Climate Modeling using a Leaky Integrator

[Admiral_Bobski]

I've averaged all the columns, and then computed the difference from the mean over the whole set and applied it to the model with a multiple. Here's the chart.

I am now down to the 5 years following 1945 of 'hand-coded' values to produce this chart.

Bleedin' 'Ell, VP!

So, 5 years to go, eh? I think it's thinking-cap time!

:lol:

CB

[VillagePlank]

Shifting the hand-coded bit forward by two years (1947-1951, now)

[VillagePlank]

I should really show the un'corrected' chart. here it is ...

Still got that dip around 1945 to contend with. Anyone got any ideas?

Also, what does everyone think? Worth pursuing? It seems, to me, to be able to demonstrate quite a few things about climate without recourse to a huge multitude of variables whether man-made or otherwise? Does anyone disagree with that notion?

Edited April 21, 2009 by VillagePlank

[Admiral_Bobski]

Out of interest, VP, when you popped in the ice data did you use the annual average?

Is there any change in the shape of the graph if you use only the summer data for each year (my rationale being that the greatest energy is received in the northern hemisphere during summer)?

You know me, VP - I definitely think this is worth pursuing. I'm still thinking about the period 1945-1950 (or thereabouts) in terms of what could have depressed temperatures (perhaps the late 1940s was just a depressing time).

I'll let you know when (if) I come up with something.

CB

[Devonian]

I should really show the un'corrected' chart. here it is ...

Still got that dip around 1945 to contend with. Anyone got any ideas?

Also, what does everyone think? Worth pursuing? It seems, to me, to be able to demonstrate quite a few things about climate without recourse to a huge multitude of variables whether man-made or otherwise? Does anyone disagree with that notion?

I'd like you to, as you've said, write this up, in accessible form, so that those of us who haven't followed every twist and turn can see what you've done/are doing here.

[Admiral_Bobski]

I'd like you to, as you've said, write this up, in accessible form, so that those of us who haven't followed every twist and turn can see what you've done/are doing here.

It's already written up in accessible form on this thread - clicking the link below should whisk you away to post #101 by VP where he starts from the absolute basics.

http://www.netweather.tv/forum/index.php?s...t&p=1491449

Have a read through and feel free to ask any questions you may have.

CB

Edited April 21, 2009 by Captain_Bobski

[Thundery wintry showers]

I recently posted about the 1945 dip (maybe I posted it in another thread- I can't find it in here!). I have vague recollections of research that showed that there was a discontinuity in the way ships measured temperatures around 1945, related to the end of World War II, which caused a step-change in readings in some regions of the globe, perhaps exaggerating the cooling trend that occurred around then. I can't go into detail as I only vaguely remember it, but I think that will have something to do with it.

[Devonian]

It's already written up in accessible form on this thread - clicking the link below should whisk you away to post #101 by VP where he starts from the absolute basics.

http://www.netweather.tv/forum/index.php?s...t&p=1491449

Have a read through and feel free to ask any questions you may have.

CB

Ok, my question is what does VP think he's found with his LI? It seems to me an explanation for warming that doesn't require an anthropogenic effect? Can it really be that VP has found something all the rest of the worlds scientists have missed - leading me to 'extraordinary claims requiring ...' and all that? Or have I misunderstood the claims - if any?

Edited April 21, 2009 by Devonian

[Chris Knight]

I should really show the un'corrected' chart. here it is ...

Still got that dip around 1945 to contend with. Anyone got any ideas?

Also, what does everyone think? Worth pursuing? It seems, to me, to be able to demonstrate quite a few things about climate without recourse to a huge multitude of variables whether man-made or otherwise? Does anyone disagree with that notion?

If you subtracted the LI output from the temperature curve data, you would know exactly what shape of curve would be required to bring the LI into perfect alignment. Perhaps we could then look at some of the teleconnections. There are certainly indices out there like the winter PCI (The atmospheric circulation index for the Pacific-North American region) and the Aleutian Low Pressure Index that have big inflections around 1945 (and 1976), but I am guessing that you would need something more subtle.

Edited April 21, 2009 by Chris Knight

[Thundery wintry showers]

Edit: re. the 1945 discontinuity, the relevant paper has been posted into the general climate discussion thread- it's peer reviewed and all- here.

