I received an email yesterday morning advising me that Muller et al (2013) had been published. (Thanks, Marc.) The title of the paper is “Decadal variations in the global atmospheric land temperatures”. The abstract is here and a preprint version of the paper is available from the Berkeley Earth Surface Temperature website here. The primary finding of the paper is that land surface temperature anomalies are more closely correlated with the Atlantic Multidecadal Oscillation (AMO) than they are to NINO3.4 sea surface temperature anomalies (as a proxy for El Niños and La Niñas or ENSO) and the Pacific Decadal Oscillation (PDO) index. We’ll discuss the very obvious reasons for this.
Muller et al also briefly discussed a couple sea level pressure-based indices. They will not be discussed in this post.
We’ll discuss why Muller should have included detrended North Pacific sea surface temperatures instead of the PDO in their comparisons with the AMO data, and we’ll use correlation maps to help show what the PDO represents—and what it doesn’t represent.
ON THE CORRELATION OF LAND SURFACE TEMPERATURES WITH THE AMO
The abstract reads:
Interannual to decadal variations in Earth global temperature estimates have often been identified with El Niño Southern Oscillation (ENSO) events. However, we show that variability on time scales of 2–15 years in mean annual global land surface temperature anomalies Tavg are more closely correlated with variability in sea surface temperatures in the North Atlantic. In particular, the cross-correlation of annually averaged values of Tavg with annual values of the Atlantic Multidecadal Oscillation (AMO) index is much stronger than that of Tavg with ENSO. The pattern of fluctuations in Tavg from 1950 to 2010 reflects true climate variability and is not an artifact of station sampling. A world map of temperature correlations shows that the association with AMO is broadly distributed and unidirectional. The effect of El Niño on temperature is locally stronger, but can be of either sign, leading to less impact on the global average. We identify one strong narrow spectral peak in the AMO at period 9.1±0.4 years and p value of 1.7% (confidence level, 98.3%). Variations in the flow of the Atlantic meridional overturning circulation may be responsible for some of the 2–15 year variability observed in global land temperatures.
My Figure 1 is Figure 3 from Muller et al (2013). We’ll concentrate on it for a few moments.
The caption for their Figure 3 (from the preprint) reads:
Figure 3. Decadal fluctuations in surface land temperature estimates and in oceanic indices. The long-term variability was suppressed by removing the least-squares fit 5th order polynomial from each curve. (A) shows the 12-month smoothed land surface temperature estimates from the four groups. The decadal variations are very similar to each other. The Berkeley Earth data were derived from 2000 sites chosen randomly from a set of 30964 that did not include any of the sites from the other groups. (B) shows the AMO index compared to Tavg, the average of the four land estimates. (C) shows the ENSO index compared to the average of the four land estimates. (D) shows the AMO and ENSO directly. Note that the AMO agreement in (b) is qualitatively stronger than the ENSO agreement in (c).
When looking at their Figure 3, keep in mind that the data have been modified. Muller et al “suppressed” the “long-term variability” by subtracting the values of a 5th order polynomial curve from the data. See the example of the AMO data from the ESRL and the resulting 5th order polynomial curve here. They then smoothed the remainders with a 12-month filters for their Figure 3.
It should be very clear that El Niño and La Niña events (ENSO) are responsible for many of the year-to-year variations in the Average Land Surface Air Temperature data (Tave in cell C) and in the Atlantic Multidecadal Oscillation data (AMO in cell D). The reason: ENSO is the dominant mode of natural variability on annual and interannual timescales. Muller et al (2013) in fact note that the AMO signal lags the ENSO signal:
For reference, the maximum correlation between AMO and ENSO in these data is 0.50 ± 0.04; with AMO lagging ENSO by 0.70 ± 0.25 years. This is a somewhat larger lag than previously reported in a more detailed analysis of ENSO by Trenberth et al. .
We can see in cell B that the annual variations in the AMO and global land surface temperatures are remarkably similar. And we can see that the yearly changes in the land surface air temperatures (cell C) and the AMO (cell D) do not correlate as well with the ENSO signal.
Let’s examine the period from the mid-1980s to the early 2000s in cells C and D. The reasons for the differences show up quite well during that period. We can see that the Tave and AMO signals do not cool proportionally to the ENSO signal during the 1988/89 and 1998-01 La Niña events. Also, the Tave and AMO both respond to the eruption of Mount Pinatubo in 1991 with a multiyear dip and rebound, while the ENSO signal does not—or shows little response to the eruption.
In other words, the reason the land surface air temperatures and AMO agree so well is that, in addition to responding to ENSO, they’re also responding similarly to volcanic aerosols and to ENSO residuals, the latter of which prevent the land surface air temperatures and the AMO from cooling proportionally during the 1988/89 and 1998-01 La Niñas.
Nothing magical there. And it introduces an interesting question for other papers and blog posts?
