While there is little disagreement between scientists that intra-decadal natural variability (e.g. ENSO, solar cycles, etc.) plays a large role in determining global temperatures, multidecadal variability is considerably more controversial. Some argue that the Atlantic Multidecadal Oscillation (AMO) and Pacific Decadal Oscillation (PDO) may have played a large role in the modern warming period, and may represent a source of variability not properly reproduced or taken into account by GCMs. Others have claimed that almost all modern warming is due to the AMO.
Unfortunately, the AMO in its readily available form is not particularly useful for assessing the degree of natural (unforced) variability during the last few decades in the North Atlantic. A more nuanced examination of the data suggests that about half of the increase in North Atlantic sea surface temperatures (SST) between 1975 and present in the detrended AMO series is due to external forcings rather than internal variability.
To start out, lets examine how the AMO is defined. The AMO is calculated using Kaplan SST data between 1856 and present for the North Atlantic Ocean between 0 and 70N latitude. It is linearly detrended over the entire 154-year period to remove the anthropogenic warming signal.
If we download the Kaplan SST data for the North Atlantic, use a land mask to calibrate grid weights, and create a weighted average of available grid cells for each month, we get the following series:
Each black dot represents a monthly average temperature for the North Atlantic, and the purple line is a lowess smooth (bandwidth = 0.1) to show the trend over time. The data does appear to have a cyclical trend, with troughs in the early 1900s and 1970s and peaks in the 1950s and 2000s. This is even more apparent when we apply the 1856-2010 linear detrending:
The strong increase in Atlantic SSTs in the detrended series nicely corresponds to the modern rise in temperature, and its easy to conclude that natural variability in the Atlantic may have helped drive modern warming from 1975 to present.
This picture has an flaw, however. To remove the anthropogenic forcing, the AMO is linearly detrended from 1856 to present. For this to be effective, however, requires that anthropogenic forcings be constant linearly increasing over the last 154 years, something that is demonstrably false as the majority of greenhouse gas forcing occurred after 1970:
This long-timeframe linear detrending makes it impossible to determine how much of the rise in North Atlantic temperatures over the past four decades is unforced.
Thankfully, there is another way to approach the problem. Instead of linearly detrending North Atlantic temperatures, we can compare them to sea surface temperatures for the rest of the world. Assuming that the multidecadal variability we are examining is limited to the North Atlantic, the difference between North Atlantic SSTs and Global SSTs excluding the North Atlantic will effectively separate out the forced component from the natural variability. Indeed, the PDO already takes this step:
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.
To start with, lets compare Global SSTs excluding the North Atlantic to those of the North Atlantic using the Kaplan SST data:
We see a pronounced rise in global SST data post-1970 that, while smaller than that experienced by the North Atlantic, is still significant. If we define an alternative AMO series as the difference between the North Atlantic SSTs and the SSTs of the rest of the world (differenced AMO), we can explicitly compare it to the conventional detrended AMO:
In the differenced series, the rise between 1976 and present is only about half the magnitude of that seen in the detrended series. This is even more obvious if we replicate the traditional presentation of the AMO data:
This strongly suggests that an analysis of the effects of the AMO should take care to properly remove anthropogenic forcing from the later part of the series. As far as determining how big an impact the positive phase of the AMO has made for global temperatures over the past few decades, I really don’t know. The best I can do is compare Global SSTs with and without the North Atlantic included:
Update
By comparing the differenced AMO series to the original North Atlantic temperatures, we can estimate the portion of warming in the North Atlantic over the past three decades due to natural variability and anthropogenic forcings:
Also, I’d like to thank Steve Mosher for data processing assistance in converting the Kaplan SST data into a more usable format.
Nice work Zeke! The PDO should be the greater of the big two. Have you tried this with the PDO? There is a lot over crossover at the pole (influence on the AO?), it would be neat to see a comparison of the two, AMO and PDO.
Captdallas,
The PDO is already calculated relative to global SST (rather than detrended).
Its also much less correlated with modern warming than the AMO:
http://i81.photobucket.com/albums/j237/hausfath/Screenshot2011-02-23at55917PM.png
Thanks for doing this. Your difference-based approach seems like a reasonable way to try to account for the non-linear anthropogenic forcing. Your last figure attempts to assign north Atlantic warming to either natural (AMO) or anthropogenic causes. Have you tried to similarly assign global temperature warming to AMO – i.e., what fraction of total global temperature change could reasonably assigned to AMO since 1975?
Thanks.
Owen,
It really depends how you approach it. If you assume that Atlantic SSTs only affect the Atlantic, it doesn’t make much of a difference (a few percent less warming) as shown in the prior figure that has full Global SSTs compared to Global SSTs with the Atlantic area removed. I’d be open to trying other methods though.
Interesting that I posted about using the Gulf Stream instead on another thread just before this was posted.
I don’t think it needs to be detrended (the temperature impact of the trend would only be about 0.02C over 150 years).
http://img833.imageshack.us/img833/6932/gulfstreamjan11.png
The rationale for using the Gulf Stream versus the AMO can be seen in the surface ocean currents over the last 30 days.
http://www7320.nrlssc.navy.mil/global_nlom32/navo/WHOSP1_nlomw12930doper.gif
Thanks. I hadn’t see that one, just the U of Washington PDO index. What I had noticed was the surface temperatures in coastal Alaska followed the PDO (somewhat, just eyeball), but the decreases were about half the increases. The PDO and AMO just moves heat around, what happens to the heat would impact the surface temperature anomaly. So I would not expect direct relationship between the oscillations and GTA. Precipitation change with the oscillation may be a better of what impact on GTA they may have. Just thinking a typing (aloud?).
Zeke,
Interesting post. Only a couple of comments.
.
“For this to be effective, however, requires that anthropogenic forcings be constant over the last 154 years, something that is demonstrably false as the majority of greenhouse gas forcing occurred after 1970”
.
Woa! I think you mean “linerarly increasing” not “constant”.
.
“Demonstrably false” I assume is based totally on the graphic you show. It is a substantial leap (huge leap?) to accept any particular estimate of net radiative forcing, as you apparently do with the graphic above. Since each GCM group has their own aerosol forcing history, I wonder what makes you like the particular one you have shown. I may have missed your description of the source, but if that source wasn’t in the post, then it would be good to add a reference. In any case, I for one have very serious doubts about the confidence level of historical indirect aerosol off-sets. If the size of the cloud albedo (indirect aerosol) effect is in fact much smaller than your graph shows (and I personally think it may well be), then the net GHG forcing seems to follow the temperature trend rather nicely. In this case, the forcing increase is in fact a lot closer to linear with time, so the “error” introduced by linear detrending is much less. It is important to note that much of the argument that the AMO index is mostly an effect of global warming and not a driver (as you and Tamino have now both argued) hangs on the validity of some very questionable aerosol forcing estimates. IMHO, “demonstrably false” is a claim that is more than one bridge too far.
There is, of course, a large body of research that shows the AMO is natural quasi-periodic oscillation in north atlantic sea surface temperatures.
In good faith, we should detrend the upward cycle of the last 150 years to make sure it is not contaminated by a global warming signal, but if you go back even further in its history, you would find it has a mostly unpredictable up and down cycle that last for several decades at a time.
The surface ocean currents in the Atlantic, for example, do a figure 8 around the Atlantic with the top and bottom of the figure 8 touching the powerful Antarctic bottom water forming region and also the Arctic bottom water forming region. This would be expected to form some type of oscillation up and down. One might even expect a see-saw between the north Atlantic and the south Atlantic (which actually seems to show up in the records).
Of course, one could dismiss all these studies in an effort to discount the AMO as a natural cycle which affects the global climate. For example, the up-swing which occurred between 1975 to 2000.
Zeke, in your graph above,Global SST from 1906 to present, excluding the Atlantic, appears to be increasing with a nearly linear trend, yet you at the same time claim the AGW has not been on a linear trend (with that slope) for that long – more like since the 1970s. What caused the SST minus the Atlantic to warm at that rate?
SteveF,
You are correct that I should have said linearly increasing, not constant.
As far as aerosols go, it really doesn’t matter that much. If you have CO2 forcing kick in strongly in 1950 instead of 1970 it doesn’t make a 1856-2010 linearly detrending much better at removing an anthropogenic signal.
man that was fast dude. It was fun working with you.
One of your AGW icons Judith Curry has finally fallen off the perch of AGW
http://judithcurry.com/2011/02/22/hiding-the-decline/
and it looks like Mann/Penn will have to answer after all…
http://climateaudit.org/2011/02/23/new-light-on-delete-any-emails/
Good job Zeke, this is a much better method however it still does not account for the possibility that the North Atlantic may warm and cool more(or less) than the global sst in response to forcing and that forcing are not spatially uniform, for instance aerosol may have cooled the North Atlantic more than the global ocean; there are several possible approach to distinguish between forced and unforced variability all inevitably have some kind of limitation, i think that Ting et al. methodology(i.e. Signal to Noise maximizing EOF PC1 as the forced response) is still a better choice.
A description of different methodologies is available here:
http://www.cgd.ucar.edu/cas/cdeser/Docs/DPWG_submitted2BAMS_6Nov09.pdf
Also the PDO despite the global anomaly is subtracted from the local ssta is still influenced by the global trend:
http://www.springerlink.com/content/w661873236444q18/fulltext.pdf
Fascinating thread, Zeke.
One of the problems in disentangling AMO/PDO from CO2 forcings, is that the variation in the AMO/PDO itself acts as a second order forcing. I’ve shown this before, but here it is again:
Given a climate feedback $latex f_0 + f(t)$ where $latex f_0$ represents the slow-secularly varying feedback associated with GHGs and $latex f(t)$ is the feedback (having a mean of zero) associated with atmospheric-ocean oscillations, we can write:
$latex {1\over 1- [f_0 + f(t)]} = {1\over 1-f_0} + {f(t)\over (1 – f_0)^2} + {f(t)^2 \over (1-f_0)^3} +\cdots$
The point is if you average the fluctuations in the climate feedback over time,there is a residual associated with the mean-squared value of the natural forcings.
The relationship between $latex f(t)^2$ and GMT is also an interesting question: It certainly appears that the data support the notion that the various atmospheric-ocean oscillations grow in strength as the temperature increases (one of the drivers for this may be the increase in polar ice melt in arctic summers).
It’s clear that an interaction exists and one that still needs some teasing out.
Zeke wrote, “To start out, lets examine how the AMO is defined. The AMO is calculated using Kaplan SST data between 1856 and present for the North Atlantic Ocean between 0 and 70N latitude.â€
FYI, the NOAA ESRL uses the coordinates of 0-70N, 80W-0 for the Kaplan SST-based AMO data. But the AMO can be calculated using any long-term SST anomaly dataset.
You wrote, “It is linearly detrended over the entire 154-year period to remove the anthropogenic warming signal.â€
You’re assuming the warming signal is anthropogenic. Big assumption since the linear trends of North Atlantic SST anomalies during the early and late 20th century warming periods are very similar…
http://i56.tinypic.com/2vmep3k.jpg
…while the linear trends of radiative forcings are not:
http://i56.tinypic.com/140m8om.jpg
I wasn’t sure if your post made that point.
