UPDATE 10-25-2012: My single word answer was found to be confusing after the heading of DO GLOBAL TEMPERATURES RISE AND FALL IN RESPONSE TO A POSITIVE AND NEGATIVE PDO? I’ve changed the answer to be a complete sentence to avoid the confusion.
>UPDATE (September 5, 2010): I’ve added a comparison of Detrended North Pacific SST anomalies (north of 20N) to detrended North Atlantic SST anomalies (the AMO) at the end of the post. And I corrected the title of the subheading in the following to read “The PDO Is Not Calculated Similarly To The Atlantic Multidecadal Oscillation (AMO)”. It was missing the word calculated.
This post presents an overview of the Pacific Decadal Oscillation (PDO) and is intended to provide the reader with a basic understanding of what the PDO represents, and, just as important, what it does not represent. The Sea Surface Temperature (SST) side of the El Niño – Southern Oscillation (ENSO) was discussed in An Introduction To ENSO, AMO, and PDO – Part 1, and An Introduction To ENSO, AMO, and PDO — Part 2 presented the Atlantic Multidecadal Oscillation (AMO).
WHAT THE JISAO PDO WEBPAGE SAYS
Someone new to discussions of climate and weather who is looking for information about the Pacific Decadal Oscillation (PDO) would find a link the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) PDO webpage at or near the top of their search engine results. (JISAO “is a Cooperative Institute between the National Oceanic and Atmospheric Administration and the University of Washington…”) The JISAO PDO webpage introduces the PDO as, “The ‘Pacific Decadal Oscillation’ (PDO) is a long-lived El Niño-like pattern of Pacific climate variability. While the two climate oscillations have similar spatial climate fingerprints, they have very different behavior in time.” Figure 1 is the first illustration on the JISAO PDO webpage. It is described as “Typical wintertime Sea Surface Temperature (colors), Sea Level Pressure (contours) and surface windstress (arrows) anomaly patterns during warm and cool phases of PDO.” With the Sea Level Pressure and windstress representations, the maps are busy. If you were to follow the PDO Index Monthly Values link at the top of the PDO webpage you’d discover the following description of how the PDO is calculated: “Updated standardized values for the PDO index, derived as the leading PC [Principle Component] 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.”
Figure 1 (PDO maps from JISAO webpage)
WHAT DID THAT MEAN?
In the first quote above, the “long-lived El Niño-like pattern of Pacific climate variability” does NOT mean that the North Pacific (north of 20N) has a separate El Niño-like event.
A typical El Niño event creates a PATTERN in the North Pacific where it is warmer in the east than it is in the central and western portions, and a typical La Niña event will create the opposite pattern, cooler in the east than it is toward the center and west of the North Pacific. These can be seen in the two maps of the North Pacific Sea Surface Temperature (SST) anomalies in Figure 2. The top map presents the average SST anomalies during the 11-month period of May 1997 to March 1998. It captures the development and decay of the 1997/98 El Niño event. Again, during an El Niño, the PATTERN in the North Pacific typically has warmer SST anomalies in the east and cooler SST anomalies in the central and western portions. (There are a number of interacting ocean-atmosphere processes that cause the pattern, but that discussion is beyond this scope of this post.) The opposite holds true during the typical La Niña event. This can be seen in the lower map of SST anomalies. That map presents the average SST anomalies during the 11-month period from March 1998 to January 1999, and it captures the development stage of the 1998/99/00/01 La Niña.
Let’s look at the JISAO introduction to the PDO again in a slightly different way. The “long-lived El Niño-like pattern of Pacific climate variability,” means that the pattern of sea surface temperature anomalies that is normally associated with El Niño and La Niña events lasts longer than those El Niño and La Niña events. This could mean that another variable or process is impacting the pattern or causing the pattern to persist. Ongoing research is attempting to close the “loop” between the ENSO and PDO.
Nate Mantua of JISAO provides a slightly different description of the PDO in his (1999) paper The Pacific Decadal Oscillation and Climate Forecasting for North America.” It adds an important aspect. He writes, “The SST pattern highlights the strong tendency for temperatures in the central North Pacific to be anomalously cool when SSTs along the coast of North America are unusually warm, and vice-versa (Graham 1994, Miller et al 1995, Zhang et al 1997, Mantua et al 1997).” “Strong tendency” is a great choice of words, because it implies that the PDO pattern is not the only pattern of SST anomalies that appears in the North Pacific. Written another way to reinforce the point, the North Pacific SST anomalies tend to have that pattern. The PDO pattern is also said to be the “dominant pattern”.
Another important point to keep in mind: Many times the entire Pacific Ocean is shown during presentations of the PDO. However, JISAO uses only the SST anomaly data for the North Pacific north of 20N to calculate the PDO. The PDO represents nothing more than that. That is, the PDO only represents the pattern of SST anomalies in the area shown in the two maps in Figure 2.
HOW DO RESEARCHERS DETERMINE WHICH PATTERN REPRESENTS THE PDO?
Researchers use a method of statistical analysis called empirical orthogonal function (EOF) analysis to determine the pattern that represents the PDO. Wikipedia describes EOF analysis as “a decomposition of a signal or data set in terms of orthogonal basis functions which are determined from the data. It is the same as performing a principal components analysis on the data, except that the EOF method finds both time series and spatial patterns.” Further discussions of this are well beyond the scope of this post. But I did ask someone on a recent thread at WattsUpWithThat if he could simplify the description of Principle Component Analysis (PCA) and EOF analysis for readers without science backgrounds. And while the description is complex, for those who are interested, it is worth reading. Here’s a link:
(Thanks, Tom Vonk.)
GLOBAL SST ANOMALIES ARE REMOVED FROM THE PDO
The JISAO description of the PDO data also includes the following sentence: “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.” Let’s clarify why and how they do that. The PDO was first calculated in Zhang et al (1997) ENSO-like Interdecadal Variability: 1900–93. In that paper, the PDO was identified as “NP”. Zhang et al explain why they remove the global average SST anomalies on page 8, under the heading of “Analysis for the period 1900-93.” They write, “When Parker and Folland (1991) performed conventional EOF/PC analysis on the global SST field based on the longer period of record 1900–90, their leading mode was dominated by the upward trend in global mean SST prior to the 1940s. The mathematical constraint that subsequent PCs be orthogonal to this ‘global warming mode’ seems physically unrealistic.”
To isolate the pattern of variability from the changes in global SST anomalies, Zhang et al subtracted the Global SST anomalies from the SST anomalies of every grid (5 deg latitude by 5 deg longitude) in the global SST dataset. Then they performed the EOF/PC analysis on the residuals.
PDO INDEX DATA
Figure 3 is a time-series graph of the JISOA PDO Index data. In its “raw” form, it is a noisy dataset.
In Figure 4, the PDO data has been smoothed with a 13-month running-average filter to reduce the noise.
Let’s look again at the description of data: The PDO Index is “derived as the leading PC [Principle Component] 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.” In simple words, the PDO is a statistically prepared dataset. It does not represent the Sea Surface Temperature (SST) or SST anomalies of the North Pacific, north of 20N. The differences and the importance of those differences will be discussed later in this post.
Referring to Figure 5, if we compare the PDO data to NINO3.4 SST anomalies, which are commonly used to represent the frequency and magnitude of El Niño and La Niña events, we can see that the magnitude and timing of the major short-term swings in the two datasets are similar. (Refer to An Introduction To ENSO, AMO, and PDO – Part 1 for a discussion of NINO3.4 SST anomalies and ENSO events.) This shows that the El Niño and La Niña events impact the strength of the PDO pattern. But there are differences between the two datasets. There is an additional long-term (low frequency) variation in the PDO data.
If we recall the discussion of the Atlantic Multidecadal Oscillation (AMO) (Refer to An Introduction To ENSO, AMO, and PDO — Part 2), the NOAA Earth System Research Laboratory (ESRL) presents the AMO data smoothed with a 121-month filter to highlight the low frequency variations in that dataset. So we’ll also use a 121-month filter to show the differences in the low frequency variations between the PDO and NINO3.4 SST anomalies, Figure 6. While the NINO3.4 SST anomalies do exhibit multidecadal variability, the magnitude of the variations in the PDO data is much greater. Recall, however, that the NINO3.4 SST anomalies represent exactly that, the SST anomalies of an area of the tropical Pacific, while the PDO is a statistically manufactured dataset.
A REMINDER ABOUT SST OBSERVATIONS
Always keep in mind that the source SST data can be very sparse in early parts of the instrument temperature record. ICOADS data is the source for long-term SST datasets before the satellite era. Figure 7 shows six maps of ICOADS SST observation locations in the tropical and North Pacific. It presents Januarys every ten years from 1900 to 1950. The contours were set to emphasize the reading locations not the values. This sparseness of readings should be considered when examining any early SST-based dataset.
In an effort to dispel some existing misunderstandings, let’s look at what the PDO does not represent.
THE PDO DOES NOT REPRESENT SST ANOMALIES OF THE NORTH PACIFIC
Figure 8 compares the PDO data to the SST anomalies of the North Pacific, north of 20N. The North Pacific SST anomalies have much less year-to-year and long-term variability than the statistically manufactured PDO.
If we scale the PDO data by multiplying it by a factor of 0.2, Figure 9, we can see that the year-to-year variations are not similar. Also, the linear trend of the PDO is flat while the North Pacific SST anomalies have a positive linear trend as one would expect.
THE PDO IS NOT CALCULATED SIMILARLY TO THE ATLANTIC MULTIDECADAL OSCILLATION (AMO)
As discussed in An Introduction To ENSO, AMO, and PDO — Part 2 , the NOAA Earth System Research Laboratory (ESRL) Atlantic Multidecadal Oscillation webpage describes the calculation of the AMO as, “Compute the area weighted average over the N Atlantic, basically 0 to 70N,” and “Detrend that time series.” To detrend the North Atlantic Sea Surface Temperature anomalies, the monthly values of the linear trend are subtracted from the North Atlantic SST anomalies.
If we detrend the North Pacific SST anomalies and scale the PDO, Figure 10, we can see that the detrended North Pacific SST anomalies (north of 20N) have no short-term or long-term relationship to the PDO.
DO GLOBAL TEMPERATURES RISE AND FALL IN RESPONSE TO A POSITIVE AND NEGATIVE PDO?
Yes, but it’s an inverse relationship. (The text has been updated between Figures 12 and 13.) We can illustrate this by examining the SST anomalies of the North Pacific and comparing them to Global SST anomalies. Refer to Figure 11. Keep in mind the PDO represents a pattern of SST anomalies, not the SST anomalies of the North Pacific, north of 20N. In order for the SST anomalies of the North Pacific to be contributing to the rise in Global SST anomalies, the North Pacific SST anomalies have to be rising faster than the Global SST anomalies. (Note that in Figures 11 through 13 the data has been smoothed with a 37-month filter.) Or during periods when the North Pacific SST anomalies are above the Global SST anomalies, they are adding from the Global average, and the opposite is true when the SST anomalies of the North Pacific are below the Global SST anomalies. At those times they are subtracting to the Global average.
And we can illustrate the relationship between the North Pacific and Global SST anomalies by subtracting Global SST anomalies from the North Pacific SST anomalies, Figure 12. I’ve identified this as the North Pacific Residual.
Comparing the North Pacific Residual to the PDO, Figure 13, the two datasets have no relationship with one another. This means that the contribution of the North Pacific (north of 20N) to Global SST anomalies is independent of the PDO.
UPDATE (September 14, 2010): It was recently pointed out to me that the two curves in Figure 13 appear to be negatively correlated. In other words, while one curve rises, the other falls, and vice versa. I confirmed this is true, so the two curves are related. I discussed this in the follow-up post An Inverse Relationship Between The PDO And North Pacific SST Anomaly Residuals.
FOR THE PDO TO BE POSITIVE, MUST THE SST ANOMALIES BE WARM IN THE EASTERN NORTH PACIFIC?
No. During a positive PDO, the North Pacific SST anomalies are warmer (comparatively) in the east than they are in the central and western portions, and vice versa, but that does not mean North Pacific SST anomalies are warm or cool as discussed earlier. Just in case you’re not convinced, Figure 14 shows the PDO data from November 1981 to June 2010. The data has been smoothed with an 11-month filter to reduce the impact of seasonal variations in the SST anomaly maps that follow (Figure 15 and 16). And I used an 11-month filter so that I could “center” the maps on an individual month. I also noted the positive and negative PDO peaks on Figure 14.
Figure 15 shows the SST anomaly maps for the North Pacific that correspond to the four positive PDO peaks shown above in Figure 14. The lower right-hand corner map shows what some might consider the typical PDO pattern: warm in the east and cool in the center and west. Note, however, that the upper right-hand corner map has the highest peak PDO (Figure 14), but of the four maps, it shows the lowest SST anomalies in the east, and the coolest SST anomalies in the center and west. So, to reinforce earlier observations, a positive PDO shows the SST anomalies are warmer in the east than in the center and west, not that it’s warm in the North Pacific.
And we can see similar results in the SST anomaly maps in Figure 16. They show the three negative PDO peaks from Figure 14. The SST anomalies in the upper left-hand map appear warmer than the other two maps, yet that map illustrates a peak negative value of the PDO. In fact, if you were to cycle between Figure 15 (Positive PDO Peaks) and Figure 16 (Negative PDO Peaks), the basin-wide SST anomalies appear higher at the Negative PDO peaks than they do at the Positive Peaks. The exception is the map in the lower right-hand corner of Figure 15 (the map with the anomalies centered on October 1997).
There is good reason for that. The SST anomalies for the North Pacific (north of 20N) in the short term, like the long term, do not correlate with the PDO. I’ve also highlighted the positive and negative peak months of the PDO with red and blue, respectively. It makes the graph “busy”, but it does help to reinforce the point. (Let me know if it’s too confusing, to the point that it detracts from the post. If it does, I’ll delete it.)
A SIDE NOTE
For those who are wondering what North Pacific SST anomaly patterns might look like when the PDO is not strongly positive or negative, I’ve created Figure 18. The PDO curve in Figure 14 crosses “zero” a number of times between 1981 and now. The maps in Figure 18 show the SST anomaly patterns the first and last two times the PDO data (smoothed with an 11-month filter) crossed “zero.” And as in the similar Figures, the maps represent the average SST anomalies for the periods shown. As illustrated, there are few to no similarities between SST anomaly patterns shown in the four maps.
DOES THE PDO DRIVE ENSO?
There are posts and comments around the blogosphere that state something to the effect of “when the PDO is positive, El Niño events are more frequent, and the PDO is negative, there are more La Niña events.” The authors of those comments have cause and effect reversed. Keep in mind, the PDO represents the El Niño-like pattern of the SST anomalies in the North Pacific north of 20N. So during periods when the frequency and amplitude of El Niño events outweigh those of La Niña events, the positive PDO pattern (warmer in the east and cooler in the central and west) will tend to appear more frequently and the PDO will be positive. The reverse occurs when the frequency and amplitude of La Niña events outweigh those of El Niño events.
The PDO also lags ENSO, so it would be difficult for the PDO to initiate the variations in ENSO. Recall that Zhang et al refer to the PDO as “NP”. For an ENSO index, they use the Cold Tongue Index (CT) in place of NINO3.4 SST anomalies, which are used more frequently now. The Cold Tongue Index represents SST Anomalies of 6S-6N, 180-90W, where NINO3.4 SST Anomalies represent the area of 5S-5N, 170W-120W. In Figure 7 of Zhang et al, shown here as Figure 19, they illustrate the cross-correlation functions between the Cold Tongue and the other time series they examined. Note how in the bottom cell NP (PDO) lags (CT) ENSO by approximately 3 months.
Figure 19 (Zhang et al Figure 7)
Confirmation of the lag: In ENSO-Forced Variability of the Pacific Decadal Oscillation, Newman et al (2004) also found that the PDO lags ENSO. Figure 20 is cell d of Figure 1 from Newman et al. They describe it in text as, “ENSO also leads the PDO index by a few months throughout the year (Fig. 1d), most notably in winter and summer. Simultaneous correlation is lowest in November– March, consistent with Mantua et al. (1997). The lag of maximum correlation ranges from two months in summer (r ~ 0.7) to as much as five months by late winter (r ~ 0.6). During winter and spring, ENSO leads the PDO for well over a year, consistent with reemergence of prior ENSO-forced PDO anomalies. Summer PDO appears to lead ENSO the following winter, but this could be an artifact of the strong persistence of ENSO from summer to winter (r = 0.8), combined with ENSO forcing of the PDO in both summer and winter. Note also that for intervals less than 1yr the lag autocorrelation of the PDO is low when the lag autocorrelation of ENSO (not shown) is also low, through the so-called spring persistence barrier (Torrence and Webster 1998).”
Figure 20 (Figure 1 from Newman et al)
THERE ARE OTHER USES (MISUSES?) OF THE TERM PDO?
Many times bloggers, climate scientists and meteorologists will use the term Pacific Decadal Oscillation (PDO) to refer to the decadal and multidecadal SST variability in the Pacific as a whole. Unfortunately, this practice is becoming common practice. This use of PDO is very confusing to those who are new to the term, who would then check references on the internet and discover the original definition. It’s also confusing to those who understand the original definition and can lead to drawn out debates when the non-classical use of PDO is used by one of the parties.
PACIFIC DECADAL VARIABILITY?
My use of the phrase “the decadal and multidecadal SST variability in the Pacific” will raise obvious questions.
Keep in mind that El Niño and La Niña events are the second most dominant causes of year-to-year changes in global temperatures. The most dominant natural factors are explosive volcanic eruptions. They can easily offset the impacts of the strongest El Niño. Refer to the discussion of El Niño – Southern Oscillation (ENSO) in An Introduction To ENSO, AMO, and PDO – Part 1.
And when we look at a long-term graph of “raw” NINO3.4 SST anomalies (commonly used to represent the frequency and magnitude of ENSO events), Figure 21, we see a noisy dataset.
Earlier in this post we compared the PDO to NINO3.4 SST anomalies where both datasets were smoothed with 121-month filters, Figure 6. The additional magnitude of the variations in the PDO may have detracted from the variability of the NINO3.4 SST anomaly data. So in Figure 22, I’ve presented the NINO3.4 SST anomalies alone. The red line simply highlights “zero deg C”. We can see that there are decadal and multidecadal periods when El Niño events are dominant, and periods when La Niña events are dominant.
Those using PDO may also be referring to the multidecadal variations in detrended North Pacific SST anomalies. When smoothed with a 121-month filter, the detrended North Pacific SST anomalies show variability that runs in and out of phase with the AMO. I’ve also included NINO3.4 SST anomalies smoothed with the same filter so show that they too can run in and out of phase.
Many of the topics covered in this post were also presented in Misunderstandings about the PDO – REVISED and in Revisiting “Misunderstandings About The PDO – Revised”. Additionally, I examined the difference between NINO3.4 SST anomalies and the PDO in the post Is The Difference Between NINO3.4 SST Anomalies And The PDO A Function Of Sea Level Pressure? and showed that a North Pacific sea level pressure dataset appears to correlate with the difference between the PDO and NINO3.4 SST anomalies. This very simple analysis indicates that the additional natural factor that exaggerates the decadal variability of the PDO MAY BE sea level pressure.
All data used in this post is available through the KNMI Climate Explorer:
I also used the KNMI Climate Explorer to create the maps.
The PDO data from JISAO is available through the KNMI Climate Explorer “Climate Indices” webpage, but I used the data directly from the JISAO website for this post: