Thursday, February 13, 2014

Turned out Wet (and still raining)

My last blog-piece sparked off a twitter debate between Zoe Williams of the Guardian and one of the brigade of chicken headed deniers (See recent speech by Prince Charles). Richard Snape kindly recorded the conversation on this storify, and then went on to write a return blog-piece of his own. Thank you Richard, it’s very helpful and perhaps treats my piece with more gravitas than it deserved. The following is really a response so you’ll need to read it to make much sense of this post.
Importantly, mine is not a statistical analysis. The trouble is there aren't really any data to analyse. Rainfall records only go back ~250 years and I just made the wild assertion that this is the wettest such period since the Neolithic. That may or may not be true and probably we’ll never know.
It’s good not to cherry pick the data and I've not defined what period or events I'm talking about. Southern England December 2013, January 2014, into February 2014 and still counting, will do for time and space, rainfall, depression intensity and frequency, wind-speed, temperature, the whole gamut of weather will do for the 'event'. Sadly, we don't have the data set for all that lot going back to the Neolithic!
So is it 5-sigma? Well, frankly, I can't be certain; depending on where one draws the boundaries, it might be. Or not! You're quite right, Richard, about 'generic'. I'm wielding a broad brush.  So never mind the data, your three points of 'implied argument' are spot on:
1. If the observation is very unlikely, then the distribution must have changed.
2. The distribution changing implies climate change (and often the anthropogenic forcing element thereof).
3. If the climate is changing - that cannot be coped with using Business-as-usual methods.
I love Nick Taleb; if you haven't yet, read Black Swan.  Expect the unexpected.
But Richard, you say "climate science appears to indicate that warming would increase the amount of water that can be carried in the atmosphere". I’d put it more strongly. We know that global warming leads to more rainfall, (in as much as we know that apples fall downwards). And we’re not just dealing with Somerset. It’s reached Berkshire so must be important now. The bigger the unusual event is the more unusual it is, but of course there I go, cherry-picking boundaries.
“I think that Climate Change probably is happening." Good. But the IPCC put that more strongly. They use the word 'unequivocal' with reference not just to climate change but to the assertion that climate change is largely being caused by man-made greenhouse gas emissions. I agree with the IPCC position in this regard.
"Fundamentally, the rarity or otherwise of individual weather observations cannot, in my opinion, provide conclusive evidence for or against climate change." Very true, but the thing is, we've done and dusted that debate. We know, like we know that apples fall downwards, that our greenhouse gas emissions are causing global warming, which is, in turn, causing climate change (see my earlier piece on these phrases). There are plenty of details to research and argue about, which is what climate scientists do, (we're not quite sure which daisy the apple will roll onto) but all the scientists working in the field accept the basics. There is consensus. My political purpose is to get folks to realise this.
It's really very difficult for climate scientists to communicate what they really believe. Everyone, at least in public, has to be very careful, trying to please their employers and not risk their careers. (Be particularly wary of what some meteorologists say. Weathermen are not always climate scientists, failing to see the wood for the trees.)
One of the trickier areas is with the global climate computer models. The numerical models are largely built on the underlying assumption that the climate is stable and they do not sufficiently account for thresholds and tipping points. But don't be tempted to think that therefore the models are rubbish. That would be ignoring the vital rider that the probability space of error is all on the bad side. The models underestimate the likelihood of catastrophic change. If you are not frightened of serious mathematics you might like to dig into the work of Lenny Smith at the Grantham Research Institute on Climate Change and the Environment at the London School of Economics.

This diagram is for the comment below:


Anonymous Anonymous said...

You are quite right to stress that the observation of a highly improbable event generally means that it is more likely that the model (stationary climate) is wrong.

Goldman Sachs' famous lament that "we were seeing 25 standard deviation moves on several successive days" is a nice example Obviously their models were wrong too.

The question in a bayesian sense is how much confidence do you have in the model? i.e, how "unlikely" an event would you have to observe, to conclude that the model is wrong? In this case even most of the skeptics tend to admit that the climate is significantly non-stationary (with or without influence of human GHG emissions).

Also I wonder if there is a useful comparison to be made with the UNFCCC probabilities, where we seem to have converged on an "acceptable risk" of 50% (!) of exceeding 2 degrees versus avoiding "dangerous climate change". Would those arguing that we need to be protected from a 0.5% chance of extreme flood apply a similar risk-aversion attitude to temperatures (which are of course an important factor in local rainfall events)? In which case extreme mitigation is implied.

Here is a nice graph showing that the summer of 2003 was a 5.4-sigma event for European summer temperature (See graph above). If you use a slightly different domain to do the calculations over, you get a similar sort of graph, eg here:

But the exact region/timescale/etc you choose will give slightly different results. It's important not to cherrypick the most extreme looking period, because although you may get a more extreme looking result, the statistical "significance" will be vastly diminished by having done so. (If you look at 100 sets of randomly distributed data, it's quite likely one of them will be a one in 100 event...).

10:36 am  
Anonymous one of the brigade of chicken headed deniers said...

"Importantly, mine is not a statistical analysis."

And yet you claimed there had been a 5-sigma event. And it caught the eye of journalist, who took it at face value.

Imagine that one of us headless chickens made the same order of claim, of statistical significance, to make a point about attribution in the UK media. Imagine that it later turned out to be groundless.

Headless chicken or not, this bird-brain had sufficient faculties to pick apart the 5-sigma claim, though the point was resisted by the journalist in question -- largely down to her lack of familiarity with the terms, but also due to her presupposition that the decapitated fowl must be wrong.

It should worry you that the headless chickens are able to debunk your claims. But what should worry you more is that you appear to be on the same side of 'science' as a man who is a powerful and vocal advocate of alternative medicine and homoeopathy.

1:01 pm  
Blogger biffvernon said...

If the headless chickens are able to debunk my claims, they haven't done so yet. As I pointed out, we cannot do a statistical analysis of a long time series of weather data to establish the variance of climate, since the Neolithic folk never got round to inventing the rain gauge. We can never know if this winter has been a 5-sigma event. But it might be.

8:30 am  
Blogger biffvernon said...

Dr Peter Stott is head of the Climate Monitoring and Attribution team at the UK Met Office. He has written a piece about connecting extreme weather events to global warming here:

11:43 am  
Blogger produkey said...

I like your topic of selection for blog but want to suggest you for sharing some more information regarding your subject so that we can understand your thought more clearly
windows 8.1 product key

8:42 am  
Blogger produkey said...

This comment has been removed by the author.

8:43 am  

Post a Comment

Subscribe to Post Comments [Atom]

<< Home