And another way for me to show you that it’s a big assumption is to remove the linear effects of ENSO and volcanic aerosols from the AMO in the short term, then compare them to scqaled and inverted NINO3.4 SST anomalies in a gif animation:
http://i56.tinypic.com/2repfo0.jpg
Note how the ENSO- and volcano-adjusted AMO data mimics inverted NINO3.4 SST anomalies before and after the upward shift (a shift of about 0.26 deg C). The upward shift occurs during the transition between the 1997/98 El Niño and the 1998/99 portion of the following multiyear La Niña.
You wrote, “By comparing the differenced AMO series to the original North Atlantic temperatures, we can estimate the portion of warming in the North Atlantic over the past three decades due to natural variability and anthropogenic forcingsâ€
Your analysis fails to consider that much of the rise in the SST anomalies for the rest of the world can also be explained without anthropogenic forcings. I’ve written numerous posts for more than two years that explain how and why the East Indian and West Pacific SST anomalies (60S-65N, 80E-180, or about 25% of the global oceans) can and do rise in response to El Niño and La Niña events when a La Niña event follows a significant El Niño event.
Anyway, thanks. You’ve reminded me that I have to finish a follow-up post to my recent post on the AMO.
http://bobtisdale.blogspot.com/2011/02/comments-on-taminos-amo-post_03.html
I’ll also discuss the points you’re trying to make.
Regards
“the difference between North Atlantic SSTs and Global SSTs excluding the North Atlantic will effectively separate out the forced component from the natural variability
I think alot of people have been confused by the smoothing used in most of the charts of the AMO on the internet (and the PDO as well).
The AMO Index has much more variation than the smoothed up and down cycles show and I don’t think the climate operates on those timescales.
Here is the weekly AMO index back to Nov. 1981. Very different than the above charts. Last week, it went down to about 0.0C and will likely go further down since it responds in lagged fashion to the ENSO (sometimes anyway and the current period is one of those times).
http://img40.imageshack.us/img40/3750/weeklyamofeb162011.png
Weekly AMO versus the weekly Nino 3.4 Index.
http://img196.imageshack.us/img196/1078/weeklyensoamofeb162011.png
There are some bizarre leaps of logic in this post.
From the data it looks like the rise since the 70’s in the North Atlantic is about twice the global average.
How you get from that to statements like
“about half of the increase in North Atlantic sea surface temperatures (SST) between 1975 and present in the detrended AMO series is due to external forcings rather than internal variability.” and
“By comparing the differenced AMO series to the original North Atlantic temperatures, we can estimate the portion of warming in the North Atlantic over the past three decades due to natural variability and anthropogenic forcings”
is a mystery.
See Bob’s comment.
Also, perhaps this what what agro was trying to say in his comment above that seems to have been truncated.
I am still stuck on how you have identified the anthropogenic forcing and detrended it?
Zeke, Steve Mosher and others here, have any of you used KNMI to obtain temperature series data in good form?
I have been using it for some time and once you learn your way around manipulating the interactive choices it is really a great tool. I believe the Kaplan SST is available there along with several other SST data sets.
http://climexp.knmi.nl/selectfield_obs.cgi?
Zeke, my observation would be consistent with the Bob Tisdale observations above pertaining to the historical trends in the Atlantic trends. That SST minus the Atlantic temperature sure looks very linear from 1906 to 2010 in your graph. Do you care to comment?
A graph that Tisdale presents above brings out what I want to call the 1998-1999 El Nino breakpoint in the temperature series.
PaulM,
Its really not that complicated. All I’m doing is suggesting that the AMO should filter out the forced component of the signal in the same way that the PDO currently does, rather than using a 154-year linear detrending that is unlikely to accurately capture the actual warming signal over the period.
Bob Tisdale,
I’m not necessarily assuming that the entire post-1975 warming is anthropogenic, just that there is likely a stronger anthropogenic component due to rapid GHG emissions than in prior periods, and a liner detrending would poorly capture this aspect. To put it another way, if you are trying to capture multidecadal variability in the North Atlantic, is there any reason why differencing the North Atlantic from the rest of the SSTs would be worse than a simple linear detrending?
Kenneth Fritsch,
Its a bit faster than linear over the last 30 years, but if you were to generate the AMO record by linearly detrending over the past 100 years instead of the past 150 years you would get a record much more similar to the differencing approach I take. I’ll play around with it later today.
Bill Illis:
Why do you think that?
I suggest this analysis might be of use once you can prove you essential assumption, vis. the measurable trend in global SSTs is “externally forced” or as you leap to imply, caused by AGW.
captdallas2 (Comment#70480): You wrote, “The PDO should be the greater of the big two. Have you tried this with the PDO?”
Contrary to popular belief, PDO data from JISAO does NOT represent the Sea Surface Temperatures of the North Pacific (north of 20N). It also does NOT represent the North Pacific SST anomalies after the global SST anomalies have been removed.
The PDO is the leading principal component of those SST anomalies (a statistically created dataset), after the global SST anomalies have been removed. On a decadal basis, the SST anomalies of the North Pacific are inversely related to the PDO.
http://i52.tinypic.com/15oz3eo.jpg
I discussed that in this post:
http://bobtisdale.blogspot.com/2010/09/inverse-relationship-between-pdo-and.html
And if we compare detrended North Pacific SST anomalies to detrended North Atlantic SST anomalies (the AMO) we can see that multidecadal variations of the SST anomalies of the two ocean basins are similar in magnitude and they run in and out of synch.
http://i56.tinypic.com/t9zhua.jpg
Refer to:
http://bobtisdale.blogspot.com/2010/09/introduction-to-enso-amo-and-pdo-part-3.html
Zeke, I don’t know whether you followed my point or not, but there is likely to be a global contribution to the observed “secular” temperature trend resulting from AMO oscillations. It’s fairly easy to see that if you end up with nonlinear contributions like $latex f_{amo}(t)^2$ that you will get a zero (or very low) frequency component surviving.
Let’s take a cosine model as an example:
$latex f_{amo}(t) = A_{amo} cos \omega_{amo} t$
then
$latex f_{amo}(t)^2 = A_{amo}^2 cos^2 \omega_{amo} t = A_{amo}^2/2 \times [1 + \cos 2 \omega_{amo} t]$
Also, in terms of the question of “what’s potentially wrong” with the “differenced AMO”:
If the AMO modulates global temperature via e.g. change in global albedo, then there should be an in-phase component of the AMO in the global (non-AMO region) SST global signal also.
In my viewpoint, what you have obtained is the forcing-only component on temperature from the AMO, rather than the AMO forcing + climate response to that forcing. That is a much more consistent way to treat the problem, and whether you follow or agree with the rest of my argument, I think we agree on this conclusion.
Others have hit on this final point, which is that not all of the warming seen since 1850 is anthropogenically driven, so that’s not the best choice of words to describe it. According to your final figure, even since 1970, about 1/3 of the warming is natural.
Carrick,
I don’t disagree that the AMO will have some global contribution. The challenge is figuring out what the magnitude of such contribution will be. The North Atlantic is a rather small portion of the total ocean area of the earth, and the apparent cyclical variability in the Atlantic is not really seen (lagged or not) in the global ocean record.
I’ll also grant you that not all of the warming since 1850 is anthropogenically driven, though a more interesting question is how much of it is forced (via solar, vulcanism, etc.) vs. intrensic. In the former case, since we are trying to tease out the natural variability in the system we are agnostic to whether external forcings are anthropogenic or not.
Zeke replied, “I’m not necessarily assuming that the entire post-1975 warming is anthropogenic, just that there is likely a stronger anthropogenic component due to rapid GHG emissions than in prior periods, and a liner detrending would poorly capture this aspect.”
You must’ve missed the upward steps in the ENSO- and volcano-adjusted North Atlantic/AMO data that coincide with the transitions from major El Nino to La Nina events. Here’s the gif animation again:
http://i56.tinypic.com/2repfo0.jpg
Between those steps, the AMO data simply mimics the inverted NINO3.4 SST anomalies. Do anthropogenic forcings only work during the transitions from El Nino to La Nina events in the North Atlantic?
Bob,
While the correlations are interesting, they hardly establish the fact that ENSO results in a step change. Eyeballing it, the match to linearly detrended SST data looks about as good as step-adjusted.
Even if they were step-wise, that’s not proof that the main driver isn’t anthropogenic.
Complex systems can have a series of neighboring metastable states. Things like volcanic action could simply be popping them out of one metastable state into a higher one.
I think that TLS more closely approximates stepwise changes. The spikes are volcanic activity of course.
With Spencer’s PDO theory he is, in effect, saying the global warming is an actual thing but caused by the PDO. Is the same the case with AMO theories or is it that the AMO has caused the appearance of warming (perhaps because of a Western Civ bias)?
If the former then accounting for global SSTs won’t help surely? That is the AMO would be causing global warming and hence surface sea temperatures would rise globally (but perhaps with some lag).
Zeke,
“I don’t disagree that the AMO will have some global contribution.”
.
I don’t think that is really the point. The Nino 3.4 value is a very good predictor of change in average global temperature trends, even though the Nino 3.4 region is a very tiny fraction of the global surface area, and is too small an area to have significant global impacts. The Nino 3.4 index happens to be a good proxy for the very wide-ranging ENSO impact on global temperatures. The real question is if the AMO index is similarly a proxy for large scale natural pseudo-oscillations. On that question, I think the jury is still out. A more productive approach is (I think) to determine if there is a plausible long term pseudo-oscillatory mechanism for which the AMO index (however you detrend it) is a reasonable proxy. I don’t think anybody is really suggesting the the AMO index itself is responsible for the apparent 60-70 year oscillation around a global trend.
Re: Carrick (Feb 24 12:16),
My hypothesis for the apparent stepwise behavior of the TLS time series is that there are two things going on. There’s a short time constant warming and recovery and a long time constant cooling and recovery associated with a major volcanic injection of sulfate aerosols into the stratosphere. I haven’t actually tried modeling this though, much less tried to define a mechanism for the cooling which would have a long time constant.
DeWitt Payne (Comment#70534),
I have also wondered about the odd stratosphere pattern. Could the volcanic aerosols be “accumulating/aggregating/collecting” smaller aerosols that have a very long residence time in the stratosphere? The lack of a significant change (no more cooling) in stratospheric temperatures since a few years after Pinatubo eruption would seem difficult to explain in other ways.
Zeke replied, “Eyeballing it, the match to linearly detrended SST data looks about as good as step-adjusted.”
Please expand on your reply.
“This long-timeframe linear detrending makes it impossible to determine how much of the rise in North Atlantic temperatures over the past four decades is unforced.”
There are estimates for anthropogenic CO2 entry into the atmosphere. Detrend using the line shape of anthropogenic CO2, assuming that increased CO2 is proportional to increased T.
Have a look at what the tend now looks like; make sure your detrending fits 1850-1980.
I’m confused. By detrending the “anthropogenic forcings” out of the data aren’t you assuming want you want to demonstrate. It seems to me that you are saying all long term upward trends are anthropogenic and only variations around that trend line can be attributed to the AMO oscillation. But clearly in the past we have had long periods of upward and downward trends that were not anthropogenic. So why make the assumption that they are in this case.
Is it not possible that a trend could be produced by have more frequent (or less) oscillations in the present era than in previous eras? (not that I have any information about oscillation frequency it’s just a theoretical question) Or maybe having more time in positive half of the cycle than in previous periods (again I have no information about this). If either of these were the case then you could argue that the variations around the trend line were results of the oscillation but the long term trend could then be, at least in part, results of long term changes in oscillation frequency or vertical translation of the cycle. But summarily assigning the trend to anthropogenic forcings seems a stretch.
Having had some time to ingest this post I believe there are several issues that should be raised. Firstly, it is very likely that the AMO can have affects on global sea surface temperatures outside of the North Atlantic basin. In particular Chylek et al (2009) note a bipolar seesaw evident between the Arctic and Antarctic which operates during the positive and negative phases of the AMO.
I will not however argue that the AMO is not contributing to the current warming and I feel that it is a contributor as to why the Arctic is so warm (as has been demonstrated by my posts at SKS) but the phase of the AMO also contributes by reducing sea ice coverage further causing feedbacks. Our anthropogenic forcings will undoubtedly continue to overwhelm the power of the AMO in the future particularly due to these feedbacks.
Nevertheless it is important to remember that the AMO was first diagnosed as a global phenomena by Schlesinger & Ramankutty in 1994. Another interesting thing to consider is that the AMO was predominantly positive during the MWP and predominantly negative during the LIA according to this abstract :
http://meetingorganizer.copernicus.org/EGU2010/EGU2010-13508.pdf
Makes me almost think broeker had a great idea…
Bob Tisdale,
I meant that I can’t easily tell if a step-change correction or a linear detrending over the time period shown would create better correlations, at least by eyeballing it. Have the data handy so I can play with it?
JKnapp,
I’m not detrending “anthropogenic forcings†out of the data. Rather, I’m removing the trend by differencing global SSTs excluding the Atlantic from Atlantic SSTs. The idea is that the AMO will mostly be confined to the Atlantic, while the remainder of global SSTs will mostly reflect exogenous forcings. Its not a perfect approach, but I’d argue that it does a better job of removing external forcings than a 154-year linear detrending.
Very interesting Zeke. Here is another interesting thought: how do theses patterns in the real world trade off against similar patterns produced by the models. If the models are accurate physically, we should get swings at least somewhat similar ( after all, this effect rises well out of the noise). Can’t you use model oscillations to CO2 concentrations to get an intuitive feel for the problem from different direction (even if not completely accurate, I would think it could serve as a starting point)
@ Bob Tisdale,
Thanks for the links. I have noticed that several NH higher lat areas have had quick surface temperature jumps followed by slow or little decline. Scandinavia around the time the first electronic surface stations were installed. At first I thought it was instrumentation, but the jump of around 3 C persisted. Then many of the Alaskan stations have fairly quick ramp ups with more gradual declines that seem to correlate with PDO. Is there a known reason why the temperatures seem to get stuck?
http://data.giss.nasa.gov/cgi-bin/gistemp/gistemp_station.py?id=645022260004&data_set=1&num_neighbors=1
There is one wicked jump 1989-1990 in the GISSTEMP plot for the whole area that just levels off?
I guess I misunderstood what you meant when you wrote, “To remove the anthropogenic forcing, the AMO is linearly detrended from 1856 to present.” and then a bit later, “Thankfully, there is another way to approach the problem. Instead of linearly detrending North Atlantic temperatures, we can compare them to sea surface temperatures for the rest of the world.”
And then your final graph has green natural forcings which came from the AMO and red which are labeled “anthropogenic forcing” which came from the detrending that you did.
Are you now saying that the label “anthropogenic” was in error and should have just said “unknown forcing resulting in trend”?
It still seems to me that you are assuming what you want to demonstrate.
Re: Kenneth Fritsch (Feb 24 09:13),
“Zeke, Steve Mosher and others here, have any of you used KNMI to obtain temperature series data in good form?”
No. I just get the data from ftp sites. Looking at KNMI it looks like there is a user form to fill out..
I dont want to use any manual process to retrieve data.
Kenneth Fritsch (Comment#70507): You asked, “Zeke, Steve Mosher and others here, have any of you used KNMI to obtain temperature series data in good form?â€
I use it all of the time. Rarely do I use the original source. AMO data is an exception.
Zeke: The Menu.
ESRL AMO data:
http://www.esrl.noaa.gov/psd/data/correlation/amon.us.long.data
Kaplan NINO3.4 SST anomalies (click on “raw data†above third graph) from KNMI Climate Explorer:
http://climexp.knmi.nl/getindices.cgi?NCEPData/nino5+NINO3.4+i+someone@somewhere
Northern Hemisphere Aerosol Optical Thickness data from GISS:
http://data.giss.nasa.gov/modelforce/strataer/tau_line.txt
As noted in the title block of the graph…
http://i56.tinypic.com/2repfo0.jpg
…to create the ENSO- and Volcano-adjusted AMO data, I used a scaling factor of 0.18 for the NINO3.4 data with a 2-month lag and a scaling factor of -1 for the AOT data.
captdallas2 (Comment#70550): There are many land surface temperature and sea surface temperature subsets with ENSO-induced upward steps. The major steps like the 1976-77 Pacific Climate Shift are the only ones that I’ve seen studied, but there is little agreement on the cause(s).
The following is a map of the annual change in global temperatures (GISS LOTI) from the start of satellite-based SST measurement to present (1982-2010).
http://data.giss.nasa.gov/cgi-bin/gistemp/do_nmap.py?year_last=2011&month_last=01&sat=4&sst=1&type=trends&mean_gen=0112&year1=1982&year2=2010&base1=1951&base2=1980&radius=1200&pol=reg
If you were to plot the temperature data of the areas with the greatest rises in temperature (Northern Canada through Greenland or the mid-latitude area that stretches from the Mediterranean Sea across Asia) you’d also note the upward steps. Never seen that addressed either. But they correlate with the AMO:
http://i55.tinypic.com/opsbv9.jpg
The correlation map is Figure 16 from this post:
http://bobtisdale.blogspot.com/2011/01/removing-effects-of-natural-variables.html
Even if you detrend the AMO, it’s cycle is still +/-0.6C.
My regression modelling shows that this variation has a +/- 0.35C impact on global temperatures. [And I can model the monthly temperatures back to 1871 very accurately compared to a climate model using the methodology].
http://img8.imageshack.us/img8/4408/hadcrut3modeljan11.png
So detrend it and you still have warming of 0.35C between 1975 to 2000.
Now that the AMO is going below 0.0C (and will probably reach -0.25C by the summer), are we going to blame the continued global cooling over the next several months on natural variation?
Why is there natural variation on the way down only and none on the way up?
Just look at the monthly global temperatures. Do you think there is no variation in those numbers (well, there isn’t, if you smooth it over 25 years, but the climate operates on daily, monthly and 3 month timeperiods, the smoothing just reduces the information which is available to understand the climate).
Isn’t there an assumption being made that the global SST increase is anthropogenic? Is there any proof of that assumption?
A very interesting PDF format presentation available on line, which includes discussion of the AMO is:
“Decadal-to-Centennial Modes of Climate Variability Over the
Past Millennium
Edward R. Cook
Tree-Ring Laboratory
Lamont-Doherty Earth Observatory
Palisades, New York
His conclusion is that the weight of the evidence points to long term historical oscillations with 35-80 year periods being most important, but with longer term periods as well. But he is not able to reject (with statistical confidence) the null hypothesis of simple red-noise driven climate variability.
Zeke, with regards to your comment about the PDO not being co-related with Global SSTs, I agree to a point. Ask anyone who lives next the ocean, especially the west coast, that ocean temperatures moderate the land temps. While it may be true that globally there isn’t a corelation but the Pac NW (Oregon, Washington, British Columbia and Alaska) temperatures have a CLEAR co-relation with PDO. (http://fnieuwenhuis.xanga.com/735341436/pacific-decadal-oscillations-effect/)
Re: SteveF (Feb 25 10:01),
But that’s probably exactly what it is, except the noise isn’t AR(1). It’s deterministic chaotic. According to Koutsoyiannis, systems with high Hurst coefficients often show quasi-periodic behavior. Dollars to doughnuts that if you take the time series and fit it with an ARFIMA (p,δ,q) model, δ will be close to 0.5.
Just to push the point a bit further.
How are you sure that you are removing a trend and not (positive monotonic) portion of a cycle?
Zeke: Now that I’ve got the other stuff out of the way I thought I advise that I agree with your method of removing the global warming signal from the North Atlantic; that is, to subtract the Rest of the World SST anomalies from the North Atlantic SST anomalies. In fact, that was one of the points I made in my recent post on the AMO.
http://bobtisdale.blogspot.com/2011/02/comments-on-taminos-amo-post_03.html
I started my discussion of that topic after Figure 6 and presented the results compared to the detrended version of the AMO in Figure 7:
http://i53.tinypic.com/2v1ukg5.jpg
So why are my results so much different than yours? You used Kaplan SST anomaly data, and I used HADISST data because that’s the long-term dataset employed by GISS in their LOTI data, which was also a topic of that discussion.
Bill Jamison (Comment#70561) February 25th, 2011 at 7:12 am
“Isn’t there an assumption being made that the global SST increase is anthropogenic? Is there any proof of that assumption?”
.
If you are only looking for a food fight, then there is no need to reply to this comment.
.
Assuming you are not looking for a food fight: The effort is to separate the possible natural variation/pseudo-oscillations from a GHG driven causal trend. There is no proof of anything, but there are lots of good reasons to believe that rising GHG levels (and not just CO2), should cause some warming, and there is lots of data which suggests natural variation is also significant. Some (like Tamino) will try in every possible way to discount the contribution of any natural warming over the last 50-100 years. Others will try in every possible way to assign all recent warming to GHG forcing. It is almost certain that reality lies somewhere in between; where between these two extremes reality lies is important for evaluating the potential risks for future GHG driven warming. If more of recent warming is due to natural variation, then less is due to GHG driven warming…. so Earth’s climate has to be less sensitive to GHG forcing.
In reading Zeke’s post and comments, it is interesting to review a couple of papers by Guan and Nigam which Bob Tisdale linked to in a post he did back in November. G&N09 remove the effects of both Pacific variability, and non-stationary global sst trend influences (by first detrending global sst IIRC) on N Atlantic variability and then do EOF analysis on the N Atlantic residuals.
@SteveF
“Some (like Tamino) will try in every possible way to discount the contribution of any natural warming over the last 50-100 years. ”
That is not true. Just because Tamino is trying to convince ‘skeptics’ that there is AGW, it doesn not mean he discounts natural variance. You are just making an assumption. The IPCC report shows what natural variance they believe there is in the past century.
re: Bob Tisdale’s question:
–
IIRC, Zeke’s “differenced” AMO series removes global sst excluding N Atlantic, while yours removes the entire global sst.
Bob,
I’m not sure why your results were that different from mine. I used Kaplan because the “official” AMO record is based on it. I validated my spatial gridding model by testing it against the AMO record (http://www.esrl.noaa.gov/psd/data/correlation/amon.us.long.mean.data). It wasn’t perfectly identical (I used 0 to 90 lat and -90 to 0 lon, where I think they use -80 to 0 lat), but it was almost the same.
