-
During transport between Antarctica and the University of Maine the core was exposed to above-freezing temperatures and some sections were partially melted and refrozen. As the core was transported lying down in the boxes, this melt and refreeze occurred in the external part of the core and did not reach the center of it. The melted and refrozen portion of the core was removed by saw and hand scraping, and only a small 10-cm piece of section 07 was discarded as it was totally refrozen.
We used an ice core light table to observe the core stratigraphy. Millimeter-thick lenses of ice were observed all along the core, probably due to summer melting. Additionally, a few depth hoar layers up to 1 cm thick were observed. There were no visible dust layers.
The core density ranged from 0.38 to 0.60 g cm−3, not reaching the firn/ice transition of 0.83 g cm−3 (Fig. 2). We averaged the TT07 density profile with the density profile of another core drilled in the same area of Antarctica (45 m deep; Schwanck et al., 2016b), fitted a quadratic trend line to the average curve, and used this trend line to calculate the snow accumulation, water equivalent (weq), and rBC fluxes for this work. We found an average snow accumulation of 0.23 ± 0.06 weq m yr−1 for the entire core, so the 20.16-m length core represents 10.65 weq m. For the 8 m analyzed in this work, the snow accumulation was 0.21 ± 0.04 weq m.
Figure 2. Density profile of the snow and firn core analyzed. Depth is presented in meters and water equivalent (weq) meters. The quadratic fit was calculated from the average density profile from this work and from Schwanck et al. (2016b).
-
The first eight sections of the core, presented in this work, were dated to 17 years by annual-layer counting using mainly the rBC seasonal variability, as this is a reliable parameter for dating (Winstrup et al., 2017). Data from Sentinel Hotspots indicate fires in Australia tend to peak in October, with the seasonal increase in fire activity occurring in August and the decrease in December/January. The Programa Queimadas data show that fires in South America tend to peak in September, with the seasonal increase in fire activity occurring in June/July and the decrease in November/December. A comparison between the seasonality of burning and the TT07 rBC record is presented in Fig. S3 in ESM.
As a support to this, we used sulfur (S), strontium (Sr) and sodium (Na), as these records show the more pronounced seasonal variability at the site (Schwanck et al., 2017), although we only had these analyzed down to ~7 m of the core. Also, the S, Sr and Na records are from a different core, retrieved a meter apart from the rBC core, and that core was subsampled and analyzed in another laboratory (CCI), meaning there could be some displacement from this record to the rBC one. The dating is presented in Fig. 3.
Figure 3. Dating of the snow and firn core based on BC, S, Sr and Na records. Dashed lines indicate the estimated New Year. In this figure, we present the BC record from the CMS, as this was smoother than that from SSC. Gray shaded areas are the dry-season (winter/spring) concentrations, while white areas are the wet-season (summer/fall) concentrations.
BC in Antarctica tends to peak during winter−spring (dry season) owing to drier conditions in the Southern Hemisphere and a consequent increase in biomass burning (Bisiaux et al., 2012b; Sand et al., 2017; Winstrup et al., 2017). Na and Sr also peak during this time, due to intense atmospheric circulation and transport (Legrand and Mayewski, 1997), while S peaks in late austral summer in relation to marine biogenic activity (Schwanck et al., 2017). We considered our new year to match the end of what we define as the dry season, as this is a reliable tying point in the record because of the abrupt drop in rBC concentrations based on the fire-spot database from Australia and South America. This is also in agreement with Winstrup et al. (2017), who stated that rBC tends to peak a little earlier than New Year in their records (Roosevelt Island Ice Core).
-
The nebulization efficiency for the Marin-5 at CWU was calculated to be 68.31% ± 5.91% (1σ), based on external calibration carried out every working day using the Aquadag standards (see section 3.1). We found a decrease in nebulization efficiency during the laboratory work period (−0.31% per working day or −13.3% over the 43 working days), but we assume the nebulization efficiency to have remained stable between the measurement of the standard and the samples measured for the day, as in Katich et al. (2017). We attribute this decrease to the Marin-5, but do not see any apparent cause. Pump flow rates were kept constant at 0.14 ± 0.02 mL min−1 at all times during analysis. This result highlights the importance of making daily Aquadag standards.
-
Samples were analyzed for 5 min each. Although a low particle count could increase the uncertainty of the method, we noticed that the measurements did not vary significantly in relation to analysis time, but much more so in relation to the sample average concentration itself.
To address this issue, we analyzed samples of varied rBC concentrations along the entire core more than once and for different periods of time. Each sample was analyzed between two and four times, for 5, 20 and/or 40 min. The samples were analyzed less than 2 h after melting to avoid rBC loss (Wendl et al., 2014).
While we observed no significant concentration variations for different analysis times (Fig. 4), our coefficient of variation (mean of all measurements of the sample × standard deviation) for concentrations lower than 0.03 µg L−1 was 25.7 ± 16.9 (1σ, n = 38), 10.4 ± 6.6 (1σ, n = 24) for concentrations between 0.03 and 0.07 µg L−1, and 7.3 ± 4.4 (1σ, n = 51) for concentrations higher than 0.07 µg L−1 (Fig. 5).
Figure 4. Changes in BC concentration (y-axis; 1 = 100%) for different BC concentrations (x-axis) for the three different analysis times. Note that when analyzing low-concentration samples (< 0.03 μg L−1), even for long times (40 min) the changes in BC concentrations are significant. Values are relative to the first measurement taken of each sample.
Figure 5. Coefficient of variation (CoV) for the samples analyzed with the CWU SP2 in the reproducibility test. Vertical axis = CoV for concentrations (gray dots).
We attribute this variation to the number of collected particles in each sample: low-concentration samples mean low particle triggers, which will lead to a higher variance in case rare particles large enough to contain a considerable fraction of total rBC mass are recorded.
-
We found a well-marked seasonal rBC cycle along the core (Fig. 6), with the same pattern of low summer/fall and high winter/spring concentrations as reported by Bisiaux et al. (2012b).
As we collected our samples in January and the drilling was carried out from the snow surface, our core starts approximately in the New Year. As mentioned earlier, BC in Antarctica tends to peak during winter/spring, and so the New Year in the record is generally viewed as a steep decrease from peak concentrations to low concentrations. This was better observed in the CMS samples than the SSC ones for the 2014−15 transition.
Both sampling methods showed similar seasonality, but the CMS provided a smoother record (e.g., less summer/fall spikes) and a generally lower summer/fall concentration. Table 1 presents the details of this comparison. We attribute the smoother record to reduced handling of the core, as with SSC the individual samples were handled after decontamination to put them in the clean vials, which could have caused cross-contamination between samples to some degree. Nonetheless, a Wilcoxon−Mann−Whitney test indicated there to be no statistical difference between the two sample datasets at p = 0.01 (N = 650; two-tailed P-value = 0.449758; see Methods S1 in ESM).
Solid-state sampling Continuous melting system Totala Geomean 0.031 0.029 1σ* interval 0.013−0.073 0.011−0.076 Lowest/highest conc. 0.003/0.701 0.001/0.262 Wet-season Geomean 0.019 0.016 1σ* interval 0.011−0.032 0.008−0.027 Lowest/highest 0.003/0.083 0.001/0.071 Dry-season Geomean 0.065 0.074 1σ* interval 0.029−0.121 0.035−0.128 Lowest/highest 0.007/0.701 0.014/0.262 aAll samples from section 1 of the core (surface) down to section 8 (around 8 m deep). Total number for solid-state sampling is 307, and for continuous melting system is 343. Table 1. Main results from the comparison between SSC and the CMS. All values are in units of μg L−1. “Geomean” refers to the geometric mean, and 1σ* is the multiplicative standard deviation, representing 68.3% of the variability (Limpert et al., 2001; Bisiaux et al., 2012a).
For SSC, concentrations ranged from 0.003 µg L−1 to 0.701 µg L−1, with a geometric mean of 0.031 µg L−1 (n = 307). Concentrations using the CMS ranged from < LOD (0.0015 µg L−1) to 0.262 µg L−1, with a geometric mean of 0.029 µg L−1 (n = 343).
Summer/fall averages for both methods were also similar, with differences regarding summer/fall highest values due to concentration peaks in the SSC method that did not alter the mean significantly. Winter geometric means were similar for both methods (CMS = 0.074 µg L−1; SSC = 0.065 µg L−1); the winter maximum showed a pronounced difference owing to an anomalous peak around the depth of 3 m, wherein the discrete sampling two consecutive samples achieved 0.701 µg L−1 and 0.568 µg L−1, while the continuous melter gave a maximum of 0.147 µg L−1 for the same depth. This almost five-fold difference did not appear anywhere else in the core, probably reflecting contamination in the samples, and thus these two SSC samples are not considered in further interpretations.
Figure 7 shows a dry- versus wet-season comparison for both methods. The results are similar for both methods: summer/fall values remain fairly steady for the entire record; winter/spring concentrations show an initial peak in 1998 and 1999 AD, followed by a low in 2002 and an increasing trend from 2002 to 2014—more visible in the CMS record (but with a weak r2 of 0.2478, not shown).
Figure 7. Comparison between dry- and wet-season average concentrations for both sampling methods: SSC (squares); CMS (triangles). Low-concentration lines are from the wet season; high-concentration lines are from the dry season.
Annual rBC fluxes were calculated to account for potential biases in annual rBC concentrations due to changes in snow accumulation rates. Fluxes were calculated by multiplying annual rBC concentrations by the annual snow accumulation. rBC annual concentrations were averaged from SSC and the CMS. Concentrations and fluxes followed a similar pattern, implying low variability in snow accumulation during the study period (Fig. 8).
