Atmospheric methane control mechanisms during the early Holocene

Understanding processes controlling the atmospheric methane (CH4) mixing ratio is crucial to predict and mitigate future climate changes in this gas. Despite recent detailed studies of the last ∼ 1000 to 2000 years, the mechanisms that control atmospheric CH4 still remain unclear, partly because the late Holocene CH4 budget may be comprised of both natural and anthropogenic emissions. In contrast, the early Holocene was a period when human influence was substantially smaller, allowing us to elucidate more clearly the natural controls under interglacial conditions more clearly. Here we present new high-resolution CH4 records from Siple Dome, Antarctica, covering from 11.6 to 7.7 thousands of years before 1950 AD (ka). We observe four local CH4 minima on a roughly 1000-year spacing, which correspond to cool periods in Greenland. We hypothesize that the cooling in Greenland forced the Intertropical Convergence Zone (ITCZ) to migrate southward, reducing rainfall in northern tropical wetlands. The inter-polar difference (IPD) of CH4 shows a gradual increase from the onset of the Holocene to ∼ 9.5 ka, which implies growth of boreal source strength following the climate warming in the northern extratropics during that period.

and Indian summer monsoons and probably reduces CH4 emission from northern tropical wetlands. Sediment reflectance record from Cariaco Basin shows increased rainfall and humiditywhich is due to southward displacement of ITCZcorresponding to each abrupt cooling event, as revealed in previous studies for different time periods (Peterson et al., 2000;Haug et al., 2001;Fleitmann et al., 2007;Deplazes et al., 2013). Moreover, 18 O enrichments of Asian (Dongge) and Indian 25 (Hoti and Qunf) cave stalagmites occurred at similar timing with abrupt cooling in Greenland, which indicate the reduction of monsoonal rainfall at northern tropical wetlands. The speleothem records from Chinese and Oman caves seem to lag by ~100-200 years after the CH4 change at ~9.3 ka, but this lies within chronological uncertainties of ~200-400 years at around 9.0 ka (Dykoski et al., 2005;Fleitmann et al., 2007).
Previously, Björck et al. (2001) found the climate cooling in the northern Atlantic and Santa Barbara Basin occurred with 30 solar-forcing change at ~10.3 ka. However, in the proxy data, there is no clear indication of southward migration of ITCZ position and reduction of Asian summer monsoon intensity corresponding to ~10.2 ka cooling (Fig. 2). Instead, there are two Clim. Past Discuss., doi:10.5194/cp-2016-75, 2016 Manuscript under review for journal Clim. Past Published: 13 July 2016 c Author(s) 2016. CC-BY 3.0 License. small decrease at ~9.9 and ~10.6 ka, but these episodes are likely beyond the chronological uncertainty, considering that the age uncertainty of Dongge Cave deposits is ±77 years (2 sigma error; Dykoski et al., 2005), and that the Siple Dome age uncertainty is likely less than ~100 years (see above and Fig. S1). This could be because the climate teleconnection between North Atlantic and tropical hydrology was not sufficiently strong enough to change the low latitude climate. Weak cooling around the North Atlantic region can be a candidate, given that NGRIP δ 18 Oice records demonstrate smaller amplitude negative 5 anomaly during ~10.2 ka event than those of 8.2 and 9.3 ka, but this is not supported by other Greenland ice core records such as Greenland Ice Core Project (GRIP) and Greenland Ice Sheet Project 2 (GISP2) (Fig. S2).
Although there appears to have been no strong change in low latitude hydrology at 10.2 ka, the amplitude of CH4 decrease at 10.2 ka is similar order to the other millennial events. This may imply CH4 reduction was controlled by other processes than the Asian monsoon intensity change. If the climate proxy from Dongge cave reflects rather regional climate changes, 10 monsoonal rainfalls and surface hydrology of other regions could be responsible for CH4 decrease. The speleothem δ 18 O records from Mawmluh Cave show no weakening of the Indian monsoon (Berkelhammer et al., 2012), moreover, there was no distinct change in ΔεLAND, a proxy of global terrestrial respiratory fractionation of atmospheric oxygen, which is affected by low latitude surface hydrology (Severinghaus et al., 2009). This evidence suggests that changes in precipitation and surface hydrology in the northern tropics may have not changed significantly during around the 10.2 ka. 15 Increased boreal source contribution is a probable compensating mechanism for smaller reduction in tropical CH4 emission.
