21 replies to this topic

[Android]

I know curve fitting is a bad thing, but I wanted to do it anyway.

In red is hadcrut3:


In blue added Multivariate ENSO Index (MEI) (http://www.cdc.noaa....aus.wolter/MEI/)


Scaled the MEI down:


Add a 0.15C/decade trend:


This is the real fudge one, to get a better match I made the trend lightly exponential and also shifted the MEI plot forward by 110 days:


In my opinion this suggests that the flat temperatures since 2002 are actually consistent with an accelerating global temperature trend in recent years that has been somewhat masked by the recent ENSO decline (similar to how pinatubo masked an otherwise warming trend). There doesn't seem to be any divergance between MEI and global temperature since 2002 like there was during the Pinatubo eruption period. There doesn't seem to be any obvious impact from solar minimum or from the negative PDO although of course either of those two could have caused the decline in MEI.

In any case if the relationship over previous decades is anything to go by it would take progressively stronger La Ninas at this point to stall temperature any futher, which seems unfeasible. So when we are already low and there's not much further to go, my guess is that it's more likely than not that we will shortly see a sudden and large step jump like we did from 2000 to 2002.

Edited by Android, 01 May 2009 - 00:59 .

[Boar Wrinklestorm]

If you look around in Excel you should able to derive a 'Pearson' score which is an index which tells you how correlated to the two lines are. Very interesting about the 110 day shift, though. Have you nosed in the LI thread, where we are investigating the possibility of sunspot energy 'lagging' ?

Off the top of your head - what are the variables that are contained in the multivariate enso index?

Great work, btw!

EDIT: the variables are: sea-level pressure (P), zonal (U) and meridional (V) components of the surface wind, sea surface temperature (S), surface air temperature (A), and total cloudiness fraction of the sky ©.

Edited by VillagePlank, 01 May 2009 - 07:37 .

[Captain_Bobski]

This follows on from something that Iceberg posted a while back, but if memory serves Iceberg didn't put in the underlying 0.15C per decade warming trend (apologies to Iceberg if you did!).

So it rather begs the question "what's causing the 0.15C per decade warming trend?"

The ENSO pattern seems to tweak a straightforward 0.15C per decade trend curve into a nice jagged curve that pulls it in line with temperature fluctuations. Which gives me an idea...!

I don't want to hijack this thread and turn it into Leaky Integrator 2: The Sequel, so I shall continue over on the LI thread.



CB

EDIT - I also don't want to ignore this thread, since it could prove extremely interesting...

Edited by Captain_Bobski, 01 May 2009 - 08:23 .

[Iceberg]

Your right CB, all I did was show that a 0.15C per a decade trend was shown but not apply it to the data per se.
I also didn't use that level of ENSO index more a straight forward plot of ENSO expressed in SST variation.

The 110 days sounds about right, a 3 month lag is most notable from ENSO wrt to global temps, which is why the decline in LA Nina in March/April will likely lead to a increase in the global temp anomaly from June/July onwards increasing as we go through the second half of the year.

Completely agree Android it looks like Spring 2007 was the low point re temps and that temps will steadily increase now. Bar a very strong -2C La Nina in the next 10 years or a massive fall in Solar input or a strong volcanic input.

I would put 2007 as the lowest year temperature globally for the next 100 years at a chance of maybe 20%.(doesn't sound high but believe me it is)

[Thundery wintry showers]

Do you mean 2008 rather than 2007? 2008 was, globally, the cooler of the two years.

[Iceberg]

Yep Sorry TWS 2008, Maybe 2007 came into my head I was thinking about Arctic ice at the same time.

[johnholmes]

what a fascinating thread-I will drop in here fairly often to read the theories as its above my little head-thanks for starting it Android.

[Android]

I was going to adjust for PDO next to see how that affected things. There is suggestion that positive PDO over the past 30 years caused a significant part of global warming, and also that it's played a large role in recent cooling.

Here's a graph of ENSO MEI index vs PDO index since 1950.



The PDO record is so similar to the ENSO record that I seem to have already adjusted for it when I adjusted for MEI.

