I’ve been perplexed for a little while about what to say.
Am I debating with:
- Someone who doesn’t think the Earth is warming?
- Someone who doesn’t think the warming is of sufficient concern or of the type which can be addressed by reducing human greenhouse gas emissions*?
- Someone that doesn’t fit into either above category?
[sub]*Not counting flatulence. Everyone loves some good farting – I’m mostly talking about industrial and transportation emissions.[/sub]
I understand the thing about CAGW – I don’t agree that it’s a useful term, but I understand it.
However, the implication that the Earth may not be warming at all – that current temperature trends are within observed random variation. That simply does not fit with what we know.
It’s useful to describe data in terms of different types of measurements – average, median, and mode are measurements of center of your data. They represent an attempt to describe the “middle” or the most common parts of your data set. Standard deviation and variance are measurements of spread - they’re an attempt to describe “how far is the data spread out?” “where is the rest of the data in relation to your measurements of center?” Measures of spread are essential in hypothesis testing, for example. If you know the general shape of your data’s distribution, you can calculate the percent chance that you would find a particular data point in a certain distribution with a known mean (mean = average) and spread.
Imagine a Texan firing away at a barn wall. After he shoots, you walk up to the inside barn wall and try to figure out what he’s shooting at. You could measure where the bullet holes are and then make a pretty good estimate of where our Texan has painted the face of Fred Phelps on his barn wall (I only give 50/50 odds on whether we could guess if he’s painted the face after or before he’s fired the shots though 
.)
You measure the center of his bullet holes, calculate how spread out they are, and that will give you a pretty good idea of where the face on the outside of his barn is (possibly even either how big the face is or whether he’s a good shot, but your estimate of one will depend on your estimate of the other.)
Now he fires off a new magazine of bullets. How would we decide if he’s aiming for a new face or at the same old face? I would find the center and spread of his new magazine’s worth of bullet holes, and compare that with the center and spread of his older set of bullet holes.
If I see that the measure of center of our Texan’s new group of bullets is further away than the measure of spread from the center of his old group of bullets, I’m going to be very suspicious that he’s shooting at his new face (or possibly that he’s drunk – we can’t discount that at this point.)
To put that into simple terms – when the mean of your new set is outside the variation of the old set, you should be suspicious of anyone who tells you you’re looking at random variation.
We have a data set of global temperature anomalies running for a bit over a century. We know its center and spread. In 1983, the individual data point for the global temperature anomaly (which itself is an average – but let’s not get all Babushka doll on this ) went over one standard deviation from the 1980 global temperature anomaly. That wasn’t a big deal. Each individual data point in the data set of temperatures centered around 0.1887°C with a standard deviation of 0.1°C has a 16% probability of being 0.2887°C and over, assuming the variation is normally distributed. What happened in 1983 was just random (not really – it was actually a powerful El Niño – but for the purposes of this line of thought it might as well be random) variation, and not even terribly improbable random variation.
In 1988, what happened was different from what happened in 1983. In 1988, the 5 year running mean of global temperature anomalies went outside the standard deviation of global temperature anomalies for the past century by 0.016°C**. Now recall that means are measures of center. Individual years would have a 16% probability of appearing more than one standard deviation above the 1980 value, but the 5 year running mean is different because it is a measure of center. We can’t easily calculate the probability of the 5 year running mean going outside the standard deviation, but we do know that it has to be less than 16%, because more than one year has to be outside the standard deviation.
[sub]It’s also important to recall that when talking about the 5 year running mean, it’s appropriate to re-center your standard deviation on the 5 year running mean of 1980 rather than using the individual data point for 1980. In this case, it doesn’t really matter because the individual data point for 1988 also is greater than one standard deviation above the individual data point for 1980.[/sub]
If we really want a number to attach to that probability (and I like quantifying things,) we could look for the set of five years that produced the 1988 five year running mean. We can also look at the first set of five years that were all at least one standard deviation higher – that occurred from 1997 to 2001. To produce the 1988 five year running mean, three of the five years (1987, 1988, and 1990) were above the 1σ level for the five year running mean. The probability of seeing this in a normally distributed data set centered around 0.1444°C with a standard deviation of 0.1°C* is 0.3%. Not 3%, 0.3%. The probability of seeing the 1997-2001 temperatures in a normally distributed data set centered around 0.1444°C with a standard deviation of 0.1°C*** is 0.01%. My natural response is to say we’re not looking at a data set normally distributed around 0.1444°C – we’re looking at a data set with an upward trend.
[sub]***A horrible Bayesian crime against nature, carefully phrased to avoid the non-independence of the measured data - for which R.A. Fisher would have me taken out and shot at dawn.[/sub]
Think back to our shooting Texan. He’s fired off a sequence of rounds, changed magazines, and fired off a second set. The midpoint of his second set of rounds has a probability of 0.3% - three times in a thousand – of appearing given the midpoint and spread of his first set of rounds. And yet here you are telling me it’s random variation:
No, it’s not random variation. Temperatures are increasing. You may not have noticed that I haven’t yet addressed at all why – the Texan could be changing targets or he could have chugged a six-pack of PBR between magazines, but there’s little point in discussing that unless we at least agree that the GW is happening. Once we agree on GW, only then does it make sense to talk about AGW (which I haven’t talked about yet) or “CAGW.”
Pointless pedantry. If you agree that greenhouse gases increase the planet’s average temperature, you need to curtail emissions at some point.
You can construct hypothetical worlds which alter fundamental characteristics of the planet such that the statement would not be true, but in the real world, where even a skeptic like Richard Lindzen agrees climate sensitivity to CO[sub]2[/sub] doubling is not less than 1°C, the statement is true.
Even in your hypothetical world, you’ve still left a possibility that the statement would be true. For it to be certainly false, you would have to specify that:
- Climate sensitivity to CO[sub]2[/sub] doubling is less than 1°C
- Climate sensitivity to CO[sub]2[/sub] tripling is in a linear relationship or a relationship of decreasing slope to doubling.
- There is insufficient CO[sub]2[/sub] in fossil fuels to raise the atmospheric concentration of CO[sub]2[/sub] above 1000 ppm for long periods of time****.
[sub]****There may be enough CO[sub]2[/sub] to raise atmospheric concentrations to 1400 ppm temporarily in the real world, but it wouldn’t stay that high.)[/sub]
You only specified #1 in your post above. If you’re going to construct a hypothetical in which my statement is false, you could at least make sure it’s false!
Humorous, but completely devoid of a serious point.
If you were to say “President-Elect Obama will be the first black American President.” And my response was “that’s not true in the hypothetical case that both his parents were white.” You would know the mindset in which I am replying.