[VillagePlank]

Ok, my question is what does VP think he's found with his LI? It seems to me an explanation for warming that doesn't require an anthropogenic effect? Can it really be that VP has found something all the rest of the worlds scientists have missed - leading me to 'extraordinary claims requiring ...' and all that? Or have I misunderstood the claims - if any?

Well, no.

It might be construed that way, but I haven't really drawn that conclusion, although I have asked that question.

One might say that the data has been cherry picked to suit the HadCrut3. That is true - the main premise here is that 'energy' from the sun is somehow stored in the Earth and is released over time providing for some sort of latency effect. We know this does happen, because when the sun goes down the temperature doesn't plummet to absolute zero overnight. Whether it happens on the scale of climate is yet to be determined. So, in my view, it's worth a go to have look and see what happens, and I selected sunspot data as the first attempt.

It's not there yet, either. Plenty of work to do, and plenty of stuff to reason out and rationalise, and, of course, that there is a reasonably high correlation means nothing until the statistical significance work is done, either.

If you subtracted the LI output from the temperature curve data, you would know exactly what shape of curve would be required to bring the LI into perfect alignment.

Here's the error chart ...

Edit: re. the 1945 discontinuity, the relevant paper has been posted into the general climate discussion thread- it's peer reviewed and all- here.

Certainly on today's reading list!

Edited April 22, 2009 by VillagePlank

[VillagePlank]

Here's the error chart ...

... which gives me a development idea.

If we strip back everything we've done and go back to just sunspots we can play with the LI's constants until we minimise the absolute error. Then we can add volcanoes back in, and so forth, and do the same? We can plot the error graph each step of the way to see if it jogs any recollections of climate phenonena too. Any objections?

(The absolute error simply means that if there's a -4 error you count it as 4, and a +4 error is counted as 4, too)

Edited April 22, 2009 by VillagePlank

[VillagePlank]

Back to the beginning, then, with annotations on the graph so all can follow progress ....

Basic idea is that any additions should increase Pearson, and reduce |Error|

[VillagePlank]

I want to go back further; we are currently stuck on starting at 1850 because that's where the Hadley series starts, but we have sunspot data going back another 100 years. Does anyone know of a good peer reviewed global temperature anomaly record that goes back this far? More particularly I'd like to see the LI output follow (if indeed it does) the ups and downs of climate variability

On a quick investigation of volcanic forcing, it can be construed as either a positive or negative forcing depending on the location of the volcano. Simple explanation is ...

"if surface albedo is very high, warming can occur if the eruption causes approximate equality between absorbtion and backscatter, because the radiation that is now being absorbed was previously being lost to space. Over a surface with a lower albedo, however, most of the incident radiation was being absorbed anyway so that the only way to produce a net heating is if the absorbtion is altered to be greater than the backscatter. It is more likely in this case that backscatter from volcanic particles will cause a local cooling"

Contempary Climatology, P Robinson, and A Henderson-Sellers, ISBN 0 582 27631 4

Yikes - could that be the reason why my mistake of integrate volcanic forcing as a positive effect on temps yielded better results than the reciprocal?

Edit: re. the 1945 discontinuity, the relevant paper has been posted into the general climate discussion thread- it's peer reviewed and all- here.

It is interesting that this paper identifies a downward shift of magnitude between 1942-1945 of around 0.3C. I originally thought it to be a problem with the LI output not incoporating enough data but it turns out to be an error in the Hadley construction - simply, some temperatures were taken from engine rooms at sea with the equivalent ambient heat associated with being in an engine room. The paper states that the MetO/Hadley are currently assessing the adjustments required to compensate for this.

What are the chances that the LI would so clearly, and independently, demonstrate that error? (I am referencing this version of the LI, here)

Edited April 22, 2009 by VillagePlank

[Admiral_Bobski]

Hi VP

I've been looking around for temperature data sets that extend back prior to 1850 and so far the only one I can find is the good ol' CET, which goes back as far as 1659:

http://www.newtownweather.co.uk/cetdata/cetdata.html

Obviously this is a local record, and there are issues with its validity when applied to the whole globe. Is there a way that the CET data can be normalised to match up as closely as possible to the HadCRUT data, then the CET data 1659-1850 being spliced onto the HadCRUT data? Would that be any use, or is it a fudge too far?!

CB

[VillagePlank]

Would that be any use, or is it a fudge too far?!

I think that if the CET was a good global analogue, then we'd see wider use of it. It would be an interesting exercise, but not one that I'd want to undertake!

[Admiral_Bobski]

I think that if the CET was a good global analogue, then we'd see wider use of it. It would be an interesting exercise, but not one that I'd want to undertake!

I'll see what else I can find then - from what I've been reading so far, data prior to 1850 is somewhat flakey. The instrumental record wasn't far-flung enough to give a good global overview and proxy data don't mesh very well with the instrumental record at that time.

Perhaps there is some proxy data that extends up to more modern times and joins up with instrumental data more seamlessly...

I'll be back after I've got the kiddies to school

CB

[VillagePlank]

Here's where we are ...

With all parameters, so far ....

Note the LI going off into space for modern times - this is because there is no volcanic forcing past 1994 included.

With the Hadley correction (leak is 3 times bigger for period stated) ...

Any comments or ideas ?

Here's me playing with parameters - to make the curve fit better the historical data ...

Edited April 22, 2009 by VillagePlank

[Admiral_Bobski]

Here's an idea, VP - since there is a known problem with the temperature dataset can we not just keep your manual adjustment in there (so long as we state that it's there and why)?

CB

[VillagePlank]

Here's an idea, VP - since there is a known problem with the temperature dataset can we not just keep your manual adjustment in there (so long as we state that it's there and why)?

I don't see why not. It has to be on the proviso that when the MetO publish either their new dataset that adjusts for it, or their adjustment technique, then we must do the same.

I suppose the next thing to consider is that big bump around 1962 ...

Edited April 22, 2009 by VillagePlank

[VillagePlank]

Here's the best chart I think I can get at this stage. Note the Pearson score.

I've meddled with the way sea-ice is incorporated, and played with the constants, quite considerably and factored out that hump - using the data and it's multipliers - in 1960ish

Anyone got any volcano data post 1940 that follows the Mann et al forcing index?

Edited April 22, 2009 by VillagePlank

[VillagePlank]

Here's the latest 'best'

It's getting very difficult, now, to increase the Pearson score, and reduce the error, so, I think, now is the time to ask all for comments. Please feel free to add, challenge, anything on this thread. EDIT - I may try to evolve this parameter set and see what I get, but that'll take a week or two ...

I have to be honest, I didn't think we'd get as close as we have done. Not quite compelling without a formal statistical analysis, but certainly worthy, I think, of going in the grand mix of things, now; that is we can create a function using only sunspots, enso, sea-ice, and volcanoes, that correlates well with the HadCru3 output.

What do you guys think?

Edited April 22, 2009 by VillagePlank

[Essan]

Beware of one size fits all - it usually fits nothing

As far as I can see this model does not take into account contrails, deforestation, black soot, aerosols, brown clouds, urbanisation etc etc (some of which obviously have both +ve and -ve effects).

Therefore, if it attempts to explain all warming/climate change it must logically be wrong. Just as any model which does not take into account solar activity, volcanism, ENSO etc must be.

If the model produces an approximation of observations then it may of course be more likely to be correct once other factors are included.

Same of course applies to IPCC models.

[Chris Knight]

I don't see why not. It has to be on the proviso that when the MetO publish either their new dataset that adjusts for it, or their adjustment technique, then we must do the same.

I suppose the next thing to consider is that big bump around 1962 ...

Just posted in General - the 1960's hump was also probably due to USN SST irregularities.

[VillagePlank]

As far as I can see this model does not take into account contrails, deforestation, black soot, aerosols, brown clouds, urbanisation etc etc (some of which obviously have both +ve and -ve effects). ... snip ...

Yup - I agree, and that is why the LI model is not a 100% correlation - try to see it has modelling climate direction, and not detail. There are plenty of errors to be seen and had against the two ... here's the chart:

Just posted in General - the 1960's hump was also probably due to USN SST irregularities.

It smoothed out by itself (it's actually still there - but the magnitude is greatly reduced) when I factored down the LI output so it would fit on an anomaly scale similar to the HadCru3

Edited April 22, 2009 by VillagePlank