Terrestrial proxies in boreal regions showed the formation of peatlands and thermokarst lakes, which occurred first in Alaska and northern Canada from ~11.0 ka, followed by Siberia and Asia (MacDonald et al., 2006;Brosius et al., 2012;Yu et al., 2013). Exposing a new land by deglaciation (Dyke, 2004) and temperature increase over northern high latitude (Marcott et al., 2013) established a favourable climate condition for thawing permafrost, thus releasing CH4 from there. We will discuss it 20 further below.

External forcing
Figure 2(a) shows a possible cause of the observed millennial scale climatic changes and abrupt cooling recorded in Greenland ice cores. Four large ice-rafted debris (IRD) drift deposits occurred during the early Holocene at ~8.5, 9.3, 10.3 and 11.3 ka (Bond et al., 2001). This record lacks a large IRD deposit that corresponds to 8.2 ka cooling (Bond et al., 2001). Later 25 study found that increase of hematite stained glass (HSG) at the timing of 8.5 ka should be revised to 8.2 ka based on quartzto-plagioclase ratio analysis (Moros et al., 2004). Additionally, Bond et al. (2001) found that 1500-year cycle of IRD in the North Atlantic are concurrent with the global climate cooling and the negative solar activity inferred by ice core 10 Be and Δ 14 C records. From this evidence the authors speculated that the solar influence should be amplified by changes of sea-ice and/or in deep water formation in the North Atlantic. However, the forcing mechanism of solar activity on the North Atlantic and global 30 climate is not well understood during the early Holocene. Renssen et al. (2006) suggested that low solar activity (in terms of total solar irradiance) can induce sea-ice expansion around the Nordic Seas and weakening of deep water formation and cooling Clim. Past Discuss., doi:10.5194/cp-2016-75, 2016 Manuscript under review for journal Clim. Past Published: 13 July 2016 c Author(s) 2016. CC-BY 3.0 License.
in North Atlantic region. Nevertheless, the anti-correlation between solar forcing and sea-ice expansion (and hence deep water formation weakening) is not strong during the early Holocene due to relatively warm climate conditions. Jiang et al. (2015) also found a negative correlation between North Atlantic SST and solar forcing proxies ( 14 C and 10 Be), which is statistically significant for the last 4000 years, while the correlation disappeared during the mid-and early Holocene. They hypothesized that climate sensitivity to solar forcing is high for cooler climate.

Inter-polar difference of CH4 and source distribution model
We calculated inter-polar difference (IPD) of CH4 to trace the latitudinal source distribution change during the early 20 Holocene. In this study, the IPD was calculated by using our Siple Dome CH4 record and a NEEM high resolution discrete CH4 record (Chappellaz et al., 2013). Precise synchronization is crucial for direct comparison between data sets which have high frequency variations. The NEEM CH4 record is chosen as a reference because the mean time resolution is higher than our data set. Synchronization was done by two steps: First, we made initial synchronization between the Siple Dome and NEEM data by setting 7 match points, and then we linearly interpolated the age offset of each match point for the rest of data points. 25 Then we applied a Monte Carlo simulation to find a maximum correlation. Both data sets were resampled every 30 years, and each point was randomly disturbed (assuming a normal distribution with 1 sigma of 30 years). By doing so 1000 different time series were created, and one set having a maximum correlation with NEEM data was chosen. Criteria for "best fit" is correlation coefficient of 0.8 with NEEM original age scale, so that a maximum correlation less than 0.8 was discarded. This procedure was repeated to make 20 sets of maximum correlation time series, and the mean ages of 20 replicate simulations were set to 30 synchronized age scale. Temporal uncertainty (synchronizing error) was determined for each point as 1 standard deviation of 20 replicates and CH4 uncertainty includes analytical error of the both records (4.3 ppb for NEEM and 1.0 ppb for SDMA).
