This is part B of Chapter 5 of an ongoing series. Other parts of the series are here:
- Greenhouse Gas Theory
- A) Actual Temperature Data; B) Problems with the Surface Temperature Record
- Attribution of Past Warming: A) Arguments for it being Man-Made; B) Natural Attribution (this article)
- Climate Models vs. Actual Temperatures
- Are We Already Seeing Climate Change
- The Lukewarmer Middle Ground
- A Low-Cost Insurance Policy
In part A, we discussed the main line of argument for attributing past warming to man-made CO2. In essence, scientists have built computer models to simulate the climate (and global temperatures). When these models were unable to simulate the amount of warming that occurred in the two decades between 1978 and 1998 using only what they thought were the major natural climate drivers, scientists concluded that this warming could not have been natural and could only have happened if the climate has a high sensitivity to man-made CO2.
This argument only works, of course, if the climate models are actually a correct representation of the climate. And that can only be proven over time, by comparing climate model output to actual weather. Back in chapter 4A, we briefly discussed how actual temperatures are in fact not tracking very well with climate model predictions, which should throw a substantial amount of doubt on the current quality of climate models (though the media still tends to treat model predictions as authoritative).
In this section, we will focus on some of the natural factors that are missing from most climate models. Obviously, if important natural drivers have been left out of the models, then one cannot conclude from the inability of the models to match historical warming that the historical warming couldn't have been natural. After discussing some of these factors, I will take my owns swing at the attribution problem.
Long-term Climate Shifts
We will begin with long-term climate variations. These are most certainly left out of the models, because no one really understands why they occur (though theories abound, of course). Mann's hockey stick not-withstanding, the consensus picture of past climate continues to include a strong warming period in the Middle Ages and a cool period, called the Little Ice Age, in the 16th and 17th centuries.
Imagine you were a climate modeler in 1600. Your model would probably have under-predicted temperatures over the next 200 years, because you were trying to model starting at the bottom of a long-term cyclical trend. So clearly leaving this trend out in 1600 would get the wrong answer. Wouldn't leaving it out in the year 2000 also get the wrong answer? All too often scientists tend to assume (though not always explicitly) that this long-term natural recovery of temperatures ended around 1950, at the same time they believe man-made warming started. A metaphorical hand-off occurred from natural to man-made factors. But there is no evidence for this whatsoever. We don't know what caused the Little Ice Age, so we don't know how long it can last or when it ends.
Changes in the Sun
Since we have mentioned it, let's discuss the sun. The sun is the dynamo that, along with a few smaller effects like the rotation of the Earth, drives the climate. We have known for some time that the Sun experiences cycles of variation, and one of the ways one can observe this variation is by looking at sunspots. We have more sophisticated ways of measuring the sun today, but we still count the spots.
Sunspots are cyclical in nature, and follow an eleven or so year cycle (you can see this in the spikes in the monthly light blue data above). But when one take this cycle out of the picture, as was done with the 10.8 year moving average above, there also appears to be longer cyclical trends. Since it is generally thought that more sunspots correlate with higher solar activity and output, one might expect that there could be some correlation between this solar trend and temperatures. As we can see above, by the sunspot metric the sun was more active in the second half of the last century than in the first half.
Today, we don't have to relay on just the spots, we can look at the actual energy output of the sun. And it turns out that the types of variations we have seen over recent decades in sunspots do not translate to very large changes in solar output on a percentage basis. Yes, there is more solar output but the extra amount is small, too small to explain much temperature variation. There is, though, an emerging new theory that a complex interaction of the sun with cosmic rays may affect cloud formation, acting as a multiplier effect on changes in solar output. A lot of skeptics, eager to support the natural causation argument, jumped on this theory. However, though the theory is intriguing and could turn out to be correct, I think folks are getting well ahead of the evidence in giving it too much credence at this point.
At the end of the day, while solar variation may explain very long-cycle climate variations, it does not do much to explain our 1978-1998 warming period, so we will move on to another natural factor that does appear to have some explanatory power and which is also not in most climate models -- ocean cycles.
This is a complicated topic and I am far from an expert. In short: As mentioned in an earlier chapter, the oceans have far more heat carrying capacity than the atmosphere. It turns out that oceans have cycles, that are decades long, where they can exchange more or less heat with the atmosphere. In their "warm" periods, these cycles tend to leave more heat in the atmosphere, and in their "cold" periods they bury more heat in their depths. Once such cycle is called the Pacific Decadal Oscillation (PDO), which will be familiar to most Americans because "El Nino"and "La Nina" climate patterns are part of this PDO cycle. If one plots global temperatures against the PDO cycles, there is a good deal of correlation:
When the PDO has been in its warm phases (the red periods in the chart above), global temperatures rise. When it is in its cool phases (the blue zones), temperatures are flat to down. As you can see, the PDO was in a warm phase in our 1978-1998 period. Surely some of that steep rise in temperature may have come from the effect of this ocean cycle, yet this cycle was not included in the climate models that supposedly ruled out the possibility of natural causes for warming in this period.
A number of scientific studies have tried to remove these (and other) cyclical and event-based drivers from the historical temperature record. Here is one such attempt (ENSO and AMO are ocean cycles, large volcanoes tend to have a global cooling effect for a few years after their eruption)
With these natural effects removed, much of the cyclical variation from the Hadley CRUT4 data are gone, and we are left with a pretty constant linear trend. Aha! There is the warming signal, right? Well, yes, but there is a problem here for the effort to attribute most or all of this warming to man -- specifically, this is not at all the trend one would expect if the long-term trend were primarily from man-made CO2. Note the very linear trend starts around 1900, long before we began burning fossil fuels in earnest, and the trend is really quite flat, while man-made CO2 production has been growing exponentially. Supporters of man-made attribution are left in the uncomfortable position of arguing that there must have been natural warming until about 1950 which stopped just in time for man-made warming to take over.
My Attribution Solution
A number of years ago I decided to take a shot at the attribution problem, largely just for fun, but it turned out so well I still keep it up to date. I decided to assume just three factors: 1. A long term linear trend starting even before the 20th century, presumably natural; 2. A new added linear trend, presumably from man-made effects; and 3. A decadal cyclical factor, from things like ocean cycles. I let the optimization program control everything -- the slope of the linear trends, the amplitude and period of the cyclical factor, the start date of the second modern trend, etc, to get the best fit with historic temperatures. As before, I used monthly Hadley CRUT4 data.
This is what we ended up with. A 66-year sine wave:
Plus a long-term linear trend of 0.36C per century and a new linear trend beginning around 1950 that adds another 0.5C per century (for a total linear trend after 1950 of 0.86C per century).
The result was a pretty good fit 8 years ago and more importantly, still continues to be a good fit up to today (unlike much more complicated climate models)
Though the optimization was based on monthly data, you can see the fit even better if we add on a 5-year moving average to the chart:
That is, then, my solution to the attribution problem. Take the 0.5C per century since 1950 that this model has as a modern linear trend, and we will for argument sake attribute it all to man. From 1950-2016 (66 years, coincidentally my sin wave period) that is 0.33C of historic warming due to man-made CO2.
In the next chapter, we return to the climate forecasts we discussed in chapters 2 and 3 and ask ourselves whether these make sense in the context of past warming.