Climate Oscillations 8: The NPI and PDO

By Andy May

The North Pacific Index (NPI) and the Pacific Decadal Oscillation (PDO)

The North Pacific Index (NPI) is computed from the area-weighted sea level air pressure (SLP) over the region 30°N-65°N and 160°E-140°W. It measures interannual to multidecadal variations in Pacific atmospheric circulation. As explained in Trenberth and Hurrel, the winter Aleutian low pressure system moves on a decadal time scale and changes the climate and sea surface temperature (SST) along western North America and in the Northern Central Pacific. These changes are closely related to the PDO (Pacific Decadal Oscillation), which describes the same multidecadal weather and SST pattern in the same region but is calculated with SSTs using a different statistical method. Other oscillations that describe this pattern or something similar are the Interdecadal Pacific Oscillation (IPO) and the North Pacific Oscillation (NPO).

The NPI and PDO are compared in figures 1 and 2. They are different, but since they describe the same weather pattern, we will discuss them together in this post. The North Pacific SST changes tend to follow sea level air pressure changes by one to two months (Trenberth & Hurrel, 1994). Shifts in either the NPI or PDO cause major changes in the migration patterns of many fish species, as well as many weather and environmental patterns in northern Asia and in North America (Ebbesmeyer, et al., 1990) and (Lluch-Belda, et al., 1989).

As figure 1 shows, the two indices tend to track one another when the NPI SLP scale is inverted. A higher NPI area SLP is called a negative NPI, lower SLP is positive. A positive PDO Index (up on the graph) and lower average SLP over the NPI region indicates warmer SSTs off the west coast of North America and an El Niño pattern in the tropics. A negative PDO and higher NPI SLP indicates cooler temperatures on the west coast of North America and warmer SSTs in the central northern Pacific (see figure 3). Negative PDO values and higher SLP correspond to a La Niña pattern in the tropics. Figure 2 is the same as figure 1, except it only includes the winter months. Winter is when most weather events occur and it is also when the meridional transport of energy to the polar regions is maximal. Since most heat transport is via mid-latitude storms, the NPI and PDO effects in winter are more easily seen.

Figure 2 shows that the two oscillations visually correlate better in the winter months. Of the Pacific oscillations, NPI shows the best correlation to HadCRUT5, with the PDO not far behind, and both are closely related to ENSO (aka the ONI). In fact, all the Pacific Oscillations are closely related to one another. Some of them may be teleconnections, or channels, for distributing the energy flux generated by ENSO.

Figure 3 shows the NPI region and compares the positive and negative PDO to the high and low NPI SLP. The positive PDO and a positive NPI (low NPI region SLP) indicate the same SST pattern. According to Trenberth and Hurrell, the sluggish response of the large mid-latitude NPI region of the North Pacific to changes in ocean forcing effectively serves as a low-pass filter and causes this region to show long-term changes in Pacific climate.

The Pacific is the world’s largest ocean, and one would think these oscillations would have a huge influence on the HadCRUT5 global mean surface temperature (GMST), but it isn’t seen in the PDO itself. The data used to create the PDO has the global climate signal removed by subtracting the global mean SST from each point prior to the analysis (Zhang, et al., 1997) & (Hare, 1996). This is done because the leading PC (principal component) of the raw PDO is very heavily correlated (0.7) with the global mean SST (Zhang, et al., 1997). So, the correlation is there, we just don’t see it in the PDO. This removal is not done to the NPI, which is probably why it is more strongly correlated to HadCRUT5.

As we saw in the last post, the mean total Pacific SST only correlates with HadCRUT5 in the middle 20^th century. In particular, it has a poor correlation in recent decades when we have the best data. Whether this is because HadCRUT5 is flawed, or the Pacific is unrelated to recent global warming is an open question.

The Pacific Oscillations (see the list of the top 10 in table 2 here) impact global climate, especially around the Pacific Ocean, but do not correlate well with HadCRUT5. This is odd, especially considering the Atlantic Oscillations correlate better, see figure 1 here.

Characteristics of the NPI and PDO

While Trenberth and Hurrel named and described the NPI in 1994, the very similar PDO was not named until 1996 when Steven Hare suggested the name. It was later more fully described by Yuan Zhang, Nathan Mantua, Steven Hare, and colleagues in 1997 and 2002. A more detailed history of the frantic “race to describe” the well-known North Pacific weather pattern during the 1990s is given in Mantua and Hare (2002). The best early statistical descriptions of the formal PDO used today are by Zhang, et al. (1997) and Hare (1996).

A weak mirror image of the anomalies illustrated in figure 3 occurs across the South Pacific and will be discussed in a later post on the Interdecadal Pacific Oscillation or the Tripole Index (TPI). Overall, the PDO’s spatial pattern resembles that of ENSO, also discussed in a later post. The largest distinction between the PDO and ENSO are their timescales. While ENSO is primarily an interannual phenomenon, the PDO is decadal to multidecadal in scale. Thus, relatively long data records are needed to define and understand the PDO (Deser, NCAR, 2015, link).

Periodic multidecadal changes in the North Pacific climate were first recognized by studying patterns in salmon, sardine, and anchovy populations and other environmental factors along western North America and the Japanese east coast (Hare, 1996), (Lluch-Belda, et al., 1989), and (Ebbesmeyer, et al., 1990). Steven Hare recognized the connection between the environmental and fishing patterns in the North Pacific and the patterns in Pacific SSTs (sea surface temperatures) in 1996 while working on his PhD thesis and named the oscillation the “Pacific Decadal Oscillation” or PDO. We have established that the NPI is very similar climatically to the PDO, so in the rest of this post we will only refer to the PDO, but we mean both.

Figure 4 shows the ERSST v5 PDO further into the past. The PDO climate and environmental shifts that are well recognized are shown with solid lines and one that is sometimes mentioned in the literature and sometimes overlooked is shown with a dashed line. These climate and environmental shifts are most easily seen in fishing records, for example, after the 1947 shift shown in the PDO in figure 4, the salmon crop in Alaska dropped between 33% and 64%. In the same area, it increased 208% to 252% after the 1977 step change (Mantua N. J., et al., 1997).

Salmon fishing is just one Pacific environmental variable that changes with the PDO. Curtis Ebbesmeyer and his colleagues document 40 environmental variables that changed in a coordinated fashion at the 1977 climate shift (Ebbesmeyer, et al., 1990).

The best data for determining the PDO index is after 1950 and from 1950 to today the largest and most dramatic feature is the 1976-77 shift (Zhang, et al., 1997). Data from 1900 to 1950 is decent and both the 1926 and 1947 features are well established. The data prior to 1900 is not very good and that is probably why the apparent shift in 1898 is not discussed more often.

Figure 5 shows that the PDO can signal a significant change in global climate, but not always. The two exceptions are in the 19^th century and since 1997. The climate shift in 1926 signaled a period of rising temperature, the 1947 shift signaled a period of cooling, and the 1977 shift a period of warming. The shift in 1997 seems to have no global warming effect.

If one assumes that the global mean SST is mostly due to CO₂, then this effect needs to be removed, as is done in the PDO calculation. If we assume that global circulation patterns, like the PDO are contributing to global warming, then removing the global mean SST is a mistake. Since we don’t know whether either assumption is correct, we should consider both results. The global mean is not removed from the NPI, and it still shows a good correlation to the PDO. The NPI does not drop as dramatically as the PDO in recent years, see figures 1 and 2.

Discussion

The Pacific Decadal Oscillation is not directly based on trends in the average SST of a region like the AMO is, and the inputs to the calculation have the global mean SST removed, so it is less likely to reflect GMST. Trenberth and Shea recommend that the global mean SST be removed from the AMO region SSTs rather than detrend them with a least squares line, but I do not since it assumes that global trends are due to CO₂, when that may not be the case.

The PDO/NPI reflect the pattern of SSTs and wind patterns across the Northern Pacific as shown in figure 3. Which pattern exists, positive or negative, has a large impact on fishing and many other environmental factors in the Pacific region, the northwestern U.S., and off the east coast of Japan. Thus, the wind/ocean circulation patterns, along with SST, radically affect the Pacific environment.

According to Franco Biondi and colleagues’ tree-ring chronology created using trees in Southern California and Baja, the PDO extends back at least to 1661. Their chronology clearly shows the climate shifts in 1947 and 1977 and has a prominent bi-decadal oscillation (Biondi, et al., 2001). The most significant oscillations in their reconstruction were in 1750, 1905, and 1947. These are all dates of significant global cooling. If the PDO has had the global mean SST removed and it still strongly identifies global climatic events, does this mean it helped cause them? Seems logical.

It is common to hear the “consensus” say that circulation patterns don’t matter regarding global warming because all they do is move thermal energy around Earth’s surface and it is only the radiation in and out of the climate system that matters. Well, global warming is not climate, and climate is not global warming, especially when we all know that oceans limit Earth’s mean surface temperature to less than ~30°C (Sud, Walker, & Lau, 1999). Further, the climate system provides energy storage capacity that varies, it also uses some of the energy to power storms, which transport energy from one location to another. For more on Earth’s natural thermostat and “global” warming’s impact on humans see here, and the references therein.

The Northern Pacific environment appears to be largely controlled by circulation patterns and only secondarily by global average warming and cooling. Circulation patterns matter, it is possible that the Pacific significantly affects global warming and cooling and not the other way around. Both conclusions are possible and reasonable. Attempts to remove the CO₂ portion of warming by removing global average SST from indices are invalid because we cannot assume that all global warming over multidecadal periods of time is due to CO₂.

A Word about the Pacific Oscillations in general

As discussed in my last Climate Oscillations post, The Pacific Mean SST, the Pacific Climate Oscillations are an enigma. They are a very important influence on climate and the environment in the Pacific itself, the Americas, and all of eastern Asia and Australia, but they do not correlate very well with the leading surface temperature records. This is only partially by design. The leading estimates of surface mean temperature are HadCRUT5 and BEST. Figure 6 compares them.

It is clear that the HadCRUT5 (gray) and BEST (blue) reconstructions are nearly identical, yet they do not compare very well to any of the Pacific Ocean Oscillations or to the warming trend of the Pacific Ocean, which covers 33% of Earth’s surface and clearly influences weather over at least half the globe. In recent decades HadCRUT5 and BEST do compare well to the AMO, but the AMO has been in a warming trend relative to the rest of the world since the 1970s (see figure 1 here). For these reasons, it is not clear if the Pacific is telling the most accurate story about global warming or HadCRUT5/BEST are. The accuracy of both HadCRUT5 and BEST is in doubt. In any case, we will continue to discuss the remaining important climate oscillations, most of these are in the Pacific.

Download the bibliography here.