Coyote Blog

I titled my very first climate video "What is Normal," alluding to the fact that climate doomsayers argue that we have shifted aspects of the climate (temperature, hurricanes, etc.) from "normal" without us even having enough historical perspective to say what "normal" is.

A more sophisticated way to restate this same point would be to say that natural phenomenon tend to show various periodicities, and without observing nature through the whole of these cycles, it is easy to mistake short term cyclical variations for long-term trends.

A paper in the journal Water Resources Research makes just this point using over 200 years of precipitation data:

We analyze long-term fluctuations of rainfall extremes in 268 years of daily observations (Padova, Italy, 1725-2006), to our knowledge the longest existing instrumental time series of its kind. We identify multidecadal oscillations in extremes estimated by fitting the GEV distribution, with approximate periodicities of about 17-21 years, 30-38 years, 49-68 years, 85-94 years, and 145-172 years. The amplitudes of these oscillations far exceed the changes associated with the observed trend in intensity. This finding implies that, even if climatic trends are absent or negligible, rainfall and its extremes exhibit an apparent non-stationarity if analyzed over time intervals shorter than the longest periodicity in the data (about 170 years for the case analyzed here). These results suggest that, because long-term periodicities may likely be present elsewhere, in the absence of observational time series with length comparable to such periodicities (possibly exceeding one century), past observations cannot be considered to be representative of future extremes. We also find that observed fluctuations in extreme events in Padova are linked to the North Atlantic Oscillation: increases in the NAO Index are on average associated with an intensification of daily extreme rainfall events. This link with the NAO global pattern is highly suggestive of implications of general relevance: long-term fluctuations in rainfall extremes connected with large-scale oscillating atmospheric patterns are likely to be widely present, and undermine the very basic idea of using a single stationary distribution to infer future extremes from past observations.

Trying to work with data series that are too short is simply a fact of life -- everyone in climate would love a 1000-year detailed data set, but we don't have it.  We use what we have, but it is important to understand the limitations.  There is less excuse for the media that likes to use single data points, e.g. one storm, to "prove" long term climate trends.

A good example of why this is relevant is the global temperature trend.  This chart is a year or so old and has not been updated in that time, but it shows the global temperature trend using the most popular surface temperature data set.  The global warming movement really got fired up around 1998, at the end of the twenty year temperature trend circled in red.

They then took the trends from these 20 years and extrapolated them into the future:

But what if that 20 years was merely the upward leg of a 40-60 year cyclic variation?  Ignoring the cyclic functions would cause one to overestimate the long term trend.  This is exactly what climate models do, ignoring important cyclic functions like the AMO and PDO.

In fact, you can get a very good fit with actual temperature by modeling them as three functions:  A 63-year sine wave, a 0.4C per century long-term linear trend  (e.g. recovery from the little ice age) and a new trend starting in 1945 of an additional 0.35C, possibly from manmade CO2.

In this case, a long-term trend still appears to exist but it is exaggerated by only trying to measure it in the upward part of the cycle (e.g. from 1978-1998).

In this post, I want to discuss my just-for-fun model of global temperatures I developed 6 years ago.  But more importantly, I am going to come back to some lessons about natural climate drivers and historic temperature trends that should have great relevance to the upcoming IPCC report.

In 2007, for my first climate video, I created an admittedly simplistic model of global temperatures.  I did not try to model any details within the climate system.  Instead, I attempted to tease out a very few (it ended up being three) trends from the historic temperature data and simply projected them forward.  Each of these trends has a logic grounded in physical processes, but the values I used were pure regression rather than any bottom up calculation from physics.  Here they are:

• A long term trend of 0.4C warming per century.  This can be thought of as a sort of base natural rate for the post-little ice age era.
• An additional linear trend beginning in 1945 of an additional 0.35C per century.  This represents combined effects of CO2 (whose effects should largely appear after mid-century) and higher solar activity in the second half of the 20th century  (Note that this is way, way below the mainstream estimates in the IPCC of the historic contribution of CO2, as it implies the maximum historic contribution is less than 0.2C)
• A cyclic trend that looks like a sine wave centered on zero (such that over time it adds nothing to the long term trend) with a period of about 63 years.  Think of this as representing the net effect of cyclical climate processes such as the PDO and AMO.

Put in graphical form, here are these three drivers (the left axis in both is degrees C, re-centered to match the centering of Hadley CRUT4 temperature anomalies).  The two linear trends (click on any image in this post to enlarge it)

And the cyclic trend:

These two charts are simply added and then can be compared to actual temperatures.  This is the way the comparison looked in 2007 when I first created this "model"

The historic match is no great feat.  The model was admittedly tuned to match history (yes, unlike the pros who all tune their models, I admit it).  The linear trends as well as the sine wave period and amplitude were adjusted to make the fit work.