-
Table 2 compares our results with other rBC records in Antarctica. East Antarctica cores [NUS0X from Bisiaux et al. (2012a)] present the highest elevations and annual rBC concentrations, but the lowest snow accumulation, in recent times (~1800−2000). The authors found a linear positive correlation between site elevation and rBC concentrations for the NUS07 cores, of 0.025 µg L−1 (500 m)−1, and hypothesized that rBC inputs to the atmosphere over East Antarctica are not controlled by the intrusion of marine air masses and that transport in the upper troposphere may be more important.
Source Core name Location in Antarctica Lat./Long. Elev. (MSL) Period covered Annual BC conc. (μg L−1) BC conc. range (2σ)a Annual accum.
(weq m yr−1)Annual BC fluxes
(μg m−2 yr−1)BC flux
range (2σ)This study TT07 West 79°55'S, 94°21'W 2122 1998−2015 0.029 0.01−0.07 0.21 ± 0.04 6.1 2.6−14.6 Bisiaux et al. (2012b) WAIS West 79°46'S, 112°08'W 1766 1850−2001 0.08 0.05−0.12 0.20 ± 0.03 16 9.8−24.4 Law Dome East 66°73'S, 112°83'E 1390 1850−2001 0.09 0.05−0.2 0.15 ± 0.03 13.5 7.3−30.6 Arienzo et al. (2017) WAIS West 79°46'S, 112°08'W 1766 14−12 kBP 0.12b − − 25b − 12−6 kBP 0.2b − − 45b − B40 East 70°0'S, 0°3'E 2911 2.5 k−0 BP 0.3c − − 20c − Bisiaux et al. (2012b) NUS07-1 East 73°43'S, 07°59'E 3174 1800−2006 0.16 0.09 to 0.26 0.05±0.02 8.3 4.6 to 14.2 NUS07-2 76°04'S, 22°28'E 3582 1800−1993 0.12 0.07−0.19 0.03 ± 0.01 3.9 2.5−6.2 NUS07-5 78°39'S, 35°38'E 3619 1800−1989 0.14 0.08−0.26 0.02 ± 0.01 3.4 1.8−6.3 NUS07-7 82°49'S, 54°53'E 3725 1800−2008 0.18 0.12−0.27 0.02 ± 0.01 5.3 3.5−8.0 NUS08-4 82°49'S, 18°54'E 2552 1800−2004 0.1 0.06−0.18 0.04 ± 0.01 3.7 2.1−6.9 NUS08-5 82°38'S, 17°52'E 2544 1800−1993 0.11 0.07−0.18 0.03 ± 0.01 3.9 2.2−6.5 a Multiplicative standard deviation representing 95.5% of the confidence interval; b 50-year average, not annual; c 7-years media, not annual Table 2. Coordinates, elevation, period covered and BC information for this study and previous works on Antarctic ice cores. To enable direct comparison, we only list studies that used the SP2.
Arienzo et al. (2017) found an even higher annual rBC concentration for the coastal site B40 (0.3 µg L−1), where the flux was calculated to be 20 µg m−2 yr−1. As BC is primarily deposited through wet deposition (Flanner et al., 2007), the authors attributed the higher accumulation in coastal areas to the scavenging of most of the BC, with fluxes lowering inland as the accumulation rates decreased.
Arienzo et al. (2017) also found high rBC fluxes for the WAIS ice core for the end of the last glaciation termination (14−12 kBP, 25 µg m−2 yr−1) and for the mid-Holocene (12−6 kBP, 45 µg m−2 yr−1). The authors attributed the high rBC fluxes in the past to a period of relatively high austral-burning-season and low growing-season insolation.
The WAIS ice core (Bisiaux et al., 2012b; Arienzo et al., 2017) is the closest to TT07 (350 km apart). Although the annual snow accumulation is similar at both sites (0.21 ± 0.04 weq m yr−1 for TT07 in this work; 0.20 ± 0.03 weq m−1 for WAIS), our annual rBC concentration is less than half that of WAIS during 1850−2001 (0.031 µg L−1 for TT07; 0.08 µg L−1 for WAIS). The rBC flux is also lower (6.1 µg m−2 yr−1 for TT07; 16 µg m−2 yr−1 for WAIS), although we acknowledge there is not a large temporal overlap between the cores (three years, 1998−2001).
Solid-state sampling | Continuous melting system | |||
Totala | ||||
Geomean | 0.031 | 0.029 | ||
1σ* interval | 0.013−0.073 | 0.011−0.076 | ||
Lowest/highest conc. | 0.003/0.701 | 0.001/0.262 | ||
Wet-season | ||||
Geomean | 0.019 | 0.016 | ||
1σ* interval | 0.011−0.032 | 0.008−0.027 | ||
Lowest/highest | 0.003/0.083 | 0.001/0.071 | ||
Dry-season | ||||
Geomean | 0.065 | 0.074 | ||
1σ* interval | 0.029−0.121 | 0.035−0.128 | ||
Lowest/highest | 0.007/0.701 | 0.014/0.262 | ||
aAll samples from section 1 of the core (surface) down to section 8 (around 8 m deep). Total number for solid-state sampling is 307, and for continuous melting system is 343. |