Uncertain Uncertainty and Variable Variability

Graph prepared by John Kennedy illustrating the effect of some of the uncertainties. Any one of the blue of the blue lines - or an un-drawn similar line - could be what actually happened. We don't know - but all of them show significant warming.

Graph prepared by John Kennedy illustrating the effect of some (but not all) of the uncertainties in the data. Any one of the blue lines – or an un-drawn similar line – could be what actually happened. We don’t know – that’s the nature of uncertainty. The significant thing is that even considering the confounding factors, all of the estimates show significant warming.

Variable Variability

One of the real pleasures of attending WMO CIMO TECO last week was the chance to meet some of my heroes. And among them I finally met Victor Venema.

Victor is climate scientist whose primary interest is in identifying and removing biases from the instrumental temperature record. He is – in the very best sense of the word – a sceptic.

His blogVariable Variability – is one of my few ‘must reads’.

Uncertain Uncertainty

Victor’s last article drew together many representations of our instrumental temperature record to ask the question: what makes people pay attention to the fact that OUR PLANET’S SURFACE IS WARMING UP!

This shocking fact has gone from being widely denied or ignored to being widely accepted and ignored.

The aim of all the presentations Victor draws together is to fairly communicate the reality of the uncertainty of the conclusions drawn – but also that the warming trend is strong when compared with these uncertainties.

Alternative Reality

But Victor’s page does not (yet!*) contain the beautiful representation at the head of this page.

The animated graph was devised by John Kennedy from the UK’s Met Office and illustrates many of the possible curves – alternate realities – that are consistent with the data.

There are more curves that would be consistent with the data but John wasn’t quite sure how to represent them.

One of the most important ‘uncertain uncertainties’ that John didn’t include is called ‘coverage uncertainty’. It arises from the fact that the instrumental record derives from thermometers that are not optimally positioned around the globe.

When I wrote to him to ask permission to use the graph he said:

The coverage uncertainty has, I suspect, an important low-frequency component. We know HadCRUT4 has a tendency to slightly under-represent Arctic areas, which have been relatively warm these past 10 years. Over time, the balance of land and ocean changes too and we know these warm at different rates. The coverage uncertainty also has a high-frequency component too.

I will get round to writing a blog post about the wiggles at some point, but in the meantime I’m interested in what people think about it. Lots of the animated presentations that I see don’t obviously add anything beyond what the standard static time series graph would show, so one concern I have is, does it add anything to that? Is there any way we can improve the representation of uncertainty in our graphs and other visual aids, particularly where there are more complex error structures that can affect the interpretation?

I love John’s attitude. Critical of his own work and looking for feedback to improve it.


Certain Certainties

I think the animation does add something. Each line represents a possible ‘reality’ that is consistent with the data we have.

The animation shows which features persist from one ‘possible reality’ to another.

In general, a year which is hotter than it’s predecessor, stays hotter in all ‘realities’.

Critically, none of the realities consistent with the data reverse or cancel the overall warming trend.

And that makes it essentially certain that the warming trend is real. And that the world really is hotter than it has been for a long, long time.


*The image is there now!


2 Responses to “Uncertain Uncertainty and Variable Variability”

  1. Victor Venema (@VariabilityBlog) Says:

    Fortunately, in cyberspace no one can see you blush. Did you get inspired by today’s #YouAreAmazing hashtag?

    The pleasure was all mine. I love the work you do. Maybe I should have become metrologist or statistician. Methodological questions are fascinating and a large part of what brings science forward. Unfortunately, as an atmospheric scientist you often have to act as if you are studying the atmosphere, as if the methodological advances are just a small side study.

    I am sorry, but I included the animation of John Kennedy this afternoon. The main additional information over normal error bars is that you see how strong the uncertainty is correlated in time. That helps you not to overinterpret too small wiggles as being real. Such as that wiggle some people call “hiatus”.

  2. Wiggles | Diagram Monkey Says:

    […] are already some, rather more coherent thoughts on this plot by Michael de Podesta and Victor […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: