Gestion des ressources génétiques forestières

Causes of millennial variations of 10Be flux 

Climatic influence 

It is unlikely that the millennial variability identified in the 10Be spectra is due climatic variations as suggested from the comparison of the 10Be flux and δD wavelet spectra calculated using the squared wavelet coherence (WTC) reported on the Fig. 5.3. The WTC can be defined as a localized correlation coefficient in time frequency space (see section C.1 in the appendix for the description of the wavelet package from Grinsted et al. [2004]). No significant correlation is observed between these two variables during this climatic cycle at EDC (r = 0.0595 with 95% confidence interval [0.0035 ; 0.1152]). The WTC of 10Be flux and δD allows us to investigate the correlations between both signals more locally. Again, no correlation is detected, except a significant region during the MIS 9.3 with an anti-phase relationship between 10Be and δD for a period of ∼9000 years. This period corresponds to the time interval where δD, and hence accumulation rate, varies more significantly. This probably means that this anti-correlation is an artifact due to the estimate of accumulation rate from δD during the studied period and not the presence of a climatic component in the 10Be flux. For the shorter periods, significant regions are very sparse without a dominant mean phase angle. Periods shorter than 1200 years cannot be considered because of the resampling applied on 10Be and δD data (∼345 years).

Transport and deposition effects

 Because 10Be atoms are likely attached to aerosols after their production, they can be influenced by transport and deposition effects before their trapping in the ice. We study here their influence on the millennial variations of our measured 10Be signal via a comparison with a record of aerosols concentration, that is high resolution sulfate data at EDC [Traversi et al., 2009]. Sulfate record has been resampled to the resolution of our 10Be measurements (11 cm) before calculating its flux (in ng.m−2 .s−1 ) in the same way than for 10Be (Fig. 5.4). The correlation coefficient between 10Be and SO4 2− moves from r = 0.65 (with 95% confidence interval [0.6278 ; 0.6758]) for the concentrations profiles to r = 0.1165 (with 95% confidence interval [0.0751 ; 0.1575]) for the flux. We indeed directly observe that there are no common millennial variations in our 10Be profile and in the SO4 2− flux record. This means that the millennial changes in our 10Be profile cannot be mainly explained transport and deposition effects, but come from another source. 

Dating uncertainties 

To test the sensitivity of 10Be changes to chronology, we compare by spectral analysis 10Be flux on three different age scales: the conventional EDC3 age scale, DFGT–2006 age scale from Dome Fuji [Kawamura et al., 2007; Parrenin et al., 2007a,b] and the new multi-proxy chronology AICC2012 [Bazin et al., 2013]. The comparison of the different 10Be flux spectra on the three different chronologies (top of the Fig. 5.2) shows that the most significant period of ∼10500 years, and so the millennial changes of 10Be flux, are not affected by chronology. 

Validity of the interpretation of glacial–interglacial accumulation and temperature change in the EDC ice core

 It is interesting to test the assumption of a strictly constant 10Be flux, consequences on the accumulation rate variations and whether or not it allows to keep the EDC chronology consistent with the ascribed time markers. Over this period, the time markers for the EDC ice core are deduced from total air content aligned with an integrated summer insolation curve [Raynaud et al., 2007]. To examine the impact of the assumption of 10Be flux on the chronology we have used using the bayesian dating tool Datice [Lemieux-Dudon et al., 2010]. Fig. 5.5 shows the accumulation rate curves after the processing by Datice to obtain the best compromise between the background chronology (built from the proposed accumulation rate and the thinning obtained by glaciological modeling) and observations (age markers). The Datice tool easily converges towards the prescribed age markers. Such a result is not surprising since our 10Be record is sufficiently deep in the core with a large uncertainty on the value of thinning which can easily compensate for changes in accumulation rates. For such an extreme case, the resulted amplitude of the accumulation rate during the glacial–interglacial transition is only different by 28% from the EDC3 one. In turn, we do not retain this assumption of a constant flux as much as the millennial variability of the 10Be flux derived using the EDC3 accumulation (Fig. 5.1) is rather small and quite comparable with what is observed during the Holocene. If the assumption of a strictly constant flux is unrealistic, it is reasonable to expect that this parameter does not show any correlation with δD at the time of glacial-interglacial transition. The relationship between 10Be flux and δD between ∼340 and 320 kyr BP indeed directly depends on the amplitude of the accumulation change between glacial and interglacial periods. In the formulation adopted by Parrenin et al. [2007a] this accumulation, A, is defined as A0 exp(β∆δDsmo) in which A0 is the present-day accumulation rate (2.841 cm-ie/yr at EDC) and ∆δDsmo the deviation from the present-day δD values (corrected for the variations in isotope and temperature at the source of the air masses, and 50-yr smoothed). β is an adjustable parameter which is optimized during the construction of the ice core chronology in order to respect thinning and age markers constraints. For example the EDC3 chronology uses a β of 0.0157. Keeping the consistency with age markers, we have searched to adjust β in order to minimize the variance of the signal (which is easily compensated by changes in the thinning function as shown in the previous paragraph). We have limited this search on the time interval from 341 to 319 kyr BP which although encompassing the full amplitude of glacial–interglacial changes, exhibits a small millennial variability. The variance is minimized for an increase by 7% which corresponds to a larger glacial-interglacial amplitude by the same amount (the variance remains around its minimal value for 0.0163 < β < 0.0173). When we apply this increased value of β to our whole record, we also notice a general decrease of the variance which validates this revision of accumulation rate estimate from δD over one glacial–interglacial cycle. Still, this change in the amplitude of the accumulation rate does not modify the general shape of the 10Be flux record and preserve its spectral properties. Earlier chronologies developed for Antarctic ice cores were based on a different approach, e.g. on the assumption that accumulation changes are driven by the derivative of the saturation pressure, as would be the case in a simple 1D model [Ritz, 1992]. Under this assumption, the accumulation change depends on the temperature change in the atmosphere which is linearly related to surface temperature change, ∆Ts [Jouzel and Merlivat, 1984], itself assumed to be related to δD, ∆T = δD/α. With this formulation, the variance of the 10Be flux is minimized for an α value 12% higher than that corresponding to the present-day observed isotope/temperature slope used in the conventional approach known as the “isotopic paleothermometer”. As for the earliest formulation, the variance of the 10Be flux is minimized over the whole period 270 – 355 kyr BP. Despite the limitations of this simple approach, this result gives support to this conventional approach which may however slightly underestimate the estimated glacial-interglacial temperature changes, e.g. by about 10%, a value similar to that derived by Jouzel et al.