The Atlantic Multidecadal Oscillation - Correcting My Mistake

I’ve moved to WordPress.  This post can now be found at The Atlantic Multidecadal Oscillation – Correcting My Mistake

################
Figure 1 illustrates the NOAA Earth System Research Laboratory (ESRL) Atlantic Multidecadal Oscillation (AMO) dataset from 1856 to present, smoothed with a 37-month running-average filter. The ESRL AMO webpage is here:
http://www.cdc.noaa.gov/data/timeseries/AMO/
The unsmoothed AMO data is here:
http://www.cdc.noaa.gov/data/correlation/amon.us.long.data

http://i40.tinypic.com/2uqlw28.jpg
Figure 1

I’ve recently been providing monthly AMO updates (Mid-April 2009 NINO3.4 SST Anomaly and AMO Update and March 2009 SST Anomaly Update), after noting A Recent Drop in the AMO in yet another post. Those posts presented a short-term (November 1981 to present) view of the data, so the long-term oscillation was not visible.

I did, however, post a long-term “AMO” graph recently in my post Individual Ocean SST Anomalies In Perspective. Refer to Figure 2. Note the major differences between that curve and the ESRL AMO curve in Figure 1. The data in Figure 2 is missing the rise in the late 1800s, and the timing of the decrease then increase in the mid to late 20th century is incorrect.
http://i41.tinypic.com/2n055lk.jpg
Figure 2

MY METHOD OF CALCULATING THE AMO WAS WRONG

My error: I had calculated the AMO as the difference between the North Atlantic SST anomalies and Global SST anomalies (North Atlantic SST anomalies MINUS Global SST anomalies).

THE NOAA ESRL AMO DATA IS SIMPLY DETRENDED NORTH ATLANTIC SST ANOMALY DATA

The ESRL describes the method for creating their AMO data as follows:

“Use the Kaplan SST dataset (5x5).
“Compute the area weighted average over the N Atlantic, basically 0 to 70N.
“Detrend that time series”

Figure 3 illustrates raw North Atlantic SST anomaly data and the corresponding linear trend. As noted above, ESRL uses Kaplan SST anomaly data for the AMO, so I’ve used it in Figure 3. To detrend the North Atlantic SST anomaly data and create the AMO, the monthly values of the trend are subtracted from the monthly SST anomaly data. (I have no idea how I mistakenly got it stuck in my head that the AMO was a residual.)
http://i44.tinypic.com/10funtv.jpg
Figure 3

In Figure 4, I’ve detrended the North Atlantic SST anomaly data, using the simple method described above. I’ve also smoothed it with a 37-month filter. The curve now reproduces the cycles of the AMO.
http://i39.tinypic.com/2crm2s5.jpg
Figure 4

To confirm that my method was correct, I created a comparative graph of the ESRL AMO data and the detrended Kaplan North Atlantic SST anomalies, Figure 5. Note that the Kaplan SST data through the KNMI Climate Explorer has not been updated since March 2003, but that’s fine for this examination.

There are some minor differences that likely result from differences in the coordinates, climatologies, and base years. KNMI and ESRL may also infill missing data differently. Considering all those possible differences, the Detrended Kaplan North Atlantic SST Anomalies do correlate well with the ESRL AMO data, and the curve is a much better presentation of the AMO than my erroneous residual version.
http://i43.tinypic.com/m73nko.jpg
Figure 5

ERSST.v3b VERSION OF AMO

For information purposes, in Figure 6, I’ve provided a comparison of the ESRL AMO data and Detrended ERSST.v3b North Atlantic SST Anomalies. The coordinates for the ERSST.v3b data are 0 to 70N, 78W to 10E. The ERSST.v3b data is also up-to-date, which allows for a comparison through March 2009.
http://i41.tinypic.com/25snzma.jpg
Figure 6

The short-term comparison, Figure 7, confirms that the AMO has indeed dropped significantly since June 2008. You have to go back to late 1996, before the 1997/98 El Nino, to find AMO values as low. Keep in mind, though, that the AMO is a noisy dataset and the AMO has a long way to go before it reaches its minimum. Referring back to Figure 6, it took approximately 30 years for the AMO to rise from minimum to maximum (assuming that 2004 was the maximum), and it should take about 25 more years for the AMO to reach minimum again.
http://i39.tinypic.com/mhyupw.jpg
Figure 7

ONE LAST GRAPH

There will be those who will wonder why a map of the North Atlantic SST anomalies will show predominantly positive anomalies, while the AMO is presently in negative values. Figure 8 illustrates raw ERSST.v3b North Atlantic SST anomaly data and the corresponding linear trend. Present SST anomalies are positive, but since they are below the trend line, the AMO value is negative.
http://i41.tinypic.com/2vdqzns.jpg
Figure 8

CORRECTIONS TO PAST POSTS

Over the next few weeks, I will be correcting the graphs and text in earlier posts. I will indicate that the corrections have been made with a note at the beginning of the post. For some posts, I will rewrite and update them in their entirety.

Sorry for the misinformation.

SOURCES

The Kaplan and ERSST.v3b SST anomaly data is available through the KNMI Climate Explorer webpage:
http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere