Has Global Warming Accelerated?

I’ve moved to WordPress.  This post can now be found at Has Global Warming Accelerated?

##############
CORRECTION: In agreement with my post The Atlantic Multidecadal Oscillation - Correcting My Mistake, I edited the sentence in this post that read, “North Atlantic SST anomalies vary in a semi-periodic cycle that is normally expressed as a residual, the Atlantic Multidecadal Oscillation (AMO).” The AMO is not a residual. It is detrended North Atlantic SST anomalies, according to the NOAA ESRL. The sentence now reads, “North Atlantic SST anomalies vary in a semi-periodic cycle that is normally illustrated as the Atlantic Multidecadal Oscillation (AMO). The AMO is created by detrending the North Atlantic SST anomalies.”
##########

In The Impact of the North Atlantic and Volcanic Aerosols on Short-Term Global SST Trends, I illustrated the significant effects of two natural variables, the Atlantic Multidecadal Oscillation and volcanic aerosols, on the global SST anomaly trends for the OI.v2 SST dataset. The conclusion of that post was that natural variables caused rate of rise in SST anomalies to appear to be accelerating in recent times, when, in fact, it has not accelerated. OI.v2 SST is a short-term dataset, lasting only from November 1981 to present. This post examines the effects of those natural variables on a longer-term SST anomaly dataset, ERSST.v2, Figure 1, and also extracts the immediate impacts of ENSO and solar variations.
http://s5.tinypic.com/vhrp7s.jpg
Figure 1

Most analyses of this type examine combined land and sea surface temperatures. I’ve elected to illustrate sea surface temperature (SST), which represent approximately 70% of the surface of the globe, because:
-It contains much less noise,
-Therefore, it requires less smoothing to dampen that noise, and
-The immediate effects of the North Atlantic can be removed simply.

A Note about the Smoothing: All graphs in this post represent data that has been smoothed with a 37-month running-average filter, but the adjustments have been made to the raw data, before smoothing. Scaling factors have been selected from the cited references and due their visual impacts on the smoothed data. Adjustments to raw data or to data that has been smoothed differently may use different scaling factors.

LINEAR TRENDS OF THE TWO WARMING PERIODS

It’s well known that Global SST anomalies rose during two separate periods since 1900. To show the difference, if any, between the trends for those two periods, I identified the timing of the minimums and maximums in the smoothed SST anomalies illustrated in Figure 1. I plotted them individually and had EXCEL determine the trends. Figures 2 and 3 illustrate nearly identical linear trends during the two warming periods of January 1910 to May 1943 (Linear Trend = ~0.116 Deg C/Decade) and April 1975 to July 2007 (Linear Trend = ~0.115 Deg C/Decade). As illustrated, over the two periods when global SST anomalies have risen, there is NO difference in the rates of rise. There has been NO acceleration in recent years.
http://s5.tinypic.com/119qzk6.jpg
Figure 2
############

http://s5.tinypic.com/2vuk978.jpg
Figure 3

If natural variables are eliminated, would there be an increase in recent years? No, but the linear trends for both periods do change—and they decrease significantly.

SUBTRACTING THE EFFECTS OF THE NORTH ATLANTIC

North Atlantic SST anomalies vary in a semi-periodic cycle that is normally illustrated as the Atlantic Multidecadal Oscillation (AMO). The AMO is created by detrending the North Atlantic SST anomalies. The simplest way to account for its immediate effect on global SST anomalies is to remove the North Atlantic data from the global data. If we assume the Atlantic Ocean surface area is approximately 30% of the global ocean surface area, and assume the North Atlantic represented 50% of the Atlantic, the North Atlantic SST anomaly data can be scaled by a factor of 0.15 and subtracted from the Global SST data.

Figure 4 compares Global SST anomalies and Global SST anomalies with the North Atlantic data removed. While it has been argued at blogs that the AMO is a cycle and should, therefore, have no impact on long-term trends, it is clear that the timing of the cycles raised global SST anomalies in recent decades and lowered them in the early 1900s. The result would be an increase in trend over the 20th Century. In fact, over the term of the data, as illustrated, the variations in the North Atlantic add approximately 0.007 deg C/decade to the global SST anomaly trend.
http://s5.tinypic.com/142h4at.jpg
Figure 4

SUBTRACTING THE EFFECTS OF VOLCANIC AEROSOLS

Estimates of the drop in global temperature for the first year following the Mount Pinatubo eruption vary from 0.2 to 0.5 deg C. Using the Sato Index of Stratospheric Aerosol Optical Thickness to determine the impact of volcanic eruptions on global SST anomalies, the Mount Pinatubo eruption appears to have lowered global SST anomalies approximately 0.15 deg C during the initial year. This coincides with the peak mean optical thickness value (~0.15) of the Sato Index for that eruption, so I have not scaled the Sato Index data. To remove the impacts of volcanic aerosols from the adjusted global SST anomaly dataset, the Sato Index data was simply added (without scaling). The result is illustrated in Figure 5. Note how there is no visible effect on the adjusted global SST anomaly trend, but there were short-term periodic effects on the SST record.
http://s5.tinypic.com/2ur5zdd.jpg
Figure 5

SUBTRACTING THE IMMEDIATE EFFECTS OF ENSO

In “Evolution of El Nino-Southern Oscillation and global atmospheric surface temperatures” (2002), Trenberth et al identified a regression coefficient for global temperature of 0.094 deg C for a 1 deg C change in NINO3.4 SST anomaly. I’ve used a scaling factor (0.096) in this post that is not significantly different that the Trenberth et al coefficient. The cited Trenberth et al paper can be found here:
http://www.cgd.ucar.edu/cas/papers/2000JD000298.pdf
I also shifted the NINO3.4 data by three months to account for the lag between an ENSO event and the response in global temperature.