I got the PDO numbers from http://jisao.washing.../pdo/PDO.latest
That page says

Quote

Updated standardized values for the PDO index, derived as the leading PC of monthly SST anomalies in the North Pacific Ocean, poleward of 20N. The monthly mean global average SST anomalies are removed to separate this pattern of variability from any "global warming" signal that may be present in the data

If I understand that correctly, PDO can be described as:
Positive PDO means North pacific sea surface is warmer than global sea surface on average.
Negative PDO means North pacific sea surface is cooler than global sea surface on average.

Seems more simple than MEI which contains more factors than just sea surface temperature as VillagePlank mentioned.

Given the proximity of North Pacific Ocean poleward of 20N to ENSO regions wouldn't it be the case that a strong El Nino would result in a warmer North Pacific Ocean sea surface temperatures? So could it be that PDO tends to follow ENSO because the North Pacific sea surface temperatures are influenced by ENSO events? The 1998 El Nino for example also saw PDO go strongly positive. PDO also went strongly negative during the 1999-2000 La Nina and during the recent La Nina.

That seems too simple to me, but in any case if PDO and ENSO track so closely then there doesn't seem to be any reason to invoke PDO as a cause of climate change when this is little different from changes in El Nino/La Nina ratios and this is already provide by MEI.

The question of PDO contribution to global temperature could therefore be phrased as "In the last 30 years there have been more El Ninos than La Ninas compared to the previous 30 years, what effect has that had on temperature?"

Which I suppose can be done by removing the effects of ENSO from hadcrut temperarture record. Here is Hadcrut3 minus MEI (ie Hadcrut3 with ENSO adjusted out...somewhat..noway near perfect, it's a month for month subtraction, but it cannot be wildly wrong and the 1998 el nino is removed almost entirely)


If this is anything to go on (perhaps not - it assumes PDO is driven by ENSO among other assumptions), it seems PDO has not contributed in any significant way to the recent warming trend. ENSO has only contributed to variation in global temperature and also not in any significant way to the recent warming trend.

Added: I just remembered a better attempt at removing ENSO from hadcrut:
http://www.realclima...rends-and-enso/

Edited by Android, 02 May 2009 - 23:03 .

[Android]

Good point I will look at AMO.

Another thing I just did was to remove the MEI and the long term trend (using the out-of-thin-air 'exponential trend' used earlier of 0.0035*X^1.4) from hadcrut to see what is left to describe. The black line is the remaining temperature variation once ENSO (+PDO?) and this long term trend are removed. The pinatubo cooling is more obvious now (I assume that's what it is).


One thing I did was plot sunspot number (scaled down a lot) from http://solarscience....ch/spot_num.txt against this, which seems to line up fairly well.



If this is anything to go on (and there are far too many assumptions to get to this stage) it suggests the solar cycle causes about a 0.2C change in global temperature from minimum to maximum.

Edited by Android, 03 May 2009 - 13:19 .

[Arnost]

I completely agree with John above: this is a fascinating topic and so I can't resist and have to drop in…



Android above identified the apparent correlation between the ENSO and temperatures. I also know that curve fitting is a bad thing – but it's so tempting…Here's something to stimulate the debate:



If one extends the MEI to its starting point in Jan 1950 and maps that against HadCRU temps (or the temp index of choice) it becomes apparent that the (visual) correlation really breaks down at the time of El Chichon – a time where there was that "other super-El Nino" – see below circled.







Obviously, volcanoes make an impact, so when the impact of associated stratospheric (volcanic) aerosols is corrected for – the correlation is indeed very good. But with one exception – at the beginning the correlation is very high, but it breaks down toward the end and the temperatures climb above that of the corrected MEI – see above circled.



It must be noted that stratospheric aerosols are not the only thing that effect the temperatures. For example there are also tropospheric aerosols (and obviously greenhouse gasses).