I’m also applying a land mask in my weighting scheme, so grid boxes containing mostly land don’t get much weight. The code is fairly straightforward:
*————————————————————————————*
*Importing code (run once)
/*
clear
insheet using “/Users/hausfath/Dropbox/Energy Science/CC Data/Kaplan_SST.txt”, delimiter(” “)
rename v2 lon
rename v3 lat
rename v4 month
destring v5, replace force
rename v5 anomaly
drop v1
replace lon = lon – 360 if lon > 180 // lon was 0 to 360 in original data file
gen latseperator = ” lat ”
gen lonseperator = ” lon”
egen gridbox = concat(lat latseperator lon lonseperator)
drop latseperator lonseperator
sort gridbox
merge gridbox, using landmask.dta
drop if _merge != 3
drop _merge
gen grid_weight = sin((lat+2.5)*_pi/180) – sin((lat – 2.5)*_pi/180)
replace grid_weight = grid_weight * ocean_percent
save kaplan_sst_base.dta, replace
*/
*————————————————————————————*
use kaplan_sst_base.dta, clear
*drop if lat > 0 & lat < 70 & lon -90 // No north atlantic varient
*keep if lat > 0 & lat < 70 & lon -90 // North atlantic varient
*Average all anomalies for each month in each year in each grid.
collapse (mean) anomaly [iweight = grid_weight], by (month)
sort month
save kaplan_global_SST.dta, replace
lowess anomaly month, bwidth(0.1) gen(global_sst_lowess)
Bill Illis,
If I detrend the global surface temperature over the past 150 years, I can easily show that the detrended portion is responsible for the a good portion of warming over the past four decades. Does that mean that this warming was due to natural variability?
Zeke (Comment#70594)
February 25th, 2011 at 3:02 pm
Bill Illis,
If I detrend the global surface temperature over the past 150 years,
————————
Thanks Zeke, Great suggestion. It really shows what the “natural variability” people have been talking about for a long time now.
This could end up becoming an iconic chart. I will certainly be spreading this around.
http://img232.imageshack.us/img232/2091/hadcrut3detrended.png
I dare say it almost looks exactly like the AMO.
Breaking time series into trends and cycles is a tricky business. Terence Mills is a well known expert on this in economic time series and he has produced some climate work too. See his 2010 Journal of Cosmology article here
http://journalofcosmology.com/ClimateChange112.html
Here is the abstract
“The last two decades has seen a major research focus on modelling trends in time series, although little consensus has emerged, particularly when temperature records are analysed. This paper considers fitting a flexible model, known as the structural model, to global hemispheric temperature series for the period 1850 to 2009 with the intention of assessing whether a global warming signal can be detected from the trend component of the model. For all three series this trend component is found to be a simple driftless random walk, so that all trend movements are attributed to natural variation and the optimal forecast of the long term trend is the current estimate of the trend, thus ruling out continued global warming and, indeed, global cooling. Polynomial trend functions, being special cases of the general structural model, are found to be statistically invalid, implying that to establish a significant warming signal in the temperature record, some other form of nonlinear trend must be considered.”
Bill,
Odd that it looks so similar to the AMO, especially in recent years. Perhaps the same forcings are acting on both, and those forcings are not linear over the past 150 years? 😛
On second thought,
It is the AMO, holy moley.
http://img689.imageshack.us/img689/9449/hacrut3detrendedandthea.png
I don’t know what you think, but Zeke just put us down a path that explains the major cycles in the climate over the past 140 years.
This little chart will have to be published in Nature now.
http://img689.imageshack.us/img689/9449/hacrut3detrendedandthea.png
The remaining trend is only 0.053C per decade and would be a consistent number throughout the period.
Re: mikep (Feb 25 18:05),
From the discussion, section 4 of your reference:
Which is exactly what you would expect for Hurst-Kolmogorov statistics, which implies a fractional integration order. The real problem is that the instrumental temperature time series is too short. None of those tests are worth a dime for a series with less than 150 points. You need 1,000 or more.
@SteveF (Comment#70537)
The MSU channel4 temperature trend is mainly due to ozone concentration, carbon dioxide increase contribute to a ~ -0.1K/decade that is still lower than the 1979-2010 trend (~ -0.3k/decade) despite lack of cooling since 1993 that is due to a small ozone recover superimposed to co2 cooling, ozone concentration are also altered in the two years following large explosive volcanic eruption, for a full description see Thompson et al.2009:
http://www.atmos.colostate.edu/ao/ThompsonPapers/ThompsonSolomon2_JClimate2009.pdf
Zeke asked Bill Illis (Comment#70594), “Does that mean that this warming [of detrended global SST anomalies] was due to natural variability?”
You don’t know for certain whether or not the rise in the “undetrended” global SST was due to natural variability. You’re assuming the rise is anthropogenic. Big assumption
Bill Illis (Comment#70637) says: “I don’t know what you think, but Zeke just put us down a path that explains the major cycles in the climate over the past 140 years.
“This little chart will have to be published in Nature now.”
But you’d have to use a HADSST2 version of the AMO for the comparison to HADCRUT.
Layman Lurker (Comment#70591) says: “IIRC, Zeke’s ‘differenced’ AMO series removes global sst excluding N Atlantic, while yours removes the entire global sst.”
Nope. My differenced AMO series also “removes global sst excluding N Atlantic”. That’s why it’s identified as the “Rest of the World” in the graph:
http://i53.tinypic.com/2v1ukg5.jpg
Zeke (Comment#70592) says: “I’m not sure why your results were that different from mine.â€
It results in part from the differences between a “globally complete†SST dataset (HADISST) and a dataset that is globally incomplete (Kaplan). Kaplan excludes the Arctic and much of the Southern Hemisphere, south of 45S. Here’s a map showing the Kaplan grid coverage in 1870:
http://i53.tinypic.com/2q3zzib.jpg
And here’s the Kaplan map in 2000:
http://i53.tinypic.com/35lrp8i.jpg
So the drop in the SST anomalies that’s been taking place for the past few decades in the high latitudes of the Southern Hemisphere is excluded in the Kaplan “global†data.
After infilling, the Hadley Centre also reinserts the measured SST, but I’m not sure about Kaplan.
You continued, “I used Kaplan because the ‘official’ AMO record is based on it.â€
And as I noted earlier, I used HADISST because I was comparing it to GISS LOTI (in that post and the earlier posts) and I didn’t want to make the same mistake Tamino did in his AMO post. In fact, I wanted to illustrate the error he made when he compared Kaplan North Atlantic SST anomalies to GISS LOTI.
You wrote, “I validated my spatial gridding model by testing it against the AMO record (http://www.esrl.noaa.gov/psd/d……mean.data). It wasn’t perfectly identical (I used 0 to 90 lat and -90 to 0 lon, where I think they use -80 to 0 lat), but it was almost the same.â€
The ESRL states on their webpage that they use 0-70N for latitudes, and I confirmed via email that they use 80W-0 for longitudes. But you’re right, of course: there should be little difference. In fact, you would find there would be little difference if you had expanded the longitudes to the east so that you included the Mediterranean and Black Seas, since together they have a multidecadal variability that’s very similar to the North Atlantic.
Regards
“Moreover, the AMO shifted into its warm phase in the 1990s (Fig. 1), which may have accentuated global warming in this period. A return from a warm to a cold AMO phase could temporarily mask the effects of anthropogenic global warming47, and thus lead to possible underestimation of future warming if the variability of the AMO is not taken into account.”
Tracking the Atlantic Multidecadal Oscillation through the last 8,000 years by Kundsen et al 2011
Nature Communications
And yet the models fit the temperature behavior perfectly without taking the warm phase of the AMO into account.
Interesting.
Interesting link, Robert. The authors are concerned that a future shift of the AMO to the cold phase could mask anthropogenic global warming and lead to underestimation of future warming.
Well, the authors could also have pointed out that since climate models have hitherto ignored multidecadal internal ocean variability we may have overestimated future warming with the AMO in its warm phase since 1990 and firmly in its cold phase around the year 1900. I would like to see the authors include the AMO in an attribution study. To what extent does the suggested AMO effect on global warming crowd-out the anthropogenic effects?
JohnM:
They are also making assumptions about the forcing, in particular sulfates are a “kludge” that helps the models “fit the temperature behavior perfectly”.
(Not really “perfectly” either.)
Carrick,
Agreed. Actually, I meant “since 1960”, as I’ve pointed out before with this graph.
http://co2now.org/images/stories/ipcc/ar4-wg1/faq_8.1_fig_1_global_mean_near_surface_temperature_since_1900.gif
I know you don’t trust the pre-1950 temperature data, but if they’re going to fit a model to it, they might as well acknowledge the fit ain’t so great in the early 20th century.
Here is the monthly AR4 hindcasts and forecasts versus Hadcrut3 from 1900 to 2020.
This is the ensemble A1B mean of the 23 models available on the Climate Explorer from PCMDI. I’ve also included the monthly ocean cycle reconstruction method which appears to be much more accurate and projects much less warming occuring to date (half).
I’ve also pinned the ensemble mean forecast to 2003 which is giving the models a slight break since the cut-off was supposed to be 2000 although they had until the year 2003 to submit the forecasts. Using 2000 would have the models off by quite a bit more as of today.
1990 to 2020.
http://img9.imageshack.us/img9/5567/ar41900mmmeanvshadcrut3.png
1979 to 2020.
http://img837.imageshack.us/img837/1584/ar41979mmmeanvshadcrut3.png
I don’t think the “external forcing” methodology/theory is accurate enough to be relied on.
gp2 (Comment#70664),
I sure which Thompson et al had provided more information about the proposed chemical effects in the several years following a major eruption…. and I hope even more that they are not just ‘guessing/assuming’ a chemical reaction following El Chichon and Pinatubo eruptions is responsible for changing ozone. Did you see some explanation of the proposed chemical processes in the article that I missed?
I agree, Bill. I don’t think the models are good enough to be used as forecast tools, even if you knew the forcings. What we can hope they are adequate for (and what they were originally tooled to do) is predict the climate sensitivity to a doubling of CO2.
Could somebody point me to where the A1B forcings are located at?
Bill Illis (Comment#70808),
Don’t worry about the AR4 projections Bill; we will soon have the new-and-improved (but much less certain) AR5 projections. These will (of course) be contorted to almost exactly fit the data through ~2011. Then we will need to wait another 10 years before the models begin to look bad again.. just in time for AR6 to generate new projections…. starting in 2021.
.
The whole exercise would be quite humorous, were it not that so many people actually believe the predictions enough to want to take public action. It is as intellectually corrupt an activity as I have ever encountered in science. Better to just stop all future AR_X projections and try to make the models a better representation of reality. Until all the models pretty much agree on both historical forcings and calculated climate sensitivity, IMO it is a waste of time to even begin to consider if their projections are credible.
Re: Carrick (Feb 27 19:03),
The various scenario data are here: http://sres.ciesin.org/final_data.html
But those don’t include any estimates of aerosols or solar. It’s all ghg’s. It is interesting to total up the fossil fuel consumption through 2100 assuming a linear interpolation between the decade values in the table. Highly optimistic.
The Thompson and Solomon paper is an interesting analysis of stratospheric temperature and ozone.
I still haven’t completely digested everything in it, but I’m left with the impression that they had three goals:
1) Explain the step-change in temperature related to the volcanic eruptions, and why a new base-line is established after the step-change.
2) Find evidence to support that stratospheric ozone levels are recovering.
3) Find evidence for a ghg-caused decrease in stratospheric temperature.
The study presents a fitting procedure based on temperature and ozone measurements. Error bars are presented for trends in both (Fig. 5), yet none of the other figures show error bars. The fitting procedure is designed to separate the effects of stratospheric ozone and tropospheric ghgs.
I wonder how the trends they’ve teased out of the data to support 2 and 3 from above would compare to the errors, if they were to carry them through.
“something that is demonstrably false as the majority of greenhouse gas forcing occurred after 1970”
So what caused the 1910 to 1940 warming?