Many papers and blog posts attempt to remove the impacts of natural variables from the global surface temperature record. When they add AMO data to the ENSO, volcanic aerosol and solar data in their multiple regression analyses, do they recognize and account for the fact that the AMO data and global land surface air temperature data have similar responses to ENSO and volcanic aerosols?
MULLER ET AL COMPARED APPLES AND ORANGES
A good portion of Muller et al (2013) deals with comparisons of global surface temperatures with the Atlantic Multidecadal Oscillation index (AMO) and with the Pacific Decadal Oscillation index (PDO), to emphasize their findings.
Muller et al (2013) failed to recognize that the AMO index data is detrended sea surface temperature anomalies of the North Atlantic, while the PDO index is NOT detrended sea surface temperature anomalies of the North Pacific. The PDO is an abstract form (a Picasso version, if you will) of the sea surface temperature anomalies of the North Pacific. So Muller et al (2013) were comparing apples to oranges.
Because I’ve discussed what the Pacific Decadal Oscillation (PDO) index is—and more importantly what it is not—in numerous posts, I’m not going to go into a detailed discussion here. If this topic is new to you, refer to the following posts (listed in order of most recent to earliest):
But in scanning the preprint version of Muller et al (2013) I was struck with an idea of how to present differences between the PDO index and the North Pacific sea surface temperature anomalies. In their Figure 6, my Figure 2, Muller et al (2013) presented correlation maps of NCDC global surface temperature data to a couple of sea surface temperature-based indices. The left-hand map presents the correlation of the AMO data with global surface temperatures and the right hand map, ENSO (NINO3.4 sea surface temperature anomalies) with global surface temperatures. High positive correlations are in dark reds and high negative correlations are in dark blues. Muller et al (2013) noted in the caption the reason they included the correlation maps.
AMO is observed to have positive or neutral correlation almost everywhere, while ENSO shows both strong positive and negative correlations.
And based on our earlier discussion, the reason the AMO has a positive or neutral correlation with global surface temperatures almost everywhere is, the AMO and global surface temperatures respond similarly to ENSO, ENSO residuals and volcanic aerosols. Simple.
Back to the idea I was talking about: The top two maps in my Figure 3 are correlation maps of global surface temperatures (NCDC data, same as Muller et al) with detrended North Atlantic sea surface temperature anomalies (the AMO) and NINO3.4 sea surface temperature anomalies (ENSO). I prepared the maps using the KNMI Climate Explorer. These correlation maps are similar to the maps presented by Muller et al. I’ve added the correlation map of global surface temperatures with detrended North Pacific (north of 20N) sea surface temperature anomalies as the lower left-hand map, and presented the PDO correlations in the lower right-hand map.
Full-sized version of Figure 3 is here.
Note that on the maps I’ve marked the locations of the sea surface areas of the respective datasets with very fine black boxes. Also note that I used ERSST.v3b sea surface temperature data for all but the PDO data. The ERSST.v3b dataset is the sea surface temperature component of the NCDC surface temperature dataset. On the other hand, Muller et al used the ESRL AMO data—it is based on Kaplan sea surface temperature data, which also includes Reynolds OI.v2 data over the last decade or so. I believe Muller et al used ERSST.v3b sea surface temperature data for their NINO3.4 data (ENSO index), but the preprint version of paper provides a link to the weekly Reynolds OI.v2-based ENSO data, which starts in 1991—and they could not have used it for the comparisons from 1950 to 2010.
If Muller et al (2013) had compared the AMO data (upper left-hand map) with a comparable dataset from the North Pacific (north of 20N) they would have used detrended North Pacific sea surface temperature anomalies (lower left-hand map), because the AMO index is detrended sea surface temperature data from the North Atlantic. Instead they used the PDO, which does not represent the sea surface temperatures of the North Pacific. Notice that there are no similarities between the two lower maps but they’re both derived from the same area of the North Pacific. The PDO index basically represents the El Niño- and La Niña-related spatial patterns in the sea surface temperature anomalies in the North Pacific—for example, warm in the east and cool in the central and western North Pacific (north of 20N) during an El Niño.
That’s why the ENSO (upper right-hand map) and PDO (lower right-hand map) correlation maps are so similar in the North Pacific. The PDO is a statistically created dataset that captures the spatial-pattern effects of El Niño and La Niña events on the North Pacific sea surface temperatures. That PDO spatial pattern is important for fishermen because it impacts where fish are located. The PDO spatial pattern is also important because it impacts rainfall patterns in the United States—we discussed that in the recent post here. But the PDO does not represent the sea surface temperatures of the North Pacific.
A couple of other notes:
The time-series data for the PDO index and the NINO3.4 sea surface temperatures are different for a very basic reason: the spatial pattern of the sea surface temperature anomalies in the North Pacific (warm in the east and cool in the central and western portions of the North Pacific during an El Niño and vice versa during a La Niña) are also impacted by the wind patterns (and interdependent sea level pressures) of the North Pacific, and those wind patterns and sea level pressures vary over time.
In the PDO map (lower right-hand map), note how the area of the central North Pacific east of Japan has the highest (though negative) correlation. That area is called the Kuroshio-Oyashio Extension or KOE. The variations in the sea surface temperatures in the KOE dominate the North Pacific data, and they are inversely related to the PDO index.
Note also in the lower left-hand map that the sea surface temperatures of the North Atlantic are correlated with the variations in the sea surface temperatures of the North Pacific. I discussed this in detail in the post The ENSO-Related Variations In Kuroshio-Oyashio Extension (KOE) SST Anomalies And Their Impact On Northern Hemisphere Temperatures.
AMO HAS LITTLE IMPACT ON U.S. TEMPERATURES?
Notice above in Figure 6 from Muller et al (2013), my Figure 2, that the surface air temperatures in the United States correlate poorly with the AMO data. They mention this in the paper:
Remarkably, neither AMO nor ENSO shows a strong correlation with the temperature in the United States, although ENSO reaches strongly up the west coast of the US.
Curiously, the correlation map for the AMO that I created at the KNMI Climate Explorer (the upper left-hand map in Figure 3), shows a moderate correlation between the AMO and U.S. surface temperatures (using the NCDC global surface temperature data). In Figure 4, I used the Berkeley Earth Surface Temperature (BEST) data in the correlation maps. The U.S. surface air temperatures based on the BEST data also correlate with the AMO data.
Full-sized version of Figure 4 is here.
WOULD USING DETRENDED NORTH PACIFIC SEA SURFACE TEMPERATURE DATA INSTEAD OF THE PDO DATA HAVE CHANGED THE RESULTS OF MULLER ET AL?
Nope. The AMO still has the strongest correlation with land surface air temperatures, because they both respond similarly to ENSO, ENSO residuals and volcanic aerosols.
But Muller et al could have saved themselves some time, since the PDO data does not in any way represent the sea surface temperatures of the North Pacific and there was, therefore, no reason to compare it to the AMO data.
I mentioned ENSO residuals a couple of times in this discussion. Those residuals are basically the aftereffects of strong El Niño events, and those aftereffects are caused by the warm water that’s left over from those strong El Niños. My illustrated essay “The Manmade Global Warming Challenge” [42MB] provides an overview of the causes and impacts of those leftover warm waters. It includes links to animations of data, which confirm the existence and source of the ENSO residuals.
And, of course, if you’re very interested in learning more about the processes of El Niño and La Niña events, there’s my book Who Turned on the Heat? The free preview is available here. Who Turned on the Heat? is available in pdf form here for US$8.00.
FURTHER INFORMATION ABOUT THE AMO
The short description: the Atlantic Multidecadal Oscillation is a mode of natural variability through which the sea surface temperatures of the North Atlantic can contribute additionally to or suppress the global warming of land surface air temperatures that are occurring in response to the warming of the rest of the global oceans. And as discussed in the “The Manmade Global Warming Challenge” [42MB], the ocean heat content data and satellite-era sea surface temperature data both indicate the oceans warmed naturally.
Refer also to the RealClimate glossary webpage about the Atlantic Multidecadal Oscillation here. There, they write:
A multidecadal (50-80 year timescale) pattern of North Atlantic ocean-atmosphere variability whose existence has been argued for based on statistical analyses of observational and proxy climate data, and coupled Atmosphere-Ocean General Circulation Model (“AOGCM”) simulations. This pattern is believed to describe some of the observed early 20th century (1920s-1930s) high-latitude Northern Hemisphere warming and some, but not all, of the high-latitude warming observed in the late 20th century. The term was introduced in a summary by Kerr (2000) of a study by Delworth and Mann (2000).
An El Niño releases heat into the atmosphere, and surface temperatures around the globe in many places warm in response to the El Niño, and in other parts, they cool—with more locations warming than cooling, so the average global surface temperature warms in response to the El Niño. But the heat released into the atmosphere during the El Niño is not directly warming the surface in those remote locations. The surface temperatures outside of the tropical Pacific warm in response to changes in atmospheric circulation caused by the El Niño. The processes that cause those changes are discussed in minute detail in Trenberth et al (2002) Evolution of El Nino–Southern Oscillation and global atmospheric surface temperatures. Wang (2005) ENSO, Atlantic Climate Variability, And The Walker And Hadley Circulation discusses why the tropical North Atlantic warms in response to an El Niño.
Compared to a number of other sea surface temperature-based indices (and sea level pressure-based indices, which we didn’t discuss in this post), Muller et al (2013) found that global land surface temperatures correlate best with the Atlantic Multidecadal Oscillation. We illustrated and discussed the reason for this—the AMO data and land surface air temperatures respond similarly to ENSO, ENSO residuals, and volcanic aerosols.
We also discussed and illustrated why Muller et al (2013) should have used detrended North Pacific sea surface temperatures instead the PDO data for a proper comparison to the AMO.
And we used correlation maps to show the differences between the PDO and the sea surface temperature anomalies of the North Pacific. We also used the correlation maps of the PDO and ENSO with global temperature anomalies to help explain what the PDO represents.