article called “Exposed: How world leaders were duped into investing billions over manipulated global warming data.”  In the article, author David Rose (whom I mentioned in my previous blog post) refers to a paper titled “Possible artifacts of data biases in the recent global surface warming hiatus,” written in 2015 by a National Oceanic and Atmospheric Administration (NOAA) research team led by Thomas Karl.  Rose suggests that the paper’s re-analysis of NOAA’s temperature records, which indicated that the warming between 1998 and 2015 was larger than initially thought, was the primary reason that the 2015 Paris agreement calling for international action to combat climate change happened.  He then goes on to say that the paper exaggerated the data because another NOAA scientist, John Bates, claimed that the publication of the paper was rushed in violation of NOAA’s rules.  The article made the rounds in the media quickly — I actually found out about it when a conservative friend of mine shared an article about the article on his Facebook feed — and Karl and his NOAA colleagues now have the increased attention of the unsympathetic House Committee on Science, Space, and Technology to contend with.  Committee Chair Lamar Smith (R-Texas) even went so far as to accuse Karl of "playing fast and loose with the data."  So what exactly happened, and what are the implications for global warming and the research behind it?

To begin to answer this question, we need go back to the temperature record.  Last week I posted a graph showing the evolution of global temperatures (from the NASA GISS data set, using bimonthly means to coincide with NOAA’s MEI data) since 1950.  There is a lot of short-term variation in the data, due largely to the cycle of El Niño and La Niña, but the overall increase is unmistakable.  The variation makes evaluating the evolution of the trend in temperature complicated, however, and very highly dependent on where you start and end the analysis.  If I were going to calculate a ballpark estimate of how the trend has changed from decade to decade, I would restrict the data to times when the El Niño cycle was mostly neutral, or when the Multivariate ENSO Index (MEI) was between +0.2 and -0.2.  I would also rule out points that were influenced by the major volcanic eruptions of El Chichon (1982) and Mt. Pinatubo (1991).  This would leave a graph that looks like this:
 Figure 1

The trend lines come from a least-squares regression of the best fit, with straight-line relationships over each decade, constrained so that each decadal line segment begins where the previous one ends.  What this shows is that the warming looks more like a stepladder than a steadily increasing parabolic curve, with some time periods warming more rapidly than others.  Is this a surprising result?  Well no, actually.  In the 1988 paper “Global Climate Changes as Forecast by the Goddard Institute for Space Studies Three-Dimensional Model,” written by a NASA GISS research team led by Jim Hansen, alternating periods of relatively rapid warming and relatively slow warming (due to natural variations beyond the El Niño cycle) can be clearly seen in the five-year running means in the model’s forecast.  (The publication of this paper coincided with Hansen’s trip to Washington to become the first scientist to testify in front of the U. S. Senate that global warming was real and required urgent action.)

So what’s the problem?  The trends displayed in Figure 1 are dependent on where I arbitrarily chose the starting and ending points.  In my graph, the trend for the decade of the 2000s is approximately 0.05ºC.  This is small compared to the 1990s (0.20ºC) and the 2010s (0.32ºC projected for the whole decade, as of the end of 2016), but not zero.  However, if you include all the data points and start the trend from the El Niño-enhanced peak in 1998, the trend over the 2000s would naturally be lower, and may even dip below zero if you choose the right endpoint.  It should be obvious (if not, see my previous post) that there are objective reasons not to start any trend analysis with the peak of a very strong El Niño, but climate science deniers created a narrative over the last decade that global warming had in fact stopped in 1998, or that at worst there was an extended “pause” or “hiatus” in the warming.  These terms were used so aggressively that they frequently appeared in the scientific literature, including the 2014 report of the Intergovernmental Panel on Climate Change (IPCC).  (Note: if you are trying to combat false narratives, do not, ever, concede to use the other side’s language.)  It was obvious by 2015 that warming had started to accelerate again, boosted partially but not entirely by another strong El Niño.  This should have rendered moot any discussion over whether there had been a pause or not.  But this is where the Karl et al. paper comes in.

The NOAA research team led by Karl worked with the NOAA temperature record.  They found that surface temperature measurements obtained on the ocean were different depending on whether they came from boats or whether they came from buoys.  When they corrected for the difference, the trend in their temperature record between 1998 and 2014 increased.  This caused an uproar — not in the climate science community, but among the community of climate science deniers, who sensed that their long-running narrative was in jeopardy.  Lamar Smith demanded that Karl produce the data and all internal communications regarding this paper.  The data were already readily available, but NOAA refused to hand over the emails.  Earlier this year, a research team led by Berkely scientist Zeke Hausfather published their results after conducting their own analysis on the NOAA data. Not only did they draw the same conclusions about the NOAA data, but confirmed that the adjustments were consistent with data from other temperature records.

That brings us to last week.  John Bates had led a team at NOAA that devised a protocol for publishing results based on NOAA data and archiving the data and the methods of analysis.  He felt that Karl had published his results before the quality of the data could be fully confirmed (it eventually was), in order to get the results out in advance of the Paris Conference.  Bates also didn’t like the manner in which the data were archived, even though plenty of data from other locations are archived similarly.  (The manner in which the data had been first archived was nothing that a researcher looking for data would object to; if anything it's very user-friendly.)  Keep in mind also that the delegates to the Paris Conference looked at hundreds of papers, and were already aware of the rapid increase in global temperatures in 2015 that was far more significant than the updated trend in the NOAA data.  But Bates recently aired his grievances in the forum for a blog that climate science deniers look at, and soon David Rose had fodder for an article, complete with much hyperbole and a comically misleading graph.  Rose repeatedly refers to the Karl et al. paper as the “pausebuster”; I had never heard that one before, but I have to admit that it’s catchy.

To make a painfully long story short, a useful paper was given extremely exaggerated importance by certain people because it messed with their narrative.  Accusations of data manipulation ensued.  These accusations have now been greatly magnified because another NOAA scientist who designed a protocol for publication decided to publicly complain, in a place where it would do maximum damage to Karl and his research team, that his process wasn’t perfectly followed.  In the end, nothing is changed about the science.  Temperatures increase more quickly over some time intervals and less quickly over others.  We already knew that.  But scientists are being accused of dishonesty, even when their analysis has been peer-reviewed and then independently verified in other peer-reviewed publications.  The accusations are coming not just from misguided bloggers and journalists, but from powerful people in Congress.

It does not have to be this way.  I am confident that most Republicans who take a long, hard look at what is going on will not continue to allow people who promote abject ignorance, to the point of witch-hunting scientists who publish results that contradict their false narratives, to represent them.  In other news this week, an older generation of Republican statesmen proposed a plan than includes a carbon tax to fight global warming.  This is exactly the kind of debate the two parties should be having: over how to address the very real problem of global warming.  Accepting reality should not be ideological.

(If you want to read a good analysis of how the data analysis process works in the context of this past week’s events, Zeke Hausfather wrote this the other day.)

(For updates on new posts, please click the "Follow" button.)

Friday, February 3, 2017

Breaking the "Icy Silence"

In December, the US House of Representatives Committee on Science, Space, and Technology posted a link to an article published on the Breitbart News Network suggesting that a recent drop in land temperatures showed that the past three consecutive years of record warmth were just part of a natural fluctuation.  Both the House Committee and Breitbart are notorious for their climate change skepticism.  The House Committee is probably best known for a patently embarrassing exchange on global warming with President Obama’s science advisor John Holdren in 2014; the exchange was ruthlessly satirized by Jon Stewart on The Daily Show, but the laughs have to be tempered by the fact that these people have been given responsibility over science policy in this country.  Breibart first came to my attention in 2009 when its founder, Andrew Breitbart, posted a tweet saying “Capital Punishment for Dr James Hansen.  Climategate is High Treason.”  Jim Hansen was the director at the time of the NASA Goddard Institute for Space Studies (GISS), a major center for climate change research where I also worked.  When somebody publicly calls for the execution of your boss, you tend to notice him a bit more.  Andrew Breitbart passed away in 2012, but his successor Stephen Bannon became an instrumental figure in Donald Trump’s successful campaign last year, and now appears to rank as the new President’s most trusted advisor.  So I wasn’t expecting a whole lot of firm substance from this particular article, written by James Delingpole and titled “Global Temperatures Plunge.  Icy Silence from Climate Alarmists.”  But while there is much about the article that is misleading, there is enough of an attempt to describe actual science in the article that I think a detailed response that goes beyond a curt dismissal is warranted.

The article begins with the statement that “Global Land temperatures have plummeted by 1°C since the middle of this year — the biggest and steepest fall on record.”  There are three problems with using land temperatures over roughly half a year (keep in mind, the article came out on November 30) to infer global trends.  The first, as was pointed out by many people, is that 70% of the Earth’s surface is ocean.  The second, as a few people pointed out, is that the ocean warms and cools less quickly than land does, so the magnitude of the drop in temperature would be reduced if oceans are included.  The third problem, which I’m surprised wasn’t brought up by more people, is that 67% of the Earth’s land is in the Northern Hemisphere.  What direction would you expect Northern Hemisphere temperatures to go between the middle of the year and the end of November?  So right away you have a number presented in such a way as to make it look far more significant than it really is, and I could understand why people wouldn’t have bothered with the article any further than here.

As far as the “icy silence” from “climate alarmists” was concerned, anybody who understands how global mean temperatures fluctuate would say that a drop in global mean temperatures from the recent peak was long expected as the strong El Niño finally faded and segued into the other side of the natural cycle, called La Niña.  But what makes is this article worth examining is that, instead of ignoring the perfectly natural explanation for the temperature variation, Delingpole actually identifies the cause correctly but then mistakenly concludes that it supports the idea that there is no warming trend.  He cites another skeptical author named David Rose, who is quoted as follows:

“Big El Ninos always have an immense impact on world weather, triggering higher than normal temperatures over huge swathes of the world. The 2015-16 El Nino was probably the strongest since accurate measurements began, with the water up to 3ºC warmer than usual. 

It has now been replaced by a La Nina event – when the water in the same Pacific region turns colder than normal.  This also has worldwide impacts, driving temperatures down rather than up.

The satellite measurements over land respond quickly to El Nino and La Nina. Temperatures over the sea are also falling, but not as fast, because the sea retains heat for longer.

This means it is possible that by some yardsticks, 2016 will be declared as hot as 2015 or even slightly hotter – because El Nino did not vanish until the middle of the year.  But it is almost certain that next year, large falls will also be measured over the oceans, and by weather station thermometers on the surface of the planet – exactly as happened after the end of the last very strong El Nino in 1998. If so, some experts will be forced to eat their words.”

I’m not actually certain what experts Rose is talking about.  Most of what he says outside of the last sentence is at least defensible, although there is room for some argument on the details.  (2016 appears to have been more than slightly warmer than 2015, for example.)  To illustrate the connection between the El Niño/La Ninã cycle (generally referred to in the scientific community as the El Niño Southern Oscillation, or ENSO) and global mean temperature, I’m going to show a few graphs.

Figure 1


Figure 1 contains the bimonthly mean anomalies, measured relative to the 1951-1980 mean, of global temperatures in the GISS data set.  The temperature curve is not a smooth, steady rise, but rather a series of peaks and relative minima oscillating back and forth around a general trend.  The most pronounced peak is indeed the most recent one, reaching a bimonthly mean value of 1.33 ºC above the 1951-1980 mean in February/March 2016.  The temperature anomaly dropped over half a degree back to 0.80ºC in June/July, which is pretty steep over a short period of time, but it has actually gone up a bit since then.


Figure 2



Figure 2 shows the corresponding bimonthly mean plot of a quantity defined by the National Oceanic and Atmospheric Administration (NOAA) as the Multivariate ENSO Index (MEI).  The mathematics of calculating the MEI is complicated, but it revolves around measurements of six different weather-related quantities in the tropical Pacific, and then defining the variance in these quantities relative to median conditions (defined as an MEI of 0).  The high positive values correspond to El Niño events, while the strongly negative values correspond to La Niña events.  As you can see, many of the temperature spikes in Figure 1 coincide with El Niño events, and many of the relative minima coincide with La Niña events.  This is why scientists consider the ENSO cycle to be the dominant source of short-term, natural variations in the global temperature record.

However, while the 2015-2016 El Niño event is among the strongest on record, the MEI data suggest that it was not stronger than the one in 1998.  That particular El Niño event produced its own sharp spike in the global temperature record.  It has also produced a large amount of misunderstanding of how global mean temperatures work, as skeptics have spent much of the last decade or so declaring the 1998 peak as the point where global warming stopped.  It should be clear enough from Figure 1 that there is more at work than the cyclical El Niño variations.  The highest bimonthly mean temperature anomaly recorded in 1998, a stronger El Niño than the most recent one according to the MEI data, was 0.75ºC.  Not only is that nearly 0.6ºC cooler than the most recent peak in the temperature anomaly from last year, but the 1998 peak was cooler than the planet is right now, even after the large drop.

As for the drop, the 2015-2016 El Niño has certainly ended, but the present state is closer to neutral than to a full-blown La Niña event.  This suggests that ENSO-neutral conditions presently result in a temperature anomaly at, or maybe a little bit above, 0.80ºC.  Were this state of general neutrality to continue for the rest of the year, 2017 would wind up comfortably being the third warmest on record, but that ultimately depends on whether or not a strong La Niña ultimately happens.

So does that mean, as Delingpole concludes in his article, that “The last three years may eventually come to be seen as the final death rattle of the global warming scare”?  No, of course not.  The overall trend in the temperature anomaly data continues to be positive.  The trend may appear larger or smaller at different times due to natural fluctuations, but it is there just the same.  If ENSO were the dominant influence on global mean temperatures without any additional warming trend, then the temperature anomaly from this past year would most likely not be higher at all than the one in 1998, much less 0.6ºC higher.

(Incidentally, I told my climate-related classes in 2015 that it would only be a matter of time before the same people who insist on using the peak of the strong 1998 El Niño as the starting point in their analysis of subsequent temperature trends would start publicly dismissing the much greater peak presently in the temperature record on account of its association with a strong El Niño.  And here you go.)

(For updates on new posts, please click the "Follow" button.)