Clim. Past Discuss., doi:10.5194/cp-2016Discuss., doi:10.5194/cp- -75, 2016 Manuscript under review for journal Clim. Past Published: 13 July 2016 c Author(s) 2016. CC-BY 3.0 License. Fig. 3 shows our IPD results with 95 % significance envelope. Our IPD agrees with the previous low-resolution estimates for the earlier part of the Holocene (9.5~11.5 ka) (Chappellaz et al., 1997;Brook et al., 2000). Our results show an increase from ~10.7 ka to ~9.9 ka, which was not previously reported. Considering the long-term decreasing trend of CH4 mixing ratio in both poles during the early Holocene, the increasing IPD implies that the amount of boreal emission reduction should have been less than that of low latitude emissions. 5 Given the new high resolution CH4 records from both poles and IPD, we ran a simple 3-box CH4 source distribution model to quantify how much the boreal and tropical source strengths were changed. Here we used the same box model employed in Chappellaz et al. (1997) andBrook et al. (2000). Briefly, the model contains 3 boxes; northern high latitude (30-90°N, N-box), tropical (30°S-30°N, T-box), and southern high latitude boxes (30-90°S, S-box). CH4 concentrations in 3 boxes (in Tg box -1 ) were determined from CH4 mixing ratio of Antarctica and Greenland. To calculate the N-box CH4, we subtracted the 7 % of 10 IPD from Greenland CH4 concentration, assuming the difference between Greenland and the mean latitude of N-box is ~7 % of IPD (Chappellaz et al., 1997). T-box concentration is inferred by assuming that the S-box emission is constant of 15 Tg yr -1 (Fung et al., 1991). Emission from each box (Tg yr -1 ) is then estimated by using the concentration of the boxes, lifetime of each box, and transport times among the boxes.
The results of our model are consistent with previous estimates by Chappellaz et al. (1997) and Brook et al. (2000). Although 15 the early studies reported average value for the 11.5-9.5 ka interval, our IPD records show similar value before the IPD starts to rise at ~ 10.7 ka ( Fig. 4 and Table 1). After that, our results show an increase of boreal emission by 9 Tg yr -1 and a decrease in tropical emission. Boreal source fraction, a ratio of boreal emission to total emission, reveals an increase by 5 %. This result supports our interpretation that the boreal sources were less reduced than those in low latitudes. This conclusion is supported by proxy-based temperature reconstructions that indicate a gradual warming in northern high 20 latitude region (30N-90N) until ~9.6 ka, while tropical temperature remains stable (Marcott et al., 2013). The climate warming in northern high latitudes caused ice sheet retreat (e.g., Dyke, 2004) and enhanced CH4 emission from boreal permafrost by forming new wetlands in mid-to high latitudes (e.g., Gorham et al., 2007;Yu et al., 2013) and accelerating microbial decomposition of organic materials (e.g., Christensen et al., 2004;Schuur et al., 2015). Thermokarst lakes created by thawing ice wedges and ground ices in Alaskan-and Siberian permafrost are suggested as a source of CH4 (e.g., Walter et al., 2006Walter et al., , 25 2007Brosius et al., 2012). Indeed, the increased boreal CH4 emission of 9 Tg yr -1 is in similar order of the CH4 release of 8.2 Tg yr -1 from thermokarst lake reported by Walter Anthony et al. (2014). However, it should be noted that the CH4 release estimates from the thermokarst lakes are based on present-day CH4 flux measurements in Siberian-and Alaskan lakes and that 9 Tg yr -1 is a small change in the budget that could be driven by conventional northern CH4 emission. A recent study also argued a possibility of underestimation of such CH4 emission measurements (Wik et al., 2016). 30 We could not estimate the IPD for the later part of the record (7.7 ~ 8.8 ka) due to a lack of high resolution CH4 from Greenland ice cores. However, since the first-generation lakes produce CH4 more actively than later-generation lakes formed after drainage (Brosius et al., 2012), it is unlikely that thermokarst lake CH4 emission would remain higher after 9.0 ka. Future Clim. Past Discuss., doi:10.5194/cp-2016-75, 2016 Manuscript under review for journal Clim. Past Published: 13 July 2016 c Author(s) 2016. CC-BY 3.0 License. study should include extending high resolution CH4 record from Greenland, as well as CH4 isotope ratio data for the younger time period.

