Quantification of rapid temperature change during DO event 12 and phasing with methane inferred from air isotopic measurements

Abstract

The description of rapid climatic changes during the last glacial period at high northern latitudes has been largely documented through Greenland ice cores that are unique climatic and environmental records. However, Greenland ice isotopic records are biased temperature proxies and it is still a matter of debate whether changes in the high latitudes lead or lag rapid changes elsewhere. We focus here on the study of the mid-glacial Dansgaard Oeschger event 12 (45 ky BP) associated to a large δ¹⁸O_ice change in the GRIP (GReenland Ice core Project) ice core. We use combined measurements of CH₄, δ¹⁵N and δ⁴⁰Ar in entrapped air associated with a recently developed firn densification and heat diffusion model to infer (i) the phasing between methane and temperature increases and (ii) the amplitude of the temperature change. Our method enables us to overcome the difficulty linked with rapid accumulation change in quantifying the temperature change. We obtain a 12±2.5 °C temperature increase at the beginning of DO event 12 thus confirming that the conventional use of water isotopes in the Greenland ice cores largely underestimates the actual amplitude of rapid temperature change in central Greenland. In agreement with previous studies, methane and temperature increase are in phase at the sampling resolution of that part of our profile (90 years).

Introduction

Twenty-four rapid climatic changes during the last glacial period have been largely documented in the North Atlantic by ice cores, deep-sea and continental records during the last 20 years (e.g. [1], [2], [3], [4]). The succession of these rapid events has been related to shifts in the thermohaline circulation modes [5] drastically affecting the North Atlantic region. Some of these shifts appear to be triggered by fresh water inputs through abrupt iceberg discharges probably induced by ice sheet instabilities [6], [7]. Through the high resolution record given by Greenland ice cores, each Dansgaard-Oeschger event (hereafter DO event) is reflected as a rapid temperature increase probably caused by a resumption of the thermohaline circulation followed by a slow return to cold conditions induced by fresh water inputs in North Atlantic. However, if the role of both ocean modes and ice rafted events is central for the sequence of those Dansgaard-Oeschger events, some questions remain unresolved. Indeed, if the thermohaline circulation stability is invoked to explain the occurrence of the abrupt warming in the high latitudes [5], it is also suggested that tropical instabilities might have played a role to induce North Atlantic climatic variations [8].

Ice core records from Summit (GReenland Ice core Project, Greenland Ice Sheet Project 2) have provided a wealth of information on DO events. The water isotopes, here δ¹⁸O_ice [2], [9], [10] and chemical records both reveal dramatic and widespread reorganization of atmospheric transport. Ice cores have also revealed strong changes in atmospheric greenhouse gas levels as recorded in air bubbles: both CH₄ [11], mainly produced by tropical and boreal wetlands, and N₂O [12], [13], originating from terrestrial soils and the oceans, clearly undergo strong variations associated with DO events (more than half of a glacial–interglacial change) thus indicating large-scale climatic reorganisations. In order to better understand the causality and the sequence of DO events, it is crucial to determine whether high latitudes temperature is modified before or after greenhouse gases concentration. However, because air is enclosed in ice around 70 m below the surface in Greenland (depth of the firn close-off under present-day conditions), the air is younger than the surrounding ice at each depth level. Therefore, the difference, Δage, between the age of the ice and the age of the gas must be precisely evaluated. The use of a densification model [14], [15], [16] enables us to estimate the phasing between methane and temperature as deduced from the δ¹⁸O_ice record. However, the associated uncertainty is up to 10% of the Δage, that is to say at least 100 years for the onset of a DO event in GRIP mainly because of accumulation rates uncertainties. This is a major limiting factor to accurately determine the phasing between Greenland temperature and gas concentration increases using δ¹⁸O_ice as a temperature proxy.

A second problem that we are facing lies in the estimation of rapid DO event temperature change from the isotopic composition of the ice, δ¹⁸O_ice. Fractionations along the water trajectory between the evaporative regions and the high latitudes where snow precipitates produce a linear relation between δ¹⁸O_ice and mean annual temperature that is well obeyed in Greenland and Antarctica:

$$\delta {{}^{18}\mathrm{O}}_{\text{ice}}=\alpha T+\beta$$

Dansgaard [17] found a α_spatial value of 0.67 over Greenland. This present-day spatial relationship is also well captured both by simple isotopic model [18] and by general circulation models including water isotopes [19]. Using this relationship as a paleothermometer relies on the assumption that the spatial relationship does not change with time and consequently that the spatial and temporal slopes are similar [20] which is clearly not the case for Central Greenland as shown from comparison with independent estimates of temperature changes.

Such an independent estimate has been achieved by two borehole temperature profile inversions at GISP2 and GRIP [21], [22]. This method showed that the interpretation of the δ¹⁸O_ice profile using the spatial slope underestimates the Last Glacial Maximum (−21 ky)/Holocene temperature change by about 12 °C (i.e. by 100%). The temporal slope at Summit (GRIP, GISP2) during the last glacial maximum is then half the spatial slope: α_temporal=0.32 for the LGM. For the LGM/modern change, atmospheric general circulation model simulations suggest that this discrepancy is probably due to changes in the seasonality of the precipitation between glacial and interglacial periods with a huge decrease of winter snowfall during glacial periods [23], [24], [25]. A new method to reconstruct rapid temperature changes, based on air isotopic measurements gave estimates of the temporal relationship between δ¹⁸O_ice and temperature for particular abrupt climatic changes at the beginning and at the end of the last glacial period (DO event 19 [26], Bølling-Allerød [16], [27], [28]). The relationship appears to be intermediate between the current one and the LGM/today warming (note that we do not consider the marine isotopic correction on δ¹⁸O_ice because of uncertainties in the marine δ¹⁸O changes during rapid events [29]). Variations of the temporal slope during the last glacial period are easily conceivable because changes in the origin and the seasonality of Greenland precipitation can be modulated for instance by the volume of the Laurentide ice sheet.

In order to bring a new constraint on quantitative temperature estimates during the full glacial period, we use here this recently developed method [26], [27], [30], [31] based on the thermal diffusion of gases in the polar firn to estimate rapid surface temperature changes. This method is the only one currently available to get quantitative estimates of rapid temperature changes during the last glacial period. The method based on borehole temperature inversion is unable to resolve rapid climatic changes because of the smoothing of the temperature signal by heat diffusion in the ice sheet. We applied this method to DO event 12 at GRIP (~−45 ky). This rapid event was chosen to bring a new constraint on paleotemperature reconstructions in full glacial period at the mid-time between Bølling-Allerød and DO event 19 for which similar measurements are available [16], [26], [27], [28]. Moreover, DO event 12 is associated to a large signal in GRIP δ¹⁸O_ice (4.6‰ to be compared to the 7‰ associated with the LGM-modern change) which corresponds to a 6.8 °C change if using the spatial relationship between δ¹⁸O_ice and temperature [17]. This large warming reflects a switch from a very cold phase (Heinrich event 5 as recorded in marine sediments through ice rafted debris from iceberg discharge [32]). It is well suited for our method since it is associated with a large transient vertical temperature gradient in the firn that will in turn fractionate the gaseous species through molecular diffusion that is 10 times faster than heat diffusion. The δ¹⁵N and δ⁴⁰Ar isotopic anomalies reflecting the rapid temperature changes are then directly measured in the air trapped in the ice. In addition to a quantitative estimate of the temperature change, methane and temperature evolution are measured on the same depth scale, enabling the direct determination of the phasing by the combined gas measurements of δ¹⁵N and methane (we neglect the difference of diffusion coefficient between ¹⁵N¹⁴N and CH₄, which would amount to only a few years of age difference between both signals at close-off). To quantify such rapid surface temperature changes, previous studies have been conducted using δ¹⁵N alone [26], [30] or the combination of δ¹⁵N and δ⁴⁰Ar [27]. In the present paper we follow the method first introduced by Severinghaus and Brook [27] using both δ¹⁵N and δ⁴⁰Ar to isolate the thermal fractionation signal from its gravitational counterpart and we invert the thermal signal to quantify the past surface temperature variation through a recently developed firn densification model including heat diffusion [16]. This methodology is applied on detailed δ¹⁵N, δ⁴⁰Ar and CH₄ measurements performed on the GRIP ice core (resolution range: 20 cm to 1 m) at the depths corresponding to DO event 12. Sensitivity studies conducted on the speed of the temperature change enable us to estimate the uncertainty of this integrated method.