PDO — ENSO, Aleutian Low, or some of each?
Guest Post by Basil Copeland
In a recent post here at WUWT, as well as on his own blog, Bob Tisdale challenged the popular view that the Pacific Decadal Oscillation (PDO) is a principal driver of climate independent of the tropical El Nino-Southern Oscillation (ENSO) system. Presenting the results of his own analysis, as well as citations from published scientific literature, Bob at least made the case for some plausible relationship between PDO and ENSO. But what is the nature of this relationship, and does it reduce PDO to, as Bob put it, merely “a pattern of SST variability, not SST” itself?
Bob quoted extensively from Newman et al. (2003). While this paper was in press at the time, it was reviewed and considered by Miller et al. (2004). The latter is a comprehensive review of “Decadal-Scale Climate and Ecosystem Interactions in the North Pacific Ocean” (e.g., the title of the paper). While acknowledging a tropical forcing (ENSO) on North Pacific SSTs, they contend that there is another, distinct forcing that is independent of the tropical forcing, related on decadal time scales to the behavior of the Aleutian Low.
In this post I will be citing selected portions of the Miller et al. paper, and presenting some simple empirical data that demonstrate the complex reality of the PDO. While on an interannual to decadal scale it is possible to demonstrate a “teleconnection” between ENSO and the PDO, on multi-decadal time scales this teleconnection appears to have little power to explain the behavior of the PDO. Instead, the multi-decadal pattern of the PDO is based — or “forced” — on midlatitude climate variability reflected in the behavior of the Aleutian Low. Specifically, I will bring into the discussion the North Pacific Index (NPI). The NPI is calculated as the area-weighted sea level pressure over the region 30°N-65°N, 160°E-140°W, and provides a measure of the intensity of the wintertime Aleutian Low. I’ve taken the monthly NPI data, created a moving average, and have normalized and inverted it so that it scales similarly to ENSO and the PDO for purposes of comparison. For ENSO, I am using a series for Nino 3.4. All data sources are cited in a reference list at the end of this post.
Power Spectrum: Nino 3.1, PDO, and NPI
In leading up to a discussion of Newman et al., Miller et al. say:
Deterministic forcing from the tropics clearly has an effect in establishing decadal SST variability in the midlatitudes. The forcing of the canonical SST pattern (around the subtropic front) has long been linked to atmospheric teleconnections from the El Nino/Southern Oscillation (ENSO) events on interannual timescales. (Except as noted, citations will be eliminated in quoting from Miller et al.)
So this much is consistent with the point Bob was making about ENSO being a driver of the PDO, and is what we see in Figure 1 with the common periodicities at 4.8 to 6.5 years, and possibly at ~9 years. But Miller et al. go on to discuss a second pattern of SST’s present in the PDO that have their origin in the Aleutian Low and propagate via “Rossby waves from the central North to the region around the subartic frontal zone and the Kuroshio-Oyashio Extension (KOE) region.” In further describing these Rossby waves, they write:
These Rossby waves arrive several years after the Aleutian Low changes, resulting in a lagged response in SST in the KOE region. This second SST pattern…also projects onto the PDO although it has a lagged relationship to the canonical pattern of SST [i.e. the pattern induced by ENSO]. Hence, the PDO index should be considered an amalgam of these two physical ocean responses. (Emphasis supplied.)
Continuing the first quote above, Miller et al. go on to say:
The forcing of the KOE SST pattern (around the subartic front) is only weakly linked to tropical teleconnections. The independent behavior of the KOE SST has been clarified through its enhanced decadal variance relative to the canonical SST pattern and through its links to decadal wind-stress curl forcing.
Then they comment on Newman et al.:
Recent studies with a simple first-order Markov model with forcing specified by the tropical SST index, damping rate specified by SST persistence (with re-emergence) and white noise forcing (simulating midlatitude weather) reversals that the bulk of the PDO index is explicable by atmospheric forcing from tropical teleconnections (Newman et al., 2003).
Again, this accords with what Bob was saying. But Miller et al. go on to say:
The forcing with tropical origins clearly drives the canonical SST pattern portion of the PDO. However, the simple model result is somewhat deficient in decadal timescale energy. This suggests that the KOE SST pattern portion of the PDO is not simply driven by (or at least is not in phase with) this tropical forcing.
We see this clearly in Figure 1. Out beyond 12.8 years, there is very little power in the Nino 3.4 spectrum. There is modest power on a scale of 31.2 years, but there is simply nothing to compare to the multi-decadal power spectrum seen in the PDO and evident in the NPI. The latter show power at bidecadal, and especially at pentadecadal periodicities. On the latter, I would here just reference a series of papers by Minobe (sometimes with others); see the reference list at the end of this post for examples.
The bottom line? While ENSO forced variation may drive the PDO out to decadal time scales, it does not fully account for PDO variation on longer time scales. That latter is more likely related to long term, low frequency, oscillations in the behavior of the Aleutian Low, which set up what is really the more uniquely characteristic pattern of the PDO: its variation between cool and warm phases on a scale of 20-30 years.
What Difference Does It Make?
Based on discussions we had in comments to his post, and some private email correspondence, I do not think that Bob would have any significant disagreement with any of this. So what was his point, and why do I think it is important to clarify the matter? Bob’s point was that “global warming” cannot be attributed to the PDO. And depending on what one means by “global warming,” he is no doubt correct. That is, if by “global warming” we are speaking of the long rise in global temperatures since the earth came out of the “Little Ice Age,” then certainly, the PDO cannot likely account for that. After all, the PDO is basically an oscillation about the long term global trend, whatever that is. But if by “global warming” we were to mean calling attention to the increase in the rate of increase in global temperatures since the middle of the 20th century — and this was a cornerstone focus of the IPCC’s AR4 — then the behavior of the PDO does become extremely relevant to the so-called “global warming” of the past half century or so.
To understand my point, consider Figures 2 and 3. In Figure 2 I show smoothed trends for the NPI, PDO and Nino 3.4. These were derived using Hodrick-Prescott filtering. Since first introducing this technique to WUWT users last year, I’ve continued to investigate the properties of this technique. As some readers may recall, the outcome all depends upon the value of the “smoothing parameter” lambda. By analyzing Morelet wavelet transforms before and after smoothing, I can determine the degree of smoothing implicit in a given value of lambda. For Figure 2 I used a value of lambda (512000) which results in a degree of smoothing that filters out cycles of less than ~2^6 months, i.e ~5.3 years. It thus captures the decadal (and longer) variation in these three series, while filtering out shorter term periodicities (which are much more prevalent in ENSO than in the PDO/NPI). We can clearly see the decadal variation in ENSO, as well as the multi-decadal variation in the PDO, which is closely matched by the multi-decadal pattern of the NPI, in Figure 2.
I then extracted the first principal component from these three series, shown in Figure 3. Consider this a “weighted average” of the three series. It really does not reveal anything we do not already know – that the PDO/NPI, and a decadal influence from ENSO – has undergone major epoch changes, or regime shifts, in the 20th Century. Figure 3 simply illustrates what is cited on the JISAO web page for the PDO:
Several independent studies find evidence for just two full PDO cycles in the past century: “cool” PDO regimes prevailed from 1890-1924 and again from 1947-1976, while “warm” PDO regimes dominated from 1925-1946 and from 1977 through (at least) the mid-1990’s. Shoshiro Minobe has shown that 20th century PDO fluctuations were most energetic in two general periodicities, one from 15-to-25 years, and the other from 50-to-70 years.
The issue with respect to what impact this has on measuring the degree of “global warming” is obvious. If we simply bifurcate the 20th Century into two halves, which is more or less what the IPCC AR4 did in its assessment of 20th Century global temperature trends, the trend for the second half of the 20th Century is going to begin during a period that was dominated by a cool phase of the PDO, and end during a period dominated by a warm phase. In other words, even if the PDO is not itself, over a complete cycle, contributing to “global warming,” trends calculated this way will be biased upwards because of the transient impact of decadal and bidecadal influences of the PDO on global temperature. Measuring global warming with a starting point in the middle of the 20th Century, which is what the IPCC did in AR4, is just a variation on the theme of cherry picking, something familiar to most WUWT readers.
Perhaps a better way to look at all of this is from the global perspective of Figure 4. Figure 4 plots the HadCRUT3 series from 1850:01 through 2009:03. Shown is a smoothed trend line, blue (HP smoothing, lambda 512000), and superimposed is a smoothed trend line in red (again, HP smoothing with lambda 512000) of the UAH satellite record, rescaled to track with HadCRUT3. In this data, there appear to be two complete PDO-like cycles, when measured peak-to-peak: one from 1879 to 1941, and the other from 1942 to 2004. From 1879 to 1941, the decadal rate of growth in global temperature was 0.036°C/decade; from 1942 to 2004, the decadal rate of growth in global temperature was 0.087°C/decade. Contrast this latter with the claim in IPCC AR4 that the rate of warming for the past 50 years (1956-2005) was 0.13°C/decade.
Put differently, the real rate of warming since the last peak of the PDO is approximately 33% less than alleged by the IPCC because of the bias created in not taking into effect the suppression of global temperatures in the mid-20th by the cool phase of the PDO. So while the PDO itself may not be contributing to global warming, the warm phase of the past three decades, following on two to three decades of a previous cool phase, has allowed climate alarmists to overstate the actual degree of global warming in the 20th Century.
References and Data
- The NPI data I utilized is here: http://www.cgd.ucar.edu/cas/jhurrell/indices.data.html#npmon
- The main page for the PDO data is here: http://jisao.washington.edu/pdo/PDO.latest
I downloaded the data, though, from Wood For Trees: http://www.woodfortrees.org/data/jisao-pdo
- The Nino 3.4 series I used is here: http://climexp.knmi.nl/data/ihadisst1_nino3.4a.dat
- A full cite to the Miller et al., 2004 paper: Miller, AJ, Chai, F, Chiba, S, Joisan, JR and Neilson, DJ. Decadal-Scale Climate and Ecosystem Interactions in the North Pacific Ocean. Journal of Oceanography, Vol. 60, pp. 163 to 188, 2004. A copy can be downloaded here: http://horizon.ucsd.edu/miller/download/jgofs/JO_60-1-11.pdf . I believe many WUWT readers will enjoy this paper. The authors at least acknowledge the role of solar forcing, and cite Landscheidt, Svensmark and Friis-Christensen, and others who are often ignored in articles of this nature.
- The principle paper by Shoshiro Minobe on the pentadecadal cycle in the PDO is “Resonance in bidecadal and pentadecadal climate oscillations over the North Pacific : Role in climatic regime shifts,” Geophysical Research Letters. 26(7), 1999, 855-858. A copy can be downloaded here: http://eprints.lib.hokudai.ac.jp/dspace/bitstream/2115/21813/1/1999grl_letter.pdf . There is an intriguing indication that Minobe is working on research that would connect the Pentadecadal oscillation to LOD here: http://wwwoa.ees.hokudai.ac.jp/~mikeda/proj/iarc/sympo/8minobe.htm .
- The HadCRUT3 data was downloaded via Wood For Trees: http://www.woodfortrees.org/data/hadcrut3gl .