Graphing from 1970 is a bit of a short time-frame. Even from 1950 is a bit so-so when we can do it from 1880!. Instead of the MEI, we can use the SOI. The below is the SOI (green line) against HadCRU from July 1880 (blue line) to current:







I have (in this case) used Kevin Trenberth's Standardised SOI. The dotted line is the Greenhouse Gas and the red line is the combined (Stratospheric, Tropospheric and Indirect) aerosol forcings used in GISS Model E (which I used to adjust the MEI above…).



The SOI appears not to be a factor in temperature trends… But is this really so? If we adjust the SOI by the Aerosol forcings then we get:







Not bad – but there are a few times where the relationship breaks down a bit (circled). As pointed out above, the AMO is a player. When we adjust for that we get:







I think that's pretty good…! Now, I have not used or adjusted for solar forcing (as Android showed above), there may well be a relationship between solar and ENSO (and the AMO for that matter).


Food for thought and discussion!


Sources:

Model E forcings from: http://data.giss.nas...#climsim_table1

MEI from: http://www.cdc.noaa..../MEI/table.html

Standard SOI from: http://www.cgd.ucar..../soiAnnual.html

(Australian BoM SOI available here: ftp://ftp.bom.gov.au/anon/home/ncc/www/sc...iplaintext.html )

AMO from: http://www.cdc.noaa....on.us.long.data

HadCRU temperature: http://hadobs.metoff...l/nh sh/monthly



Methodology: All the data from above have been used as downloaded with the exception of the Std SOI index, which was inverted and scaled down by a factor of 50.



Oh yes… If I adjust the above model by the GHGs using the same methodology, this is what I get: http://i43.tinypic.com/1766on.jpg



Cheers!

Edited by Arnost, 03 May 2009 - 15:28 .

[Android]

That is interesting, more detailed than my attempt. I avoided putting in greenhouse gases because I knew that would requrie aerosol forcing too, which I couldn't find data for. Knowing the sum of both greenhouse gas and aerosol forcing is a wiggly line over time meant I just stuck to starting in 1975 where my assumption was that the forcing of both would be fairly much consistant. I could add both in now using the ModelE aerosol forcing, although you have essentially already done that. I'll take a look at SOI too when I also take a look at AMO. Something bothers me about using these ocean oscillations and deducting them from global temperature, given that they are part of global temperature in the first place.

Yes you might find a solar cycle in there when you deduct everything else from the temperature record, but I imagine the uncertainty might hide it. It's small and hard to see but it seemeed to make everything line up better, at least in my attempt since 1975. It would be interesting to extend the graph to make a prediction for the next two decades. I couldn't do that because my underlying trend was assumed. But assuming aerosl forcing, greenhouse and solar cycle can just be extrapolated it could be possible to make a prediction.

Edited by Android, 04 May 2009 - 14:14 .

[Arnost]

Android, on 5 May 2009, 12:11 AM, said:

That is interesting, more detailed than my attempt.

Thank's Android - I purposefully linked to the datasets so that anyone can replicate, and improve on this...

(PS are you also posting this elsewhere?)

[Android]

yes, on solarcycle24.

Since the last post I have added in the models forcings you used for greenhouse gases, volcanic forcing, aerosols as well as extending it over the entire 20th century.

I had to drop use of MEI as the ENSO index because it only goes back to 1950. So I am now using the global-sst enso index (http://jisao.washing...#digital_values). It's close to MEI anyway.

The scales I have chosen are all arbitary and were chosen soley in attempt to get the best alignment with hadcrut. So this only suggests these scales are consistent with the temperature record, not that they are the actual case. This goes especially for the aerosol and greenhouse gas forcing scales. Because the aerosol forcing record and greenhouse gas forcing record oppose each other, there are probably a variety of different scales that can be used to get a match. The fit is fairly good:


Note that the axis of the two graphs unfortunatly don't line up due to a pasting error.

There is a small mismatch in the 40s. Another possible flaw is that unlike the leaky integrator on the other thread, there are no delays in response to forcing in this method (apart from the response to ENSO)


I also performed a similar exercise to remove the volcanic, ENSO and solar cycle from the UAH record. It's a very rough "correction". But it does show that the recent flattening of temperature (since 2002) is consistant with the declining ENSO and solar cycle since then:

Edited by Android, 09 May 2009 - 20:19 .

[Chris Knight]

Android, on 9 May 2009, 09:18 PM, said:

There is a small mismatch in the 40s. Another possible flaw is that unlike the leaky integrator on the other thread, there are no delays in response to forcing in this method (apart from the response to ENSO)

Wartime bias - Also in the early 1960s (Cold War) - this was mentioned in VP's LI thread, due to an uncorrected error in the HadCRUT3 data series clearly caused by SST measurement bias because the Sea Surface temperature measurements were taken by engine inlet temperatures, rather than bucket measurements. Post WWI also showed a similar blip, supposedly corrected for. Thompson et al shows the bias clearly:

[Android]

interesting, plus the stuff on COWL in that paper.

Edited by Android, 10 May 2009 - 22:21 .

[Captain_Bobski]

It's been a bit quiet on here of late, so I thought I'd take the opportunity to say that this is really good stuff you guys have been doing. Kudos to you, Android and Arnost!

To quote the bomb from Dark Star, "I must think on this further..."

<retracts into bomb bay>



CB

Edited by Captain_Bobski, 11 May 2009 - 17:55 .

[Android]

Update. It's now about a year later. The main difference is that the recent El Nino is included. The solar minimum is still around! but the next cycle is clearly now ramping up. The parameters used for the graph below aren't exactly identical to the ones used above. It's still curve fitting so take it with a grain of salt anyway.






Red: HadCRUT3

Orange: Sunspot count x0.0015

Maroon: MEI (Enso Index) x 0.12 and shifted forward by 110 days (this is arbitrary)

Green: exponential background warming trend - cause unspecified - 0.0035*X^1.4

Light Grey: Volcanic forcing x 0.05 (this is merely to reduce the very obvious discrepancy around the time of the Pinatubo eruption)

Blue line: Orange + Maroon + Green + Gray

Everything is done by eyeballing a fit. No statistical procedure to determine scalings and offsets was used. I know the image quality, captioning, etc isn't that great either. But the analysis doesn't really deserve more than this. It's just rough curve fitting by eye, somewhat interesting but probably wrong in a number of ways. It's not even clear what all the assumptions are.

Below is the same graph except the blue line does not include the sunspot cycle. This is to emphasize how factoring in the solar cycle seems to yeild a better match. I posted something similar on the thread a year ago showing that when I remove ENSO and the background warming trend there does appear to be a significant 11 year cycle in the data. Again eyeballing the data, so take it with a pinch of salt.



Back to the first graph:



If this is anything to go on (and it likely is not so this is just speculation), it is the recent drop in the solar cycle and the long sustained solar minimum that is responsible for the global temperature plateau/fall in the past ~7 years. Previous descending tail ends of solar cycles also appear to result in such falls in global temperature. The recent plateau appears more obvious because it is not as heavily distorted by volcanic and ENSO variability, and the solar minimum has been drawn out for longer.

If the next solar cycle is similar to the last 3, the ramp up into solar cycle 24 (you can just see this starting at the end of the orange SSN plot) should yeild about 0.15C warming, which coupled with the ever rising green background trendline, will effectively bring global temperature to a new plateau. Ie global temperature will experience an upward step change some time within the next 3 years. I emphasize this is what will happen in to the blue plot, not necessarily to the red (hadcrut) one. It's possible the whole basis of this eyeballing is wrong and therefore that the blue and red plots will  diverge in coming years.

Edited by Android, 11 April 2010 - 15:12 .

[Gray-Wolf]

I'm sorry Android, I'm being a bit thick here, but is this looking like solar has a small input that we can see over the 11yr cycle and ENSO the same within it's cycling but that the underlying upward temp trend trumps all ?

EDIT: What could possibly be responsible for capturing the high temps and carrying them thruogh the cold ones to then have more 'warm' added on top of them again?????

It's almost like the atmosphere's got a blanket around it isn't it?

Edited by Gray-Wolf, 11 April 2010 - 16:08 .

[Android]

Yes you are correct