Figure 6 illustrates the impact of removing the scaled NINO3.4 SST anomaly data from the global SST anomalies that have already been adjusted for North Atlantic variability and volcanic aerosols. It clearly shows that the three peaks in recent years are ENSO related, as was the deep minimum around 1910. By adjusting these, the long-term global SST anomaly trend decreases slightly.
http://s5.tinypic.com/ic5pw6.jpg
Figure 6

Also note how the peak in the early 1940s has shifted and become exaggerated. This appears to reflect the anomalous response of the Indian Ocean to the multiyear El Nino at that time. Refer to Figure 7, which is a long-term graph of Indian Ocean SST anomalies.
http://s5.tinypic.com/6zn8nr.jpg
Figure 7

SUBTRACTING THE IMMEDIATE EFFECTS OF SOLAR IRRADIANCE

There are no long-term monthly Total Solar Irradiance (TSI) datasets, so I’ve used monthly sunspot numbers as a proxy for TSI.

The sunspot data was first scaled to reflect the current consensus that, for the past three solar cycles, global temperatures changed approximately 0.1 deg C as the solar cycles varied from minimum to maximum. Then I subtracted the scaled sunspot data from the global SST anomaly data that’s been adjusted for the effects of the North Atlantic, volcanic aerosols and ENSO. Refer to Figure 8. In addition to a minor reduction in trend, removing the impacts of solar irradiance produces noticeable changes in curve. It minimizes the peaks in the data during recent decades, but adds to the troughs in the late 1940s through the 1950s. Then again, it was recently discovered that the SST data during the mid-1940s contained and still contains a large discontinuity. The discontinuity was due to the approach researchers used to account for changes in sampling methods. So I’m not necessarily concerned about any overreaction during the 40s and 50s.
http://s5.tinypic.com/6yl0ki.jpg
Figure 8

A BEFORE AND AFTER COMPARISON

Figure 9 compares the unadjusted Global SST anomaly data with the dataset that’s been adjusted for the North Atlantic variability, volcanic aerosols, ENSO, and solar contributions. The long-term trend drops considerably when the natural variables have been removed, from ~0.041 deg C/decade to ~0.028 deg C/decade, or about 32% [1-(0.028/0.041)]. Note also how the SST anomalies in the mid-1940s did not change but SST anomalies at all other times did. The previously discussed discontinuity occurred in 1945/46. Curious.
http://s5.tinypic.com/6h6r20.jpg
Figure 9

THE RESULTING CURVE

Figure 10 illustrates the Global SST anomaly curve with all the previously discussed adjustments. If you, the reader, were to follow the recipe I’ve provided in the preceding, you could create your own spreadsheet. If you do, allow yourself the ability to adjust all of the variables so you can see how the curve can be altered and how the adjustments interrelate. I could illustrate that, but I wouldn’t have any bases for those adjustments.
http://s5.tinypic.com/11uuo8z.jpg
Figure 10

A few things struck me about Figure 10. How do we interpret the curve? Does it still represent, over the 20th century, a warming period, followed by a period of cooling, then another warming period? Or is it a long continuous warming with an anomalous rise and fall in the 1930s and 40s? And the last question, how would GCMs duplicate it. For example, without the volcanic aerosols forcing, the GISS Model E output, Figure 11, simply looks like an exponential curve.
http://s5.tinypic.com/2jgwig.jpg
Figure 11

The GISS Model E was discussed further in my two posts:
GISS Model E Climate Simulations, and
GISS Model E Climate Simulations Part 2
The posts also contain links and the instructions for downloading the Model E output data.

SHORT-TERM TRENDS

As noted at the beginning of this post, removing the effects of natural variables on global SST anomalies reduces the trends during the two warming periods. Referring back to Figure 10, there’s now a peak instead of a trough at 1975. Therefore, using 1975 as a starting point for the recent period of warming would result in a lower trend. I wouldn’t want to be accused of cherry-picking dates. So I’ll use the minimum at October 1957 as a start time. I also shifted the beginning and end months of the earlier period to accommodate the shifts in peaks and valleys. As shown in Figures 12 and 13, the trends for the two periods of warming are again nearly identical, so based on the adjusted SST data, there has been no acceleration in recent years.
http://s5.tinypic.com/2rntow6.jpg
Figure 12
############
http://s5.tinypic.com/rw13jk.jpg
Figure 13

Recall, also, that the trends for the warming periods in the unadjusted data were ~0.115 deg C/decade. Refer to Figures 2 and 3. After removing the natural variables, the trends of the two warming periods plummeted to ~0.065 deg C/decade. That’s a decrease of approximately 43% [1-(0.065/0.115)]. Trends of 0.065 deg C/decade are far short of the 0.2 deg C/decade rise projected by the IPCC for the next two decades.

SOURCES

Monthly Sunspot data from January 1749 to October 2008 is available through NASA’s Marshall Space Flight Center.http://solarscience.msfc.nasa.gov/greenwch/spot_num.txt

Sato Index data is available from GISS:
http://data.giss.nasa.gov/modelforce/strataer/tau_line.txt

The ERSST.v2 data are available at the KNMI Climate Explorer website.
http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere