So this has already been covered by Skeptical Science, so I’ll only make a brief comment. Owen Patterson – the UK Secretary for the Environment – on Question Time last night, apparently said
the climate’s been going up and down – but the real question which I think everyone’s trying to address is – is this influenced by manmade activity in recent years and James is actually correct – the climate has not changed – the temperature has not changed in the last seventeen years.
So, the James mentioned above is one James Delingpole and the less said about him the better.
So, is Owen Patterson correct? Has the temperature not changed in the last 17 years? The answer is No. This is so incredibly frustrating as this mistaken view should have been put to bed ages ago. Why is it wrong? Well, what they are referring to is the fact that if one uses linear regression to analyse temperature anomaly data one discovers that the 2σ error in the best-fit trend is large enough that one cannot rule out that temperatures haven’t dropped – or been flat – in the last 17/18/19 years (depending on which dataset you choose to use).
However, as I pointed out in yesterday’s post, even though it is possible that temperatures have not changed for the last 17/18/19 years, it isn’t probably. For the GISSTEMP data (from NASA) there is only a 2.9% chance that the trend since 1995 is negative – hence a 97.1% chance that it is positive (and I have not engineered these numbers to be the same as for the consensus on AGW in the literature). You can go on to show that there is an 85% chance that the trend exceeds 0.05oC per decade and a 56% chance that it exceeds 0.1oC per decade.
So, Owen Patterson’s statement is simply wrong. The data does not indicate that temperatures have been flat for 17 years. They indicate that temperatures could have been flat but that this is highly unlikely and it is much more likely that temperatures have been rising and rising at a rate close to 0.1oC per decade. If you want more details you can read another post of mine that addresses the claim that we’re in a pause or a decline – we’re not in case you were wondering.
Wotts Up With That Blog 
There’s that 97% again. I can’t help feeling that you’re pushing the 97% theme here.
As for the probability of warming or no warming, or it’s worse than we thought, it’s actually cooling… the fact remains that there has not been any statistically significant warming for quite a while now. That’s why scientists are coming up with novel
excusesreasons for the lack of warming.Using a cloak of probability to hide from reality isn’t going to change that reality. This isn’t a fringe group talking about it… it’s mainstream. The lack of warming is real and evident in every temperature dataset. This idea of basically saying that there may be some warming hidden in the noise just makes you look desperate. Yes, global warming doesn’t work well without the warming so I can understand your frustration at the atmosphere’s reluctance to continue playing the global warming game, but this obvious clutching at straws really isn’t a good look.
I assume you’re joking about the 97.1% – although I can’t really tell.
Basically, your interpretation of the temperature anomaly data is simply wrong. You cannot use it to claim there has been no warming since the mid-1990s. I don’t really know what else to say. The numbers I include in the post above are a quite detailed analysis of the likelihood of particular trend ranges. If you disagree, you can do the same yourself and check. If you can convince me they’re wrong, go ahead. But to claim that there has been no warming for 17 years when there is only a 2.9% chance that the trend has been zero or less is, quite simply, the wrong interpretation of the temperature anomaly data.
Yes, I am joking about the 97%… someone else pointed that out in your previous post so I thought I’d poke a bit of fun at it appearing again in this post.
The metric used to determine warming is ‘statistical signifance’ in the data, not the probability of warming hidden in the mist. You can continue to point to what you see as a high probability that there is a tiny bit of warming hiding in the noise, but it really isn’t going to change the fact that the warming has, more or less, peaked out.
If my interpretation of the data is wrong, then I’m not alone.
Yes, but the statistical significance is based on the magnitude of the 2σ errors. These 2σ errors determine the range of possible trends, with 95% confidence (i.e., there is a 95% chance that the actual trend is within 2σ of the best-fit trend). It’s 95% because the likely trends are normally distributed (one can show this by integrating a normal distribution from -2σ to 2σ – essentially as I did in the last post). You can therefore determine the likelihood of the trend lying in any particular range. For the GISSTEMP data, therefore, there is only a 2.9% chance that the actual trend, from 1995 to 2013, is negative. That’s what the error analysis is telling you. So I’m more than happy to accept that there is a possibility that there has been no warming. The data, indeed suggests that. However, the likelihood of this being the case is very small.
Part of the reason for writing this blog is because, in my opinion, many people’s interpretation of this data – and of other data – is indeed wrong. When I say opinion, I mean scientific, not I’ve just decided this because I want it to be true.
You just said it yourself…
A ‘possible’ trend based on probability means nothing. An event can have a 99.99999% probability of occuring, and yet may not occur… this is where your interpretation falls over.
The warming has to be of statistical significance to be regarded as warming. You saying that there is a big probability that it’s still warming has no more value than me saying that there’s a small probability that it’s cooling… either could be true.
Sorry, you’re just simply wrong. You can’t use these errors to claim there is no warming and then ignore the more detailed error analysis that tells you the likelihood of particular trend ranges.
I agree that something can have 99.9999% chance of happening and still not happen. That is essentially the point I’m making. It is indeed possible that there has been no warming. It is, however, much more likely – according to the data – that there has been warming. I am not claiming that there definitely has. I’m simply stating the likelihood based on the analysis of the errors. You, and others, seem happy to state that there has been no warming simply because the uncertainties mean that it can’t be ruled out. That is an incorrect interpretation of the data.
No, it’s not a case of claiming no warming because it can’t be ruled out… it’s a case of there not being enough warming (if any at all) to be statistically significant. You can’t find hidden warming using probability… all you can find is probability. You claim that using probability provides a ‘more detailed error analysis’ but the reality is that probability is just a game of chance.
The whole point here is that you are saying others are wrong for saying that there’s no warming when you can’t prove that there is any. Others are saying, rightly, that there is no statistical significant warming. Saying that there is a possible/probably/maybe a tiny bit of warming hidden in the noise doesn’t make them wrong…. if the warming is not statistically significant, then it’s not warming.
Don’t feed the trolls.
Thanks Victor, but I’m going to give it one more try.
Skeptical, I’m sure we’re not going to agree but I’m going to give it one one more try. Here’s where I think there is some confusion. Let’s imagine we have scientific evidence that suggests that a particular measurement/experiment should return a particular value. We then carry out an experiment/measurement and discover that our result differs from what was expected. However, we then consider the errors and discover that the expected result is statistically consistent with the measured result. Then one might say the difference is statistically insignificant.
How does this relate to temperature anomaly data. Let’s imagine that we have some scientific evidence to suggest that we would expect the trend from 1995 to 2013 to be flat (i.e., no warming). We consider the temperature anomaly data and discover that a flat trend is statistically consistent with the data (at the 2σ level). One might then conclude that there has been no warming. The problem is that there is no scientific evidence that suggests that the trend should have been flat since the 1995. There is nothing special about a flat trend. Just because it is zero, doesn’t make it more likely than any other value. That a flat trend is possible is therefore not especially relevant. For the GISSTEMP data, a trend of about 0.2oC per decade is as likely as no trend at all.
We could actually push this further. There is scientific evidence (climate models – which you may dislike, but they exist nonetheless) that suggest that the trend would be expected to be between 0.1 – 0.2 oC per decade. Given that this is statistically consistent with the data, maybe we should really be claiming that these are the correct values. Now, I’m not suggesting we do so. I’m just trying to point out that without some additional scientific evidence that a flat trend is expected, that it is statistically consistent with the data has no particular significance because I can use the same argument to suggest that the actual trend is 0.2oC per decade (which I’m not doing mind you). For the GISSTEMP data, any value between -0.006oC per decade and 0.222oC per decade is statistically consistent (at the 2σ level) with the data. You need to explain why 0 is somehow more consistent than any other value before your interpretation really has any merit.
Skeptikal, can you show that there a statistically significant temperature standstill/cooling?
Frustrating! I think you’ve explained this really well. Statistics is not my forte but even I’m beginning to get it. People like Owen Patterson are trying to create doubt in order to delay action on climate change. It is unfortunate that they are to an extent, successful. I am optimistic though that this is a short-term barrier. Bill McKibben said that “mother nature” is a far more effective educator and I think this is beginning to sway some people already. It will also become harder and harder to argue the world is cooling when all the Arctic ice is gone and when the oceans continue to heat, the permafrost melt and the sea level rise.
Thanks. It seems quite likely that it will soon be obvious that we need to do something about global warming/climate change. The hope, of course, is that when it becomes obvious it doesn’t also become too late – or at least not too late to not suffer in some significant way.
Yes, true. There is positive news out of China though – http://www.nature.com/news/china-gets-tough-on-carbon-1.13175?WT.ec_id=NATURE-20130613
They are about to launch an emissions-trading scheme and plan to cut emissions by 40-45% by 2020.
We will have to disagree on this one. Your whole argument has been that there has probably been warming but you can’t put a figure on the rate of warming since it lies somewhere inside the 2σ errors… BUT, for someone else to say that there is no statistically significant warming is wrong because you have probability on your side. No one else (as far as I know) is claiming warming based on probability, but if it makes you feel better, then go ahead and keep claiming that.
No, I didn’t say that for “someone else to say that there is no statistically significant warming is wrong”. I’m simply saying that extending that statement to “there has been no warming” is wrong.
The theory of global warming is not based on statistics, but on our physical understanding of radiative transfer though the atmosphere and the feedbacks in the climate system.
You would like to test theories on empirical data, which are always uncertain. Thus in principle, you always need statistics to compare a theory to data. For some theories we are fortunate enough that we can do (laboratory) experiments and make the uncertainties so low that it is intuitively clear that the result is statistically significant without an explicit computation.
One should also consider the length of the period for which the trend is computed. Seventeen years is extremely short for a climatological study. If you select the period short enough, as climate ostriches tend to do, no interesting, informative statement will be statistically significant.
Pingback: Another Week of Global Warming News, June 16, 2013 – A Few Things Ill Considered
Reblogged this on acckkii.