>Keep in mind when reading this post that the NINO3.4 temperatures shown in many of the graphs have been scaled drastically; the changes in NINO3.4 temperature are, in fact, more than 10 times greater than illustrated. Also keep in mind that the changes in NINO3.4 temperature precede the changes in global temperature anomaly; global temperature is responding to the El Nino or La Nina, not vice versa.
Figure 1 is a typical illustration of global temperature anomaly from 1850 to 2007. I’ve used Hadley Centre global temperature data (HadCRUT3GL) for this post.
Annual changes in anomaly, Figure 2, are calculated by subtracting the prior year anomaly value from the current year value, then repeating the calculation over the range of 1850 to 2007. Temperatures are rising whenever the “Annual Change” values are above zero, and dropping when they are below zero.
Figure 3 illustrates the magnitude of these annual variations without the visual skewing of the anomaly data. For the most part, these annual changes are driven by El Nino and La Nina events, large and small. Doubt that? Read on.
The El Nino/Southern Oscillation (ENSO) data used are NINO3.4 from NCDC, available here:
Note that the NINO3.4 data begins in 1871, so the global temperature data prior to that year will be excluded.
NINO3.4 SST and anomaly curves are shown in Figures 4 and 5, where the base period for the anomaly data is 1950 to 1979, the same base period used by Trenberth and Stepaniak in “Indices of El Niño Evolution”, J. Climate, 14, 1697-1701. Refer to: http://www.cgd.ucar.edu/cas/catalog/climind/TNI_N34/index.html#Sec5
Working backwards in time, the significant 1997/98, 1982/83, 1939/40/41/42, and 1878/79 El Nino events clearly stand out. The 1939 to 42 El Nino appears to be the source of that bump in the global temperature anomaly, Figure 1, that’s centered around 1940. It’s then followed by the two major La Nina events in 1950/51 and 1954/55/56/57, which appear to have created or enhanced the global temperature dip in the 1950s.
To put the annual changes in global temperature and NINO3.4 into perspective, Figure 6 illustrates the raw data. Please click on the TinyPic links for the full-sized graphs. The correlation between the annual variations in NINO3.4 area SST and global temperature is obvious. Changes in NINO3.4 temperature are in many cases more than 10 times larger than the changes in global temperature anomaly.
In “Evolution of El Nino–Southern Oscillation and Global Atmospheric Surface Temperatures”, JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. D8, 10.1029/2000JD000298, 2002, Trenberth et al identified the global temperature reaction to the 97/98 El Nino: “The regression coefficient based on the detrended relationship is 0.094 deg C per N3.4 and is deemed more appropriate. The N3.4 contribution is given in Figure 3. It shows that for the 1997–1998 El Nino, where N3.4 peaked at ~2.5 deg C, the global mean temperature was elevated as much as 0.24 deg C (Figure 2), although, averaged over the year centered on March 1998, the value drops to ~0.17 deg C.” Note: The figures referenced within the quotation are the figures in the Trenberth study, not this post.
I’m using annual data for this comparison, and the NINO3.4 data sources are different. The following will be used to verify the scaling factor: The Trenberth et al annual average value of 0.17 deg C global temperature and the annual NINO3.4 anomaly value centered on December 1997 (assumes a 3-month lag between NINO3.4 and global temperature) from the calculated NINO3.4 anomaly data illustrated in Figure 6, which is 1.823 deg C. The calculated annual scaling factor is 0.093 (0.17/1.823). The difference from the Trenberth et al regression coefficient is insignificant.
Figure 7 illustrates the correlation of annual changes in global temperature and NINO3.4 data, where the NINO3.4 data has been adjusted by the scaling factor of 0.093. Again, please use the TinyPic link to view the full-sized graph. And again, keep in mind that changes in NINO3.4 temperature precede global temperature and that the NINO3.4 data in Figure 7 has been reduced in scale by a factor of more than 10. It’s easy to see that ANNUAL changes in NINO3.4 temperature drive ANNUAL global temperature variations.
Why was the NINO3.4 data scaled with the value of 0.093 in Figure 7, when a larger value would have provided a better comparison? The scaling factor of 0.093 has a role in the long-term comparison.
Recall: To create the graphs of the annual changes in global temperature that were used in Figures 2, 3, 6, and 7, the anomaly from the prior year was subtracted from the anomaly of the year being calculated, and the process was repeated over the term of the data. To return the data to its original long-term state, the annual values would need to be added to each other, from year to year, using a running total. The data will shift because the base period for the anomaly data was eliminated in the first calculations, but the shape and magnitude of the curve will remain the same.
This also allows a side-by-side comparison of the long-term effects of the scaled NINO3.4 on global temperature. A running total can also be applied to that data. Refer to Figure 8.
The correlation between the running totals infers that ENSO drives long-term change in global temperature, in addition to the annual variations. From the early 20th Century on, temperatures rise in parallel until the 1950s, when they dip until the mid-1970s, then rise again, all a function of El Nino/Southern Oscillation. Using the NINO3.4 data, there was no need to magically apply a non-existent tropospheric aerosol forcing in the 1950s through 1970s in order to slow the warming.
BUT WHAT DRIVES ENSO?
I created the following graph, Figure 9, to show that the magnitude and frequency of ENSO events rose with the increase in the annual TSI changes. These annual changes in TSI were calculated in the same way used to calculate the annual changes in temperature; that is, the prior year TSI value was subtracted from the year being calculated, with the process repeating each year for the term of the data. With a few exceptions, the graph does infer that ENSO event frequency and magnitude do increase with rising variations in solar irradiance.
Then I noticed that a few of the data points appeared to correlate but were offset by a few years, so I shifted the TSI data four years. Refer to Figure 10. It looks nice, but it still only illustrates an implied relationship between the amplitude of TSI and the frequency and magnitude of ENSO events.
NINO3.4 and global temperature anomaly data correlate annually and over a long term. This provides a stronger case for cause and effect than if they had correlated in only one way.