The folly of extrapolating from short term trends

Climate change deniers routinely make the same mistake as the comic above.

Climate data are noisy.  We see wild fluctuations in temperature as the temperature rises through out day and cools at night.  On a longer (but still short) timescale we have the warming of spring and summer and the cooling of fall and winter. Thankfully no one is foolish enough (with a few exceptions) to use these trends to argue about climate change.  As George Monbiot points out:

If a UK cold snap persuades climate sceptics that global warming isn’t happening, then a heatwave must convince them that it is.

We’re still waiting. During the cold weather last winter, Gerald Warner, Peter Mullen and a host of other climate change deniers lined up to suggest that there must be something wrong with global warming theory, because some snow had fallen in Britain…

If a single cold snap in the UK persuades them that global warming isn’t happening, then a single heatwave in the same place must surely convince them that it is. Logic would dictate that the world must now be destined for a century of heating – until the next cold snap, whereupon it is obviously destined once more for a century of cooling.

Anyone who uses a single weather event to argue for or against climate change can be safely ignored. They either don’t know what they are talking about, or are being intentionally dishonest.

But just as it is immediately obvious that a single weather event says nothing about climate change, a timescale of several years (since 1998 perhaps) still isn’t enough to derive any firm conclusions. As a study Published in Geophysical Research Letters published earlier this made abundantly clear:

The reality of the climate system is that, due to natural climate variability, it is entirely possible to have a period as long as a decade or two of ‘‘cooling’’ superimposed on the longer-term warming trend due to anthropogenic greenhouse gas forcing… Claims that global warming is not occurring that are derived from a cooling observed over such short time periods ignore this natural variability and are misleading.

Tamino over at the always great (though frequently very statistical) Open Mind blog makes the point that when looking at any single decade (take you pick from 1970 to the present), the uncertainties are so great that the from these data we estimate global temperature change as somewhere between rapid cooling and oh-my-god-we’re-all-going-to-fry warming. Or in other words useless.

Perhaps it’s useful to take a look at what actual global temperature looks like on short time scales. Let’s look at some decades, the 1970s through the 2000s. I’ll plot all decades on the same scale for both axes; here are annual average temperatures from GISS for the 1970s:

Linear regression gives a positive slope, at 0.0065 +/- 0.0224 deg.C/yr, but the error limits are way too big to draw any meaningful conclusion and the visual impression of the graph doesn’t indicate warming or cooling, just a lot of jiggling around. That’s natural variation for you; a lot of jiggling around which makes trend estimates on short timescales too imprecise to be useful. The 1980s gives a nearly identical impression:

Again the linear regression slope is positive at 0.0067 +/- 0.0219 deg.C/yr, again the uncertainty is much larger than the estimate, and again the visual impression is neither warming nor cooling, just a lot of jiggling. For the 1990s we have:

The untrained eye may get the impression of a meaningful warming trend. But the linear regression trend rate is 0.0179 +/- 0.0276 deg.C/yr, so the error range is still considerably larger than the estimate. From these data, we’d estimate global temperature change as somewhere between rapid cooling (-0.0097 deg.C/yr) and oh-my-god-we’re-all-going-to-fry warming (+0.0455 deg.C/yr). For the 2000s we have:

This time the linear regression trend rate is 0.0126 +/- 0.0218 deg.C/yr, so once again the uncertainty is much larger than the estimate. It is worth noting that of these four decades, the 2000s don’t have the smallest linear regression trend rate, they have the 2nd-largest.

What’s the cure for “too short to tell” time spans? Longer time spans! Here’s the data from 1970 to the present:

Clearly a decade is too brief a time span to get a meaningful trend estimate; just as clearly the trend since 1970 is — how shall one say? — obvious. The linear regression trend rate is 0.0164 +/- 0.0028 deg.C/yr. I have good reason to believe that the “turning point” marking the start of recent warming is 1975 rather than 1970, but even with this earlier date the trend is statistically significant. Strongly. And it’s warming, not cooling.

Clearly we shouldn’t base any climate trend on a decade of data.

And it is worth noting that temperature aren’t the only noisy data that climate scientists deal with. Sea level, the heat stored in the oceans and the extent of Arctic sea ice, are all noisy data.

This should be obvious but deniers who definitely ought to know better, but it isn’t, as RealClimate points out. Roger Pielke Sr, should know better. He is a trained climatologist, yet he claims that:

  1. Since 2006 sea level rise has flattened.Has it? You be the judge:It doesn’t look flat to me. But more importantly it is obvious that these data are noisy. ‘2006 to the present’ is far to short a period to derive any meaningful conclusions.
  2. Pielke Sr. then claims that “There has been no statistically significant warming of the upper ocean since 2003”.Really? Again, you be the judge:
  3. Again it looks like an increase to me, and again it is rather obvious that ‘since 2003’ is too short a timescale to draw any meaningful conclusions:

    the trend in from 2003 to 2008 in the Levitus data (the Domingues et al data does not extend past 2003), is still positive but with an uncertainty (both in the trend calculation and systematically) that makes it impossible to state whether there has been a significant change.

  4. Peilke Sr.’s most dishonest statement is in regards to sea ice. He says that claims that the arctic sea ice is decreasing faster than model predictions are wrong because ‘since 2008, the anomalies have actually decreased’.Again here is the relevant data:
    Peilke justifies his claim based on less than one year of data! (the 2009 data isn’t in yet). Yet it is plainly clear that even though 2008 had more ice than 2007, it is still well below what is predicted by climate models.

This is simply unjustifiable for someone like Peilke Sr. He knows better; he is being deliberately dishonest.

Just as the comic at the beginning of this post is absurd, so are attempts by deniers to draw any meaningful conclusions on short timescales (especially when they lie, as Peilke Sr. has done). They both make the same fundamental error.

Leave a Reply

Proudly powered by WordPress | Theme: Baskerville 2 by Anders Noren.

Up ↑