What's up with the sun?

[marksi]

The sunspot prediction graph has been redrawn by the scientists in some sort of pretence that they have a clue what's happening.

As someone on a geekier forum than this said... "Not sure what the pay packet looks like for the prediction job in NASA but it would seem to be a great job. You can spend all day fishing and relaxing, then once a month move the prediction graph to reflect what was observed."

WAS:

NOW:


This page http://commons.wikimedia.org/wiki/File: ... t_SWPC.gif is a long, slow animation on how substantially the predictions have changed.

[Sput]

Haaaaang on. I see the problem and why it's fine to change the graphs (and at any rate, the amount they change by is nothing in this business: changing by factors of 10 and 100 are where you start to worry).

This is more like a weather forecast than a climate model. With a forecast you take the underlying science and the most recent data you have, run it bunch of times and see what the most frequent outcomes are (and of course as you get closer to the time you're forecasting, the output changes). With a climate model you take historical observations, use them to constrain some parameters that are placed there by doing the maths and run it from an arbitrary starting point to see what happens in the same statistical way. In both, the amount of data you have will affect the prediction, but only in the forecast will the recent data be more important.

[marksi]

I understand that it's like a weather forecast. But a terribly imprecise one.

As time passes I'm less clear on what the probability of this prediction is in comparison to others. The amount of older data and the reliability of it is much lower than other areas owing to the length of a solar cycle, but I think some perspective on probability is required in order to have any idea what importance to place on the prediction.

[Sput]

I imagine that's why there are 2 red lines - they probably mark the edges of the outcomes with probabilities +/- 33.5% from the most likely, which will lie in between them.

And HAH at the met office ever doing anything precise. It's all much dirtier than most people realise. It's the nature of the beast.

[cdd]

Of course the problem with using two lines to mark the boundary is that it only works for positive graphs. Because the two graphs are slightly out of sync, the reading at that point implies that there is no tolerance for error at that reading - when in fact the time component is not certain (as the graph implies).

What a rubbish explanation - it would have been simpler to just say they should have used error bars.

[cdd]

Also I know I don't have any fancy titles or a large paycheck for guessing at the future of that graph, but is it just me who thinks both of those predicted inclines look substantially too steep based on the rate of descent?

[Sput]

On your second post there, cdd: If you look at the prediction curve, it appears that they expect it rises faster than it falls. Be nice if I could see the zoomed out version to see if that's been the case historically.

On your first post: No, error bars would be less clear in this situation. They'd have to be x and y, and there could be a systematic shift to the left when there's a shift in y, for example. It's much safer (and I personally think it's clearer) to show the boundaries of the statistically relevant scenarios.

[cdd]

Well in their prediction, it seems to rise and fall at the same rate. I agree additional historical data would be useful, though - I can't find anything from the same graph, but this graph seems worthwhile:

Graph

EDIT: Re P1; Yep, I see what you mean about a possible shift in X which couldn't be represented with error bars alone.

[Sput]

Well in their prediction, it seems to rise and fall at the same rate.

Oh and I'd also say that the other graph you linked to shows the same thing, it's just hard to make out because it's squished together.

[cdd]

Hmm. Nicely illustrated.

Well I'm dyslexic. (Best I can do!)

[Stuart*]

At least Sput's awful scrawl gave me something to chuckle at in this thread, at last!

I find the subject very interesting, but my knowledge is unfortunately restricted to what little I can glean from the various Discovery/Eden/NatGeo Channel documentaries on the topic.

I think I've grasped the somewhat haphazard nature of sunspot cycles, and even the changes in the Earth's magnetic field. However, I did see a documentary a couple of weeks ago which discussed the possible gravitational impact on the Solar System of crossing the galactic equator (due to begin around 2012) which only happens every 26,000 years. The inference was that this could either initiate or increase the speed of the magnetic polar change, together with the associated effects on our protective magnetic 'shield'.

I'm hoping that 'Professor' Sput, or some other knowlegable person, may be able to explain whether this could have any element of truth.