However, it is instructive to note that a simple model of a linear trend plus sine wave matches history so well, particularly since it assumes such a small contribution from CO2 (yet matches history well) and since in prior IPCC reports, the IPCC and most modelers simply refused to include cyclic functions like AMO and PDO in their models.  You will note that the Coyote Climate Model was projecting a flattening, even a decrease in temperatures when everyone else in the climate community was projecting that blue temperature line heading up and to the right.

So, how are we doing?  I never really meant the model to have predictive power.  I built it just to make some points about the potential role of cyclic functions in the historic temperature trend.  But based on updated Hadley CRUT4 data through July, 2013, this is how we are doing:

Not too shabby.  Anyway, I do not insist on the model, but I do want to come back to a few points about temperature modeling and cyclic climate processes in light of the new IPCC report coming soon.

The decisions of climate modelers do not always make sense or seem consistent.  The best framework I can find for explaining their choices is to hypothesize that every choice is driven by trying to make the forecast future temperature increase as large as possible.  In past IPCC reports, modelers refused to acknowledge any natural or cyclic effects on global temperatures, and actually made statements that a) variations in the sun's output were too small to change temperatures in any measurable way and b) it was not necessary to include cyclic processes like the PDO and AMO in their climate models.

I do not know why these decisions were made, but they had the effect of maximizing the amount of past warming that could be attributed to CO2, thus maximizing potential climate sensitivity numbers and future warming forecasts.  The reason for this was that the IPCC based nearly the totality of their conclusions about past warming rates and CO2 from the period 1978-1998.  They may talk about "since 1950", but you can see from the chart above that all of the warming since 1950 actually happened in that narrow 20 year window.  During that 20-year window, though, solar activity, the PDO and the AMO were also all peaking or in their warm phases.  So if the IPCC were to acknowledge that any of those natural effects had any influence on temperatures, they would have to reduce the amount of warming scored to CO2 between 1978 and 1998 and thus their large future warming forecasts would have become even harder to justify.

Now, fast forward to today.  Global temperatures have been flat since about 1998, or for about 15 years or so.  This is difficult to explain for the IPCC, since about none of the 60+ models in their ensembles predicted this kind of pause in warming.  In fact, temperature trends over the last 15 years have fallen below the 95% confidence level of nearly every climate model used by the IPCC.  So scientists must either change their models (eek!) or else they must explain why they still are correct but missed the last 15 years of flat temperatures.

The IPCC is likely to take the latter course.  Rumor has it that they will attribute the warming pause to... ocean cycles and the sun (those things the IPCC said last time were irrelevant).  As you can see from my model above, this is entirely plausible.  My model has an underlying 0.75C per century trend after 1945, but even with this trend actual temperatures hit a 30-year flat spot after the year 2000.   So it is entirely possible for an underlying trend to be temporarily masked by cyclical factors.

BUT.  And this is a big but.  You can also see from my model that you can't assume that these factors caused the current "pause" in warming without also acknowledging that they contributed to the warming from 1978-1998, something the IPCC seems loath to do.  I do not know how the ICC is going to deal with this.  I hate to think the worst of people, but I do not think it is beyond them to say that these factors offset greenhouse warming for the last 15 years but did not increase warming the 20 years before that.

We shall see.  To be continued....

Update:  Seriously, on a relative basis, I am kicking ass

Want to get worked up about price collusion in the oil industry?  Don't waste your time.  No study has ever found collusion effects that raised US gasoline prices more than a few percent, and only for a very short period of time.  The reason is that in a free market, there is too much incentive for new entrants undercutting a price collusion attempt.  Railroads and airlines have probably the most severe economic incentives to collude, and they have never pulled it off for any period of time EXCEPT when the government stepped in to enforce the arrangement (e.g, airline controls pre-deregulation).

If you want to see a real cartel at work raising prices at the expense of consumers, check out this from the Mises Blog:

The raisin agricultural marketing order (AMO), with roots in the
Depression-era Agricultural Marketing Agreement Act, is rationalized as
a way to "stabilize" prices. However, it allows the Raisin
Administration Committee (RAC), controlled by producers, to determine
how much of each crop can be sold, with the rest forced into storage.
That power to jointly restrict output to raise price makes it a cartel.
A cartel with so many members would not usually succeed, and the mere
attempt would be prosecuted if antitrust laws were applied, but AMOs
are enforced by the government, through the USDA...

The RAC "stabilization" is accomplished by restricting sales, often
substantially. "Free tonnage" has been as low as 53% of the crop in
2001, and less than 80% in most years. That helps producers by harming
consumers, turning price "stabilization" into price enhancement....

The raisin cartel's effects on American consumers can also be seen
in the gap between the "free tonnage" prices and "reserve pool" prices
for raisins destined for low value markets. In 2001, those prices were
$877.50 per ton versus $250 per ton; in 1998, it was $1250 versus $357;
in 1984 and 1994, the differential approached 10 to 1.