For the past few months I have been working on an independent study with my advisor, Susan Lozier, researching the connection between the Atlantic Meridional Overturning Circulation (AMOC) and the Atlantic Multidecadal Oscillation (AMO). As the first part of this project, I conducted a literature review and in this post I would like to discuss the current state of the field.
But first a quick explanation of the AMO. The AMO is the average sea-surface temperature (SST) over the entire North Atlantic after the long-term warming trend is removed. In reconstructions of North Atlantic SST from the mid-1800s to present, the basin-wide mean time series appears to oscillate with a period of 60-80 years with warm periods in the late 1800s, 1920s to 1960s, and the 1990s to the present, and cold periods in between (Fig. 1). Some prefer to refer to this oscillation as “variability” because the 60-80 year period of the AMO is at the very edge of detection by direct measurements so “AMO” and “AMV” are used interchangeably in the literature.
Figure 1. The AMO from the NOAA Extended Reconstructed Sea Surface Temperature (ERSST) version 4 (1856-2016). The globally-averaged SST time series (black dashed) is subtracted from the local North Atlantic SST (area-weighted average shown in black solid line) to yield the detrended North Atlantic SST, which is then spatially averaged to derive the AMO (red thin). A 10 year low-pass filter is used to analyze low-frequency variability (red thick). The official NOAA AMO index (Trenberth and Shea, 2006; red dashed) is shown for comparison. In the top panel, the map of linear regressions between the local SST and the annual AMO demonstrate the spatial signature of the AMO. All regressions shown have p-values less than 0.05.
The AMOC has long been connected to North Atlantic SST in the paleoceanographic literature (which relies on ocean sediment cores for proxy-based SST reconstructions) because an increase in the AMOC would increase the oceanic heat transport and thus warm the SST. But within the shorter time-scale field of observational physical oceanography (which relies on observations of SST from ships, autonomous instruments and satellites among other platforms), there are multiple different SST forcing mechanisms to consider, including air-sea heat fluxes, vertical mixing and horizontal heat flux convergences/divergences. Though the AMOC may indirectly affect all of these terms, only this last term can be directly forced by variations in the AMOC. So the AMO lies in an interesting area between paleoceanography and observational physical oceanography – this middle ground is one of the many reasons why there is still ample confusion with the AMO and its forcing mechanisms.
Both the time scale (60-80 years) and the spatial signal (coherent across the entire North Atlantic, Fig. 1) is perplexing to oceanographers. The most logical forcing would be from the winds – the global wind patterns are responsible for most features in large-scale ocean circulation. But the dominant wind pattern in the North Atlantic, the westerly winds, bisect the North Atlantic around 45°N into the subtropical and subpolar gyres. Thus forcing by the westerly winds would cause opposing forcing in the two gyres, a feature that is captured by the North Atlantic Oscillation (NAO, Fig. 2) but cannot explain the basin-wide nature of the AMO. Another potential forcing mechanism is the AMOC – the AMOC redistributes heat meridionally so it could theoretically explain spatial pattern of the AMO. Zhang and Zhang (2015) posit that a southward-propagating positive AMOC anomaly leads to convergences of horizontal heat flux in the upper ocean that can explain the AMO north of 34°N. South of that latitude, the AMOC anomaly propagates quickly as a wave rather than an advective signal, so it no longer leads to heat convergences along its path. This explanation is consistent with the long-held belief of paleoceanographers that the AMOC is controlling the extra-tropical SST variability (albeit through a slightly different mechanism) but it doesn’t explain the AMO impact in the tropical North Atlantic. To explain that, a number of papers have used models and historical observations of low-level clouds (Bellomo et al., 2016; Brown et al., 2016; Yuan et al., 2016). Very simplistically, low clouds tend to cool a region because they reflect more radiation back out to space than they trap through their greenhouse effect (high clouds do the opposite – they trap more radiation than they reflect). These papers show that low frequency variability of low clouds in the tropical Atlantic varies negatively with respect to the AMO – so there are a lot of low-level clouds during cool AMO phases and few clouds during warm AMO phases, indicating that the presence of clouds could be forcing the tropical AMO signal. Taken together, these results point to concurrent AMOC anomalies in the northern portion of the study area and low level clouds in the southern portion forcing the AMO.
Figure 2. The official NOAA winter NAO index (1864-2016) and its spatial pattern. Linear regression coefficients between detrended ERSST and the annual NAO index (thin black). A 10 year low-pass filter (thick black) is applied to compare the NAO to the AMO at low frequencies. In the top panel, regressions with p-values less than 0.05 are outlined in black contours.
These explanations all assume that the AMO requires a physical forcing mechanism – an assumption that is challenged by Clement et al. . This paper demonstrates that the AMO can be recreated by slab ocean models in which the ocean is idealized as a constant (in time) depth slab of water and has no interannual variability in ocean heat transport. So in these simulations, vertical mixing and changes in horizontal heat transport cannot force the interannual variability in the AMO. Instead, the authors suggest that the AMO is simply noise with a peak in multidecadal frequencies (Fig. 3) arising from a combination of oceanic memory and random atmospheric forcing. This result has raised many questions with perhaps the most pertinent to OSNAP being how can the AMOC, with its associated heat transport variability at the RAPID line on the order of petawatts (Johns et al., 2011; 1 PW = 10^15 W = 100*global energy consumption), not influence the AMO? Results from the RAPID line, for example, have shown that upper ocean heat content in the northern subtropical gyre (26°N-42°N) decreased in response to a slowdown of the AMOC at the RAPID line in 2010 (Cunningham et al., 2013). One would assume that this also cooled the SST in this region. But does a cooling of the northern subtropical gyre lead to a basin-wide SST signal on multidecadal time scales? Does the AMOC even vary on multidecadal time scales? If the advective time scale of the upper limb of the AMOC (the time water particles take to go from the equator to the subpolar gyre) is on the order of 10-20 years, what would cause oscillations in the AMOC at 60-80 year periods? Some have posited that the strength of the Agulhas leakage from the Indian to the Atlantic Ocean may be forcing these changes (Biastoch et al. 2015), but a link this far afield may need more evidence for it to catch on.
Figure 3. Fourier spectra of the AMO (1854-2016). The peak value occurs at a period of 81 years so the time series (Fig. 1) shows exactly two cycles. The resolution of Fourier analysis becomes coarser as the period increases, so the location of the peak at long periods is difficult to determine exactly, but probably lies on the shorter end of the range between 54 years and 162 years.
The connection between the AMOC and the AMO is important to OSNAP because the AMO has been connected to Atlantic hurricane activity (e.g. Knight et al., 2006), North American and European climate (Trenberth and Shea, 2006; Sutton and Hodson, 2007), Sahel and Amazonian rainfall (Folland et al., 1986; Knight et al., 2006), Arctic sea ice extent (e.g. Zhang, 2015), and global temperature (e.g. Brown et al., 2015). If indeed the AMOC is forcing the AMO, then measurements like OSNAP become even more important and may even provide some long-sought after predictability in the climate system.
Bellomo, K. et al. (2016). New observational evidence for a positive cloud feedback that amplifies the Atlantic Multidecadal Oscillation, GRL, 43, 9852-9859.
Biastoch, A., et al. (2015) Atlantic multi-decadal oscillation covaries with Agulhas leakage. Nat. Comms., 6:10082.
Brown, P. T. et al. (2015), Regions of significant influence on unforced global mean surface air temperature variability in climate models, JGR, 120, 2, 480-494.
Brown, P. T. et al. (2016), The necessity of cloud feedback for a basin-scale Atlantic Multidecadal Oscillation, GRL, 43, 3955-3963.
Clement, A. et al. (2015). The Atlantic Multidecadal Oscillation without a role for ocean circulation, Science, 350, 6258, 320-324.
Cunningham, S. et al. (2013) Atlantic Meridional Overturning Circulation slowdown cooled the subtropical ocean, GRL, 40, 6202-6207.
Folland, C. K. et al. (1986) Sahel rainfall and worldwide sea temperatures, 1901-85, Nature, 320, 6063, 602-607.
Johns, W. E. et al. (2011) Continuous, array-based estimates of Atlantic Ocean heat transport, Journal of Climate, 24, 2429-2449.
Knight, J. R. et al. (2006) Climate impacts of the Atlantic Multidecadal Oscillation, GRL, 33, L17706.
Sutton, R. T. and Hodson, D. L. R. (2007) Climate response to basin-scale warming and cooling of the North Atlantic Ocean. Journal of Climate, 20, 891-907.
Trenberth, K. E. and Shea, D. J. (2006) Atlantic hurricanes and natural variability in 2005. GRL, 33, L12704.
Yuan, T. et al. (2016) Positive low cloud and dust feedbacks amplify tropical North Atlantic Multidecadal Oscillation, GRL, 43, 1349-1356.
Zhang, J. and Zhang, R. (2015) On the evolution of the Atlantic Meridional Overturning Circulation Fingerprint and implications for decadal predictability in the North Atlantic, GRL, 42, 5419-5426.
Zhang, R. (2015) Mechanisms of low frequency variability of summer Arctic sea ice extent, PNAS, 112, 15, 4570-4575.