IPD during the Pre-Boreal Oscillation (PBO)
We also observed a high IPD at the earliest part of the Holocene, where CH4 records from both poles show a large variability. This could be due to mismatching of synchronized time scales and different surface conditions of drilling site and hence signal 5 attenuation process within firn. A sensitivity test on synchronizing error has been carried out by shifting the reference age scale (in this study, NEEM chronology) 15 years back and forth given that the initial age match points were resampled every 30 years. The IPDs calculated with shifted age scales (plus 15 years, control, and minus 15 years) are plotted in Fig. S3, showing a consistent high IPD values during ~11.0 to 11.2 ka interval, while for the earlier part IPD seems to be highly sensitive to synchronization. Nevertheless, this might be a result of different gas enclosing processes within firn layers in both ice cores. 10 As the width at half-height of the gas age distribution at NEEM site was reported as ~32 years (Buizert et al., 2012), which is ~23 % narrower than that of Siple Dome (Ahn et al. 2014). It means that the NEEM signal has been less attenuated than Siple Dome one, which could result in higher (lower) IPD at the period where rapid CH4 increase (decrease) is observed. Indeed, the discrete and continuous CH4 record from WAIS Divide, which has a mean accumulation rate similar to NEEM (Buizert et al., 2013), shows ~10 to 20 ppb higher amplitude variability. 15 Previous studies that, using the stable isotopic composition of C and H in CH4 that aimed to disentangle the cause of abrupt CH4 increase during the earliest period of the Holocene have shown contradictory results. Schaefer et al. (2006) calculated isotopic (δ 13 C-CH4) mass balance model to discern major source term that caused a slight enrichment in 13 C during the Younger-Dryas termination, suggesting tropical wetland emission as a dominant source. The authors also proposed biomass burning, geologic CH4 and enhanced sink process at marine boundary layer as alternatives, but less probable scenarios. On the 20 other hand, Fischer et al. (2008) argued nearly constant biomass burning emission of ~45 Tg yr -1 throughout the last glacial termination with a slight increase in PB, and also showed that the boreal sources were expanded during the YD-PB transition.
The triple isotopic mass balance model using δ 13 C-CH4, δD-CH4 and Δ 14 C-CH4, Melton et al. (2012) suggested the biomass burning and thermokarst lakes are the most important additional sources. The enhanced biomass burning agrees with global charcoal influx, an independent proxy for wildfire, which shows intensified wildfire in northern tropical regions (Daniau et al., 25 2012). However, it is unlikely that the increased pyrogenic emission in tropics leads to higher IPD. Brosius et al. (2012), using an isotopic mass balance model including thermokarst lake sources, suggested another scenario that the enhanced boreal wetland emission contributed largely for the CH4 overshoot. In the meanwhile, the boreal emission hypothesis was refuted by a recent study of 14 C-CH4 change during the YD termination that revealed the major carbon source for abrupt CH4 doubling was not the permafrost-origin old carbon (e.g., Petrenko et al., 2009Petrenko et al., , 2015. Therefore, the cause of the high IPD at the start of 30 the Holocene still remains elusive. Clim. Past Discuss., doi:10.5194/cp-2016Discuss., doi:10.5194/cp- -75, 2016 Manuscript under review for journal Clim. Past Published: 13 July 2016 c Author(s) 2016. CC-BY 3.0 License.