May 14 2008
Timescale of the PDO, NAO, and AMO
Earlier this month, JPL announced that the Pacific Decadal Oscillation (PDO) switched from its warm phase to its cool phase. Many a blog have jumped on the bandwagon to post on this ‘news’, even Skeptical Science. Cross and Cook write show a plot of the PDO index and the global temperature anomaly on the same graph. The temperature anomalies have a definitive trend, while there is little-to-no trend in the PDO.
They write that “[w]hile PDO does have some degree of correlation with short term variations in global temperature, the striking feature of [the figure] is the contrast in trends between PDO and global temperature.” I can understand why it might seem like a good idea to compare the two, especially given the recent blogospheric reaction to the JPL news. However, their analysis is flawed for one simple reason:
By definition, the PDO cannot have a trend!
From the University of Washington, we see that the PDO is defined as “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.”
Ignore the part about being the leading principle component (PC) blah blah blah part for now. Focus on the second sentence. “The monthly mean global average SST anomalies are removed…” This means that before any processing of the data is done to determine the PDO, the trend is removed! This is why they don’t see any trend in the PDO while there is a trend in the temperatures.
Signal vs. Noise
When diagnosing global warming, (almost) any oscillation is noise. Since the total radiative forcing has increased roughly monotonically since the 1900s, the global energy should also have increased roughly monotonically. Therefore, any cycles such as El Nino will not contribute to the overall long-term trend when enough years are used to calculate the trend.
The problem arises when we are calculating trends over time scales where there are significant oscillations. During the last few months, there has been more than a few blog posts about the recent surface temperature trends; using time spans ranging from as little as 5 years. When this happens, the oscillation has a large influence on the calculated trend (but not the actual underlying trend).
NAO, PDO, AMO
In addition to the PDO, there has been discussion about the influences of the NAO (North Atlantic Oscillation) and AMO (Atlantic Multi-decadal Oscillation) on global and regional temperatures. A recent commenter pointed me towards this piece by D’Aleo. I have already shown that the PDO cannot, by definition, have a trend. The same is true for the AMO. From the CDC page on the AMO, we can see in step 3 they detrend their sea surface temperatures. Thus, there can be no trend in the AMO.
It can still be fun to look at the time series though. I’ve plotted only the data that is provided on the CDC website (except the PDO data from UW).
|
The fluctuations in the NAO (top panel) have a very short time scale. This is because it is defined as a difference in atmospheric pressures at two locations in the North Atlantic. Pressure variations, unlike sea surface temperatures, have very little “memory”; that is the average pressure this month will not be very highly correlated to last month. On the other hand, sea surface temperatures have memory; the values from month-to-month will be highly correlated. There is very little going on in the NAO data on long time scales.
The PDO does appear to have an oscillation with a period of 80 years. I’m just eyeballing the graph; it appears that there is a switchover about every 40 years. Since the last PDO ’switch’ came in about 1975, it appears that it’s about time for another.
The third panel has the time series from the AMO. Like the PDO, it appears that there is an oscillation with a long period. Just eyeballing the graphs again, it looks like the AMO has a period of about 80 years, just like the PDO. Interestingly, it appears that the AMO and PDO are about 270 degrees (30 years) out of phase.
Wavelets: like waves, only smaller
To get a better estimate of the period of these oscillations, we can use the wavelet blackbox. This method basically find the dominant periods at all of the time steps. Have a looksee at the graphs below.
|
This first plot is the wavelet analysis of the NAO. White and green mean there is little to no significance to those periods. As the colors change from green to blue to red, the significance of the periods increases. The magenta shield shaped line indicates the confidence region. “Below” that line are area of low confidence. That means there are not enough points to be sure that those periods are significant. 99% confidence contours are also shown. There are no systematic long term periods in the NAO data.
|
The PDO is a little more interesting because it shows the big bullseye in pink for most of the years at periods of 70-80 years. There are other periods above the magenta line that have significance above 99% but do not show for all of the years calculated. For instance, there are many years that show a strong period of 20 years, but only from 1930-1980.
The 70-80 year periods that are usually considered the “normal period” for the PDO exist outside the magenta confidence interval. This does not mean that there is not a period of 70-80 years in the PDO data. It does mean that we should be careful about interpreting the PDO data, since from the graph about, there is only 1.5 cycles at this period.
|
The AMO wavelet is extremely similar to the PDO. The pink bullseye shows nicely at 70-80 year periods, but below the magenta confidence line. Unlike the PDO, there is no significant periods at around 20 years.
Conclusions
It is important to be careful when making statements about long period oscillations in a time series that is short. When there are less than 2 full periods in the record, it is very difficult to say what the actual main period of the oscillation actually is.
When comparing two climate metrics, it is important to know their definition. It makes no sense to compare the temperature trend of the PDO to anything, since the PDO has no trend. If it is necessary to compare the PDO to the global temperatures, it is first necessary to detrend the temperatures or to calculate an adjusted PDO that includes the trend.
(1 votes, average: 1 out of 5)
Loading ...Related Posts:
11 Responses to “Timescale of the PDO, NAO, and AMO”
To reduce spam, comments are automatically closed 30 days after the last comment. If you would like to comment on any closed thread, please use the contact form at the top of this page.
The warming caused by CO2 is indeed a trend and PDO is a cycle. It is difficult to deconvolve a trend from one or more cycles (PDO, AMO,etc.) and thus one is skeptical about the use of models parametized over short time intervals. The models so parametized predict huge positive feedbacks yet the temperature record of over 100 years shows a trend consistent with little or no feedback. I believe that is the strong message of the graph showing the time-series data of both global temperature and PDO index.
[Reply: I'm not sure how you go from difficult to deconvolve a trend with cycles to being skeptical of model parameterizations. Can you fill in the steps?]
Atmoz, I’m afraid you’re reaping the benefits of trading blogroll listings with AW.
For our denialist friends, the PDO is a handy black box into which they will attempt to stuff just about anything. BTW, just recently the PDO acquired some competition. Long live the NPGO!
1. If there are a number of cycles with frequencies and amplitudes not well known, it becomes quite a challenge to tease out a trend. From what I am given to understand, these models have one thing in common namely, their ability to “retrodict” temperature anomalies for the last decade of the last century. If the temperature anomalies of this decade were significantly mediated by a “positive” cycle —PDO comes to mind— then any adjusted model parameters are suspect. This would not occur, of course, if the models were truly ab initio and that is impossible at this stage of sophistication.
2. Looking at the second graph presented in Skeptical Science it appears to me that the best estimate of temperature rise is either the line from crest to crest or trough to trough of the temperature anomalies and these lines give a decadal rrise in temperature of under 0.1 C.
Hello Atmoz! Here is a message to Joe D’Aleo that I attempted to post in the comment box at ICECAP, but was informed by the computer that I “was unauthorized to use this form”.
ATTN: Joe D’Aleo.
RE : Evidence for a New Pacific Ocean Oscillation
Last year about this time, I started detailed analyses of the temperature records of the weather station and lightstation at Quatsino. BC, which is located at the northwestern tip of Vancouver Island. To see the site of the weather station and lightstation, GO: http://www.fogwhistle.ca/bclights/ .
I undertook this study to test the AGW hypothesis, which I flatout reject. By analyes of the records, I hope to show that there is no empirical evidence for any global warming as the late John Daly did for a great many remote weather stations such as Alice Springs and Death Valley. The change in annual mean temperature at Alice Spring is 0 K since 1879.
I selected this station on the basis of the criteria of the late John Daly for analyses of station temperature records. I did a preliminary analysis of the records and prepared a short report which I sent to Roger Sr last summer. Since that time I completed a multi-decadal analyses of the records using unorthodox methods and have found some unusual results which are quite pertinent to your study of multi-ecadal ocean oscillations.
Here is brief summary of the results of the analyses of the records for the September Equinox Interval ( SEI, Sept 16-26) from 1895 to 2007. The data are reported as: Interval, mean Tmin (or Tmax) K.:
Tmin Temperature-Time Series : 1895-1899, 281.3, 1900-1929, 280.5, 1930-1989, 281.8, 1990-2005, 283.1, 2006, 282.7, and 2007, 282.4.
Tmax Temperature-Time Series: 1895-99, 289.7, 1900-39, 289.2, 1940-79, 290.4, 1980-2005, 290.3, 2006, 289.1, 2007, 288.5.
Note: The mean classical average deviation for the multi-decadal intervals (1900-2005) is +/- 1.5 K.
For the Tmin temperature-time series, note (1) the two temperature jumps of 1.3 K at 1930 and 1990, (2) the long interval of constant mean temperature, and (3) declining mean temperatures for 2006 and 2007.
For the Tmax temperature-time series, note (1) the single temperature jump of 1.2 K at 1940, (2) the long interval of constant mean temperature (1940-2005), and (3) the dramatic drop in mean temperatures for years 2006 and 2007 to values below that for the 1895-99 interval.
