Hans Renssen
Faculty of Earth and Life Sciences, Vrije Universiteit Amsterdam, Netherlands

Hugues Goosse and Thierry Fichefet
Institut d´Astronomie et de Géophysique Georges Lemaître, Université catholique de Louvain, Louvain-la-Neuve, Belgium

Introduction
Paleoclimatic records have revealed that around 8200 years ago temperatures in the North Atlantic region abruptly decreased during a period of about 300-400 years^1-4. This cooling was accompanied by a drastic reduction in precipitation in Northern Africa and Asia⁵. It has been proposed that this climatic event was caused by a perturbation of the thermohaline circulation (THC) in the North Atlantic Ocean^2-4. The THC is an important heat source for Europe and is driven by the sinking of surface waters with a high density (i.e. deepwater formation) in the Nordic Seas. The hypothesis is that around 8200 year ago a collapse took place of the last dome of the Laurentide Icesheet that covered the Hudson Bay (North America). Clearly, such a collapse would have considerably increased the influx of freshwater into the North Atlantic through the melting of ice and the release of freshwater from ice-dammed proglacial lakes. Such a freshwater pulse could possibly have lowered the surface water density in the Nordic Seas below a particular threshold value, thus perturbing the formation of deepwater and weakening the THC. The 8,200 yr BP event has been used by Pentagon officials (Schwartz & Randall⁶) as a 'worst case' scenario for the future, as most climate models predict that the future human-induced global warming will cause substantial weakening of the THC⁷.

Our numerical experiments
Recently, we have done a number of experiments with the ECBilt-CLIO global climate model^8-9 to test the likelihood of this hypothesis. The experimental design consisted of two steps. The first step was to bring the model into equilibrium with boundary conditions of the period just before the event (i.e. 8500 years ago). The following changes in boundary conditions were applied compared to a present-day control simulation: different insolation, lowered atmospheric concentration of greenhouse gases (CO₂, CH₄ and N₂O) and changes in surface albedo. We assumed that vegetation changes did not play a significant role in forcing the 8.2 ky event. Starting from a present-day climate state, ECBilt-CLIO reached a new equilibrium with the changed boundary conditions after 500 years of simulation. This new equilibrium climate is characterized by a stronger seasonality (warmer summers and colder winters) over low- and mid-latitude continents in the Northern Hemisphere. The increase in seasonality is an expected response of the changed insolation and is in agreement with paleoclimate reconstructions.

Subsequently, the second step consisted of perturbation of this equilibrium state by performing several experiments in which different meltwater pulses were released in the Labrador Sea (Figure 1). The amount of meltwater was kept fixed at 4.67x10¹³ m³ and was based on estimates derived from geological data. The duration of the pulses, however, was varied, because published estimates ranged from 1 to 500 years. Four different pulses were introduced in the Labrador Sea, by releasing freshwater at the following linear rates: 1.5 Sv (1 Sv =1x10⁶ m³s^-1) in 10 yr, 0.75 Sv in 20 yr, 0.3 Sv in 50 yr and 0.03 Sv in 500 yr.

Figure 1: The main site of deep convection in our model is located between Norway and Svalbard (left panel). 15 years after the release of freshwater in the Labrador Sea, the surface of the Nordic Seas has become significantly fresher, leading to a perturbation of the deep convection south of Svalbard (right panel).

Results
With the 500-yr pulse, the model’s THC was only slightly weakened, as the overturning in the Nordic Seas decreased from 16 to 15 Sv. In the other cases, however, the overturning was significantly perturbed. In the case of the 10-yr pulse, overturning in the Nordic Seas was reduced to 4 Sv, after which the model shifted to a state of reduced overturning for at least 1000 years. In the 20-yr and 50-yr cases the model’s THC recovered within a few hundred years. Thus, the duration of the latter two simulated perturbation events was in agreement with the time-scale of the 8.2 ky event found in geological records. Moreover, associated with the THC perturbation, a cooling was simulated in the North Atlantic region (Figure 2) that was similar to what was suggested by paleoclimate reconstructions. In addition, the climate became substantially drier in the tropics, for instance in North Africa. This is also consistent with widespread paleoclimatic evidence (such as a lowering in lake levels⁵). Based on this agreement, we concluded that the simulation results support the hypothesis that the 8.2 ky event is caused by a freshwater perturbation of the North Atlantic THC.

Figure 2: Annual mean surface temperature anomaly (°C) during the simulated 8.2 ky BP event.
This result closely resembles proxy evidence for the 8.2 ky BP event, supporting the hypothesis
that the event was caused by a meltwater release in the Labrador Sea, associated with the final
stages of the Laurentide Icesheet
 

It was clear from the discussed freshwater perturbation experiments that the response of the model depends on the time-scale of the applied meltwater pulses. One important question remained unanswered, however: to what extent does the model’s response depend on the climatic conditions at the time of the perturbation? To shed light on this issue, we have repeated the 10, 20 and 50-yr freshwater perturbation experiments with four different initial conditions, resulting in three ensembles of 5 experiments each. In the five 50-yr cases, the model showed a similar response: recovery of the THC within a few hundred years. In the 20 and 10-yr perturbation experiments, the model behavior was more complex (Figure 3). The THC recovered in some instances, while in other cases it shifted to the weakened overturning state observed in the first 10-yr perturbation experiment. In the 20-yr cases, the model shifted to the weakened state in 2 out of 5 cases, whereas in the 10-yr cases this portion is 4 out of 5. Thus, it became clear that when the THC is brought close to a threshold (i.e. with 10 and 20-yr perturbations), the model’s response depends also on the high frequency climate variability. This implies that the THC’s response to freshwater perturbations may be unpredictable.
 

Figure 3: Simulation of the 8.2 ky BP event (Renssen et al., 2002): response of the thermohaline circulation to freshwater pulses of 0.75 Sv released during simulation years 550-570 in the Labrador Sea, for 5 ensemble members (each offset by 20 Sv for clarity). Other forcings are representative of 8.5 ky BP. The Figure shows that one type of forcing (0.75 Sv freshwater pulse during 20 years) applied to the same climate state (8.5 ky BP quasi-equilibrium) can lead to different responses of the Thermohaline Circulation.
A pdf-version of this slide can be downloaded at the website of CLIVAR.
 

For more information, see the following papers: