03.18.2021

A potential feedback loop underlying glacial-interglacial cycles

By Els Weinans, Anne Willem Omta, George A. K. van Voorn and Egbert H. van Nes

Abstract

The sawtooth-patterned glacial-interglacial cycles in the Earth’s atmospheric temperature are a well-known, though poorly understood phenomenon. Pinpointing the relevant mechanisms behind these cycles will not only provide insights into past climate dynamics, but also help predict possible future responses of the Earth system to changing CO22 levels. Previous work on this phenomenon suggests that the most important underlying mechanisms are interactions between marine biological production, ocean circulation, temperature and dust. So far, interaction directions (i.e., what causes what) have remained elusive. In this paper, we apply Convergent Cross-Mapping (CCM) to analyze paleoclimatic and paleoceanographic records to elucidate which mechanisms proposed in the literature play an important role in glacial-interglacial cycles, and to test the directionality of interactions. We find causal links between ocean ventilation, biological productivity, benthic δ18δ18O and dust, consistent with some but not all of the mechanisms proposed in the literature. Most importantly, we find evidence for a potential feedback loop from ocean ventilation to biological productivity to climate back to ocean ventilation. Here, we propose the hypothesis that this feedback loop of connected mechanisms could be the main driver for the glacial-interglacial cycles.

Introduction

The past 0.9 million years have been characterized by large cycles in global Earth temperature with a periodicity of about 100 ky. During warm (interglacial) periods, ice volume was small and CO2CO2 concentrations were high. During cold (glacial) time periods, ice volume was large and CO22 concentrations were on average 90 ppm lower (Ruddiman 2001). Understanding the processes involved in these cycles can help to pinpoint the relevant processes in the carbon cycle. Given the rapid change that our climate system is undergoing today, knowledge on the existing feedbacks is of particular importance.

There is an ongoing debate about which mechanisms drive the reduction of CO22 during glacials and which mechanisms drive the rapid increases in CO22 during deglaciations (Menviel et al. 2018). Interactions with the ocean’s carbon reservoir are an obvious first candidate for two main reasons:

  1. 1.The ocean is by far the largest carbon reservoir interacting with the atmosphere on the relevant timescale (Fasham 2003).
  2. 2.The δ13δ13C of the ocean-atmosphere system was isotopically lighter at the Last Glacial Maximum than during the Holocene (especially in the deep ocean) (Eggleston et al. 2016; Peterson and Lisiecki 2018). Therefore, it is unlikely that the excess carbon was stored in the terrestrial biosphere or methane hydrates, which are both isotopically light carbon reservoirs (Zeebe and Wolf-Gladrow 2001).

It has been hypothesized that the excess carbon was stored in an isolated abyssal reservoir (Broecker and Barker 2007). At the glacial-interglacial transition, carbon from this reservoir would have been released to the atmosphere through upwelling (Broecker and Barker 2007; Marchitto et al. 2007). However, it has turned out to be difficult to locate a stagnant reservoir of sufficient size to account for the excess carbon (Broecker and Barker 2007; Broecker and Clark 2010). Another proposed mechanism that has received considerable attention is that changes in ocean circulation caused the increase in CO22 (Siegenthaler and Wenk 1984) by bringing carbon from the deep to the sea surface (Anderson et al. 2009). Ocean circulation or ocean ventilation could have caused the rise of CO22 in other ways as well. For example, meltwater pulses in the North Atlantic during the last deglaciation may have led to a temporary shutdown of the Atlantic Meridional Overturning Circulation (AMOC) and an associated increase in Antarctic Bottom Water (AABW) (Broecker 1998). According to a hypothesis by Toggweiler (1999), such a shift would in turn have led to outgassing of CO22 from the ocean to the atmosphere due to the poor nutrient utilization subpolar Southern Ocean, where AABW is formed. Even so, modeling studies have been ambiguous about the impact of meltwater pulses on the ocean carbon cycle. In some simulations, the addition of meltwater in the North Atlantic led to carbon release from the ocean to the atmosphere (Schmittner et al. 2007; Bouttes et al. 2012; Matsumoto and Yokoyama 2013; Schmittner and Lund 2015), whereas it led to a net uptake of carbon by the ocean from the atmosphere in others (Obata 2007; Bozbiyik et al. 2011; Chikamoto et al. 2012). According to a set of model experiments by Menviel et al. (2014), the net effect depends strongly on the detailed salt budget.

Other potential mechanisms for CO22 fluctuations during glacial-interglacial cycles are based on biological activity. For a long time, the main hypothesis was that enhanced productivity in polar regions during glacial times drove down carbon from the atmosphere into the ocean (Sarmiento and Toggweiler 1984; Sigman and Boyle 2000). An early hypothesis by Martin (1990) states that enhanced biological productivity during glacial times was caused by iron deposition (or dust deposition). This idea is supported by experiments that show that current densities of phytoplankton (biological productivity) are limited by iron, for example within the Southern Ocean Iron RElease Experiment (SOIREE) project (Boyd and Law 2001). It has been suggested that changes in iron concentration can be amplified by local feedbacks, for example by increased ocean surface temperatures. These feedbacks could further increase the effect of iron fertilization on climate (Ridgwell 2002). However, according to the iron fertilization hypothesis, primary productivity in the subpolar Southern Ocean should be higher during glacial times, which is not supported by proxy records (Kohfeld and Chase 2011; Kohfeld et al. 20052013). Furthermore, general circulation models have shown that atmospheric CO22 did not respond as much to increased productivity as suggested by earlier box models (Archer et al. 2000). A hypothesis by Omta et al. (2013) describes another role for biological productivity: it suggests that spikes in the densities of marine calcifiers may lead to a quick reduction in sea surface alkalinity which explains the observed rapid increase in temperature and CO22 during deglaciations. Instead of extracting CO22 from the sea surface and thus reducing atmospheric CO22 and temperature, marine productivity increases atmospheric CO22 according to this explanation.

It is generally assumed that the different mechanisms are linked to each other (Crucifix et al. 2017), and many of them are not mutually exclusive. Therefore, most modeling studies use a combination of proposed mechanisms (Brovkin et al. 2007). Furthermore, many geological variables are synchronized with the Milankovitch forcing (the Summer insolation at subpolar latitudes in the Northern Hemisphere), which makes it hard to distinguish between links between the variables themselves or common links with the forcing (Daruka and Ditlevsen 2016). What makes the topic even more complicated is the cyclic behaviour that is observed, which may be linked to positive or negative feedback loops involving different elements (Lenton et al. 2008). One example is the ice-albedo effect, where higher temperatures lead to reduced ice caps which leads to a reduction in the albedo, which in turn increases the temperature. Thus, two processes reinforcing each other can give a strong positive feedback loop. The addition of more processes may weaken the feedback, particularly if different scales are involved. This suggests that a potential feedback loop underlying glacial-interglacial cycles would likely be dominated by a few key causal links.

It is still unclear what these few dominant causal links would be, although many different potential causal interactions have been described (see Table 1 and Fig. 1). Here we take a statistical approach to investigate whether existing data suggest any particularly strong causal links. The recent increase in available records that go back hundreds of thousands of years and that have improved temporal resolution and quality allows for such an approach. A suitable method for unravelling causal links from nonlinear time series is Convergent Cross-Mapping (CCM) (Sugihara et al. 2012). It has been applied in ecological systems to find causal links between temperature and anchovy and sardine abundance (Sugihara et al. 2012), in physiology to distinguish between normal (healthy) and impaired cerebral autoregulation (Heskamp et al. 2014), in social systems for predicting the behaviour of users of social media (Luo et al. 2014) and in the climate system for detecting directionality in the relation between temperature and CO22 (Van Nes et al. 2015). In this study, we apply CCM to published marine and ice-core records spanning the past 800,000 years in order to shed light on the (directions of) interactions between climate, ocean circulation, biological productivity and dust deposition (Fig. 2).

The full article appeared on the Climate Dynamics website at https://link.springer.com/article/10.1007/s00382-021-05724-w

]]>

Join our effort to build a greener future.