Global Warming? Not at this site. Note also no cooling in the 1950-70 interval.
The length of the cycle is about 65 years or twice that of the PDO.
I selected September for this analysis because in BC any effects of the ENSO are minimal, which is not the case for March, June and December. For these intervals, it is nothing but a trouble-maker and messes up the data.
You can clearly see that these results are not consistent with the PDO which shifted into a cool phase ca 1940 and a warm phase at ca 1975. Since this site is at the ocean-land interface, Tmax is a measure of the sea breeze and Tmin is a measure of the land breeze. Thus it is not appropiate to use the annual mean temperaure metric. The Tmax and Tmin are measuring two different types or states of the air.
I have also analyzed the temperature records for Sept 21 to determine the effects of other natural forcings except changes in sunlight. The results are quite remarkable. I also have done calculations for the lightstation records. The lightstation has the same latitude and longitude as the weather stations but is at a higher elevation (21 vs 7 meters). The lightstation records only start at 1979, however.
-=-Harold D. Pierce, Jr, B.Sc(Hon)., Ph.D.
PS: I’m a pot-boiling organic chemist. Since I don’t want to pay BC carbon taxes. I want shut down all of this global warming gobblygook and climate claptrap ASAP!
ATTN: Atmoz, you want a look-see at QUATSINO.TXT? Its free!
The world anxiously awaits the comments of professionals who will read Harold’s work in the science journals.
Which journals, Harold, are you submitting this paper to? After all, the scientific community doesn’t get its science from comments on blogs…
Best,
D
Atmoz — This is quite nicely done. So nice that I would very much like to persuade you to do a similar analysis for the GISP2 ice core temperature study by Alley, accessible from the NOAA Paleoclimatology web site.
To keep it fairly short, I suggest using just the data for the Holocene, say 10,428 ypb to the end, 100 ypb. I suppose you’ll need to intrpolate as the samples are not at exact time intervals; I found linear interpolation between the nearest data points works as well as anything.
Having done this using a Lomb periodogram based method, I suggest you ought to find significant power for periods between 45 and 90 years, about 85% of the time,, but nothing significant for longer periods. However, I am not entirely pleased with my attempt and I’m not really setup to do a nice wavelet based presentation.
I do hope I have enticed you!
ATTN: Dano
Roger Sr has already checked the preliminary data and he liked it.
I’ll request that he post it on his blog, which is about as close as it get to a peer-reviewed journal. Besides its free, and he does all the nitty-gritty editing and formating, And I don’t want the hassle of preparing a formal ms and submitting to a journal especially one that assess page charges.
No comment and criticisms? Take your best shots, guys, so I can fix any screw-ups and deficienies, like omission of recent research results.
I wouldn’t dismiss blogs as a medium for disseminating scientific info. With one click of a mouse, it can be sent anywhere and everywhere especially to politcians who would never look at a journal. Ya know, it doesn’t take much to razzle-dazzle these guys!
[Reply: I guess I don't understand what you did. My interpretation is that you looked at temperature data from one station and then conclude that there isn't any global warming. Is that correct?]
The overlap of the AMO and PDO cycles is interesting. Recalling that constructive and destructive interference phenomena when evaluating two cyclical events occurring simultaneously, both cycles appear to be in the cool phase of the cycle during the 70’s (when global cooling was the rage) and both have been in their respective warm phase during the late 90’s (the advent of the global warming theories).
The trends look to like a moderation of the warming will be next, followed by another cooling trend.
Hey Dano, there’s a paper for you. Compare the overlapping influence of the two cycles against recorded temperature trends.
ATTN: Crashex
Joe D’Aleo has already done this. Go over to ICECAP and read his articles, and check out the graphs.
Harold, thanks for the D’Aleo reference.
The graphs illustrate a good visual trend match between the arctic temps and the AMO and PDO cyles.
Makes me wonder why all the ice extent comparisons are always to the 1979 to 2000 trends and averages. The AMO and PDO cycles move from trough to peak over that specific time frame. That’s only half the wave length–its like using a cosine wave from 0 to 180 and ignoring the 180 to 360 section of the overall cycle.
Why aren’t current ice extent conditions compared to years that were at a similar part of the cycle?
ATTN: Dano
Roger Sr has already checked the preliminary data and he liked it.
I’ll request that he post it on his blog, which is about as close as it get to a peer-reviewed journal.
So the scientific community won’t see it. Who cares? Why bother? What good is it?
Best,
D