by Judith Curry
The conclusion is that the oscillatory mode (mostly due to the AMO) is significantly more important than the monotonic mode (mostly due to increasing atmospheric CO2) in explaining the 1980–2000 U.S. temperature increase. – Bruce Kurtz
The failure of global climate models to simulate regional climate variability on decadal time scales suggests that the multidecadal ocean oscillations such as the AMO and PDO might play a dominant role in determining climate variability on these scales.
This issue is addressed in an interesting new paper published in PLOS One:
The Effect of Natural Multidecadal Ocean Temperature Oscillations on Contiguous U.S. Regional Temperatures
Abstract. Atmospheric temperature time series for the nine climate regions of the contiguous U.S. are accurately reproduced by the superposition of oscillatory modes, representing the Atlantic multidecadal oscillation (AMO) and the Pacific decadal oscillation (PDO), on a monotonic mode representing, at least in part, the effect of radiant forcing due to increasing atmospheric CO2 . The relative importance of the different modes varies among the nine climate regions, grouping them into three mega-regions: Southeastern comprising the South, Southeast and Ohio Valley; Central comprising the Southwest, Upper Midwest, and Northeast; and Northwestern comprising the West, Northwest, and Northern Rockies & Plains. The defining characteristics of the megaregions are: Southeastern – dominated by the AMO, no PDO influence; Central – influenced by the AMO, no PDO influence, Northwestern – influenced by both the AMO and PDO. Temperature vs. time curves calculated by combining the separate monotonic and oscillatory modes agree well with the measured temperature time series, indicating that the 1938-1974 small decrease in contiguous U.S. temperature was caused by the superposition of the downward-trending oscillatory mode on the upward-trending monotonic mode while the 1980-2000 large increase in temperature was caused by the superposition of the upward-trending oscillatory mode on the upward-trending monotonic mode. The oscillatory mode, mostly representing the AMO, was responsible for about 72% of the entire contiguous U.S. temperature increase over that time span with the contribution varying from 86 to 42% for individual climate regions.
Published in PLOS One [link]
Below is a summary figure, one that was not included in the final manuscript (but was in an earlier version of the paper that BK emailed to me) that combines the content of 3 figs in the final paper.
Caption. Measured and calculated global temperature change. The calculated global atmospheric temperature-time curve (orange solid line) is obtained by combining the monotonic mode (black dashed line) and the oscillatory mode (orange dashed line). The centered moving average temperature (black solid line) includes fewer than thirteen years after 2007 and is increasingly unreliable (black solid line changes to dashed).
Fig 8 shows the contribution of the calculated monotonic and oscillatory modes to the 1980–2000 increase in atmospheric temperature for the individual climate regions. The oscillatory mode is responsible for about 85% of the temperature increase in the Southeastern mega-region, compared to about 67% in the Central mega-region. The Southeastern mega-region has the smallest absolute monotonic mode contribution. Again, the NR&P doesn’t fit exactly into either the W or NW climate region since the oscillatory mode in the NR&P accounts for 60% of the temperature increase compared to 42 and 52% in the W and NW, respectively.
Fig 8. Contributions of the monotonic and oscillatory modes to the 1980–2000 contiguous U.S. regional temperature increases. The stacked bars show the contributions (K) of the monotonic and oscillatory modes for each climate region. The inset numbers are the percent of the total temperature increase attributable to the oscillatory mode. The climate mega-regions are shown by the colors.
Fig 9 shows the mega-region boundaries (red dashed lines) corresponding to the approximate locations of the NCDC climate regions comprising the mega-regions. The NR&P climate region should probably be divided between the Central and Northwestern mega-regions, but the location of the boundary separating those mega-regions within the NR&P is unknown. The figure also summarizes the mega-region characteristics in terms of oscillatory and monotonic mode influence.
Fig 9. Contiguous U.S. climate mega-regions. The three mega-regions are shown relative to the approximate locations of the NCDC contiguous U.S. climate regions. Within the NR&P climate region the location of the boundary separating the Northwestern and Central mega-regions is unclear and is, accordingly, left undrawn
Summary of data sets and methodology
Surface temperatures obtained from NOAA/NCDC [link]
AMO: de-trended, un-smoothed monthly data from the ESRL Physical Sciences Division [link] These data are calculated from the Kaplan dataset.
To calculate the North Atlantic SST trend, used the ERSL “not de-trended” dataset [link]
For the PDO used Nate Mantua’s monthly data at JISAO [link]
For both the AMO and PDO monthly datasets, first converted the monthly data to yearly averages and then converted that result to a 13-year moving average.
JC comment: Bob Tisdale has remarked that NOAA’s new SST dataset [link] has resulted in reclassification of many weak El Nino years (e.g. 2014 is no longer classified as El Nino). The calculations of AMO and PDO depend not only on the underlying data set, but also on the methodology used. It would be worthwhile to assess the uncertainty in Kurtz’s results associated with uncertainties in both the regional surface temperature data sets and also calculations of AMO and PDO.
I find this paper to be of interest on two fronts:
- the regional attribution of climate change, including ‘fingerprints’ associated with the multidecadal ocean oscillations
- projections of future regional climate variability on decadal time scales
With regards to attribution, ‘fingerprint’ detection methods do not account for the fingerprints associated with the multidecadal oscillations. Kurtz conducts a straightforward and simple attribution analysis. However his methodology may be too simple for attribution – apart from uncertainties associated with the datasets, it is not straightforward to disentangle the forced response from the ocean oscillations. This issue was raised in a recent paper by Michael Mann [link]; the stadium wave team has submitted a critique to Science that is winding its way slowly through the review process).
You may have spotted this graphic from NASA, promoted by Bloomberg [link] that purports to demonstrate that the warming was caused by CO2 (and not the sun or volcanoes). Well, I have a number of problems with that diagram, but the Kurtz paper reinforces yet again that you can’t properly do late 20th century warming attribution without considering the multidecadal (and longer) ocean oscillations.
The second application for this paper was raised in my talk at last year’s UK-US Workshop on Climate Adaptation entitled Generating possibility distributions of scenarios for regional climate change. Consider the southeast U.S. (my home territory), which is one of the globe’s cool spots (i.e. it has not been warming). Nevertheless, regional planners find it important to account for the impacts of climate change in their plans out to 2050. Something like 5+ years ago, I was invited by the Atlanta Regional Commission to give a presentation on the latest climate model simulations; they felt they couldn’t act on regional water policy until the climate model results were clarified (the AR4 simulations were all over the place for the southeast U.S.). Because of the dominance of the AMO in this region, planners would be much better off accounting for the multidecadal variability of the AMO rather than AGW.
So how unusual is the U.S. in having multidecadal ocean oscillations being a dominant factor in climate change? We don’t know, because climate scientists have been mostly focusing on the anthropogenic effects.
Bio notes: I have been corresponding with Bruce Kurtz via email for the past 5 years. Bruce Kurtz has a PhD in chemical engineering. He has conducted hands-on R&D and also has held managerial positions in companies that include AlliedSignal (now Honeywell) and one of the Unilever specialty chemical companies (subsequently acquired by ICI). After retiring he became interested in climate science, particularly aspects related to heat and mass transports; hence his interest in the AMO/AMOC.