N.Atlantic N.Pacific cross-correlation

Cross-correlation of North Atlantic Ocean and North Pacific Ocean SST.

The geographic areas chosen correspond roughly to those used for the usual AMO and PDO indices, however N.Pacific area is slightly smaller avoiding the coastal US regions that anti-correlates with PDO and having excluding the most northerly region of the Bering Sea that is lacking early 20th coverage.

To avoid unintentional distortion of the frequency spectra by various re-gridding, bias-correction and other complex processing the original ICOADS data are used.

Rate of change d/dt(SST) estimated as the difference of successive data points is filtered using a low-pass Lanczos filter [1]. This filter has a reasonably sharp transition band and does not notably attenuate the pass-band, unlike the various running-averages and gaussian alternatives.

Using rate of change removes the auto-regressive AR(1) nature of the data.

A further consequence is that d/dt is also a kind of high-pass filter with a 1/f weighting that attenuates long periodicities. The combined effect being a broad, band-pass filter that retains inter-annual and decadal periodicities.

Annotations on the graph lack precision, the following peaks are extracted at higher resolution.

Peak period in years:

65.8
34.3
24.03
20.00 # poorly resolved
16.112
11.90
9.051
6.154
5.5 # poorly resolved
5.15
4.282

The circa 9 year peak is shown in detail here:
https://climategrog.wordpress.com/?attachment_id=755

[1]
The 45 month filter used here fully removes periods of less then three years, attenuation begins at around 4.6 years. The profile is given by the normalised 1 year cut-off filter presented here:
https://climategrog.wordpress.com/2013/11/28/lanczos-filter-script/

Data source
ICOADS v2.5 as archived at KNMI:

http://climexp.knmi.nl/select.cgi?id=someone@somewhere&field=coads_sst

Advertisements