Mineral Dust Influence on the Glacial Nitrate Record from the RICE Ice Core, West Antarctica and Environmental Implications

Nitrate (NO3), an abundant aerosol in polar snow, is a complex environmental proxy to interpret owing to the variety of its sources and its susceptibility to post-depositional processes. During the last glacial period, when the dust level in the Antarctic atmosphere was higher than today by a factor up to ~25, mineral dust appears to have a stabilizing effect on the NO3 concentration. However, the exact mechanism remains unclear. Here, we present new and highly resolved records of NO3 and 15 non-sea salt calcium (nssCa, a proxy for mineral dust) from the Roosevelt Island Climate Evolution (RICE) ice core for the period 26 40 kilo years Before Present (ka BP). This interval includes seven millennial-scale Antarctic Isotope Maxima (AIM) events, against the background of a glacial climate state. We observe a significant correlation between NO3 and nssCa over this period and especially during AIM events. We put our observation into a spatial context by comparing the records to existing data from east Antarctic cores of EPICA Dome C (EDC), Vostok and central Dome Fuji. The data suggest that nssCa 20 is contributing to the effective scavenging of NO3 from the atmosphere through the formation of Ca(NO3)2. The geographic pattern implies that the process of Ca(NO3)2 formation occurs during the long-distance transport of mineral dust from the midlatitude source regions by Southern Hemisphere Westerly Winds (SHWW) and most likely over the Southern Ocean. Since NO3 is dust-bound and the level of dust mobilized through AIM events is mainly regulated by the latitudinal position of SHWW, we suggest that NO3 may also have the potential to provide insights into paleo-westerly wind pattern during the 25 events.


Introduction
Nitrate (NO 3 -), the end product of oxidation of nitrogen oxides (NO x = NO+NO 2 ) in the atmosphere, is one of the anions widely measured in Antarctic ice cores (Legrand and Mayewski, 1997;Wolff, 2013). A range of sources relevant to Antarctica has been identified for the production and release of NO x species, such as a) oxidation of N 2 O and the photo-dissociation and 30 https://doi.org/10.5194/cp-2020-151 Preprint. Discussion started: 10 December 2020 c Author(s) 2020. CC BY 4.0 License. place in the atmosphere, then dust most likely enhances NO 3 scavenging (Rothlisberger et al., 2000). Rothlisberger et al. 65 (2000) recommended that additional records for both mineral dust and NO 3 are required to resolve the role of dust in the accumulation of NO 3 during the glacial.
In this context, we present a new, well-dated and highly resolved glacial record of NO 3 and Ca 2+ from Roosevelt between the samples on a ratio of 5:1 (sample: standard) The analytical precision is better than ~5 % for Ca 2+ and better than ~10% for NO 3 for the IC samples measured within the bracketing time period . A likely reason for the reduced precision observed for NO 3 could be due to the influence of extremely small contamination from drill fluid, a mixture of Estisol-240 and Coasol (Bertler et al., 2018), as the section of the core lies in the brittle ice zone (500 -764m) that is susceptible to internal fractures (Pyne et al., 2018). 100

01
(1) where R c and R m are the crustal and marine ratio of Ca 2+ /Na + (Lambert et al., 2012) and with Ca 2+ and Na + measured from IC.
We have used the traditional values of R c and R m , which are 1.78 and 0.038, respectively (Bowen 1979), instead of 1.06 and 105 0.043 as used for nssCa 2+ calculation in EDC (Lambert et al., 2012). This is because sea salt aerosols deposited at RICE originate from open-ocean sources and are formed through mechanisms like bubble-bursting (Winstrup et al., 2019), as opposed to sites in central East Antarctica (e.g. EDC), where sea ice surfaces are identified as the major source of such aerosols (Wolff et al., 2006). Hence, traditional R m value shall be more appropriate. Also, the trajectory modelling for dust transport to RICE shows a mixture of sources such as south of South America, Australia and New Zealand (Neff and Bertler, 2015), which 110 suggests using a general value of R m may better represent the crustal ratio. However, it is also to be noted that changing R c and R m from traditional values has only a very minimal effect on nssCa 2+ calculation (mean difference of ~0.5ppb), as documented in numerous other studies (Bigler et al., 2006;Lambert et al., 2012).

Principal Component Analysis 115
Principal Component Analysis (PCA) is a multi-variate statistical technique used for data reduction and development of multiparameter proxies of climate indicators (Buizert et al., 2018;Lambert et al., 2012). PCA analysis has been performed on the whole data set and is carried out using MATLAB algorithm, pca. Prior to the analysis, outliers are removed, data are averaged to 50-years to achieve equal sample spacing and are then detrended.

Wavelet Coherence and Cross Spectrum Analyses 120
Wavelet coherence is a measure of correlation between two variables ( NO 3 and nssCa 2+ here) in a time-frequency plane and is computed using analytic Morelet wavelet in MATLAB (Grinsted et al., 2004). The computed coherence is expressed as magnitude-squared coherence (msc). Wavelet cross spectrum (WCS) analysis is performed to identify the relative lead/lag between NO 3 and nssCa 2+ and the phase of WCS is provided for the values higher than 0.6 msc.

Temporal Variability of nssCa 2+ and NO 3 -
We compare the changes in nssCa 2+ and NO 3 to RICE δD, a proxy for isotope temperature reconstructions (Dansgaard, 1964), between 26-40 ka BP (Fig. 2). Temperature records for the last glacial period, especially for marine isotope stage (MIS) 3 in Similarly, NO 3 displays concentration decrease during AIM events (Fig. 2c). However, during AIM 6, we also observe a brief rise in the NO 3 concentration after the first drop, followed by its decline. Likewise in nssCa 2+ , NO 3 record also includes two large peaks of centennial duration around 33.3 ka BP and 31.5 ka BP, respectively (black circles on Fig. 2c).
The evolution of NO 3 over this period also showcase a shift to higher concentration after 31.5 ka BP (~ 22.5 to 24.4 ppb) (grey line on Fig. 2c), which imparts a significant increasing trend to the record towards MIS2/LGM. 150

Wavelet coherence and cross spectral analysis of NO 3 and nssCa 2+
The statistically significant coupling between NO 3 and nssCa 2+ observed in their respective time series on millennial and centennial scales are also evident in the wavelet coherence and WCS analysis ( Fig. 3a, b). A significant correlation is observed in the frequency band of ~1500 -3000 years between ~39 -28 ka BP and within ~500 -700 years between 30 -35 ka BP. In 155 the latter, centennial frequency band, WCS analysis also reveals an in-phase relationship (Fig. 3b).
To understand the complex relationship between NO 3 and nssCa 2+ , a principal component analysis is performed. The full data set consists of Na + , Cl -, Mg 2+ , nssCa 2+ , K + , SO 4 2-, NO 3 and MSA concentrations. The first three principal components together 160 explain ~95 % variance of the considered ions (Table 1). The first principal component (PC1) is dominated by the most common sea salt species, Na + and Cl -, and accounts for 77.67 % variance of the data. The higher variability of sea salt aerosols in RICE is not a surprise as it is a low-elevation coastal site and, as such, will be very sensitive to any changes in the spatial with high factor loadings (0.84 and 0.45, respectively). Most of the variance of these two species are explained in this PC. As NO 3 and nssCa 2+ has completely different sources, this PC is most likely representing identical depositional mechanisms.

Role of mineral dust in the NO 3 accumulation during the glacial period
Mineral dust, represented by the nssCa 2+ , mostly originates from continents outside of Antarctica and is introduced into the Antarctic atmosphere by SHWW (Fischer et al., 2007b;Lambert et al., 2012;Röthlisberger et al., 2002). The strong association of nssCa 2+ and NO 3 observed in the time series, wavelet coherence and through their covariance in PC3, raises the question: Is nssCa 2+ an effective scavenging agent for the deposition of NO 3 or is the prevention of post-depositional leakage of NO 3 -180 from the snowpack is the dominant influence? Rothlisberger et al. (2000) noted that if NO 3 formation takes place in the atmosphere, a strong correlation between nssCa 2+ and NO 3 is expected and is to be observed across the continent, as extracontinental dust is well-mixed in the Antarctic atmosphere by SHWW.
To test this hypothesis, we compare glacial nssCa 2+ and NO 3 records from other Antarctic locations ( Fig. 1). Just as it is the case at our west Antarctic site RICE, EDC (Rothlisberger et al., 2000), Vostok (Legrand et al., 1999) and Dome Fuji 185 (Watanabe et al., 1999), three east Antarctic sites, also document a similar close association between NO 3 and nssCa 2+ (Fig.   4). Glacial accumulation rate is relatively higher at RICE (~11cmyr -1 ) (Lee et al., 2020), in comparison to the East Antarctic sites, where the rate of accumulation is understood to be very low during the glacial period (~1.4 cmyr -1 ) (EPICA Community Members, 2006). Based on a global chemical transport modelling, Zatko et al. (2016) examined the impact of the snow accumulation rate on NO 3 loss, recycling and redistribution across Antarctica. Modelled NO 3 loss and recycling is very 190 significant (up to ~ 95% and factor of 8, respectively) in the low accumulation polar plateau due to photolysis. This suggests that at Antarctic locations such as EDC, Vostok and Dome F, HNO 3 deposited on the snow surface are most likely to be lost https://doi.org/10.5194/cp-2020-151 Preprint. Discussion started: 10 December 2020 c Author(s) 2020. CC BY 4.0 License. witout diffusing into the snow and reacting with nssCa 2+ present inside. Whereas, the photolytic NO 3 loss is minimal at the coastal Antarctic locations like RICE (5 -55 %). Thus, as the NO 3 and nssCa 2+ association is found active at sites both in East and West Antarctica that are spatially well-dispersed, with different rates of accumulation and hence preservation potentials, 195 we propose that nssCa 2+ is scavenging NO 3 from the atmosphere and depositing it as Ca(NO 3 ) 2 in the snowpack.
We also note that overall during the glacial period, the hydrological cycle was weaker, continental shelves were more exposed and glacier discharges into outwash plains were higher, providing conditions conducive to increased dispersion of dust from the source regions (Fischer et al., 2007a;Lambert et al., 2012). Such dustier conditions can lead to more efficient scavenging of NO 3 -. Studies have also shown that the lifetime of HNO 3 in the atmosphere through the dust removal can be 200 well limited to just two days during dustier periods like glacial times (Hanish and Crowley, 2001), re-affirming its effectiveness in the scavenging process. As SHWW shift southward, away from these source regions in the mid-latitudes, AIM events are characterized by a reduced nssCa 2+ concentration in the Antarctic ice core records (Fischer et al., 2007a;Lambert et al., 2012). The southward displacement of SHWW may also change the precipitation pattern in the predominant source regions such as Patagonia by reducing its strength (Boex et al., 2013;Moreno et al., 2012). Furthermore, in response to the atmospheric reorganization, oceanic fronts re-arrange and shift southwards during AIM events, bringing warmer subtropical waters to the mid-latitudes 220 (Barker et al., 2009;De Deckker et al., 2012). This causes sea surface temperature in the mid-latitudes to rise (Barker et al., 2009;Lamy et al., 2007), which then could trigger regional warming (Boex et al., 2013). This warming when combined with reduced precipitation in prominent source areas like Patagonia, can possibly lead to glacier melting (Boex et al., 2013) and formation of pro-glacial lakes which can act as dust-traps in such areas (Sugden et al., 2009), and also facilitate the vegetation growth on surfaces previously covered by snow, thereby stabilizing them. Hence, these feedback processes can further reduce 225 the dust-mobilization during the events. https://doi.org/10.5194/cp-2020-151 Preprint. Discussion started: 10 December 2020 c Author(s) 2020. CC BY 4.0 License. RICE also captures a decrease in nssCa 2+ level broadly during all AIM events. However, during AIM 6 and AIM 3, we also observe centennial-scale peaks within the drop. To put our data into a wider context, we compared our records to the available EDC and EPICA Dronning Maud Land (EDML) data (Fig. S2). And during AIM 3, we observe a similar pattern at both the sites. The increase in nssCa 2+ flux after the initial drop is observed to be approximately twice the magnitude of the fall at both 230 the locations. Fogwill et al. (2015) suggest that the obliquity minimum around ~28.5 ka BP may have resulted in circum-Antarctic sea ice growth, influencing SHWW and forcing it to shift northward with higher speed. This might explain the centennial-scale jump in nssCa 2+ concentration after the early drop during AIM 3. SHWW is likely to re-adjust later as the stadial conditions continue to be prevalent in the North Atlantic (Buizert et al., 2018) forcing SHWW to shift poleward, which could describe the following fall back in the concentration. During AIM 6, the pattern of the rise in concentration after the 235 initial reduction observed in RICE is not visible at EDC or EDML. This questions the continent-wide representation of this dust signal and perhaps even suggest that its appearance is limited only to the Ross Sea region/West Antarctica.

Potential of NO 3 as an additional paleo-westerly wind proxy?
Experimental studies show that amongst calcium-containing minerals, carbonates such as calcite (CaCO 3 ) and dolomite 240 (CaMg(CO3) 2 ) are the most reactive in the presence of nitric acid (HNO 3 ) in the atmosphere (Gibson et al., 2006;Krueger et al., 2004;Vlasenko et al., 2006). The reaction can lead to the formation of calcium nitrate (Ca(NO 3 ) 2 ) following reactions (2) and (3) below (Gibson et al., 2006;Krueger et al., 2004). In addition to the reaction with HNO 3 , CaCO 3 and CaMg(CO3) 2 can also react with nitrogen pentoxide (N 2 O 5 ) to form Ca(NO 3 ) 2 given by reactions (4) and (5)  The mineral composition of both southern South American and Australian dust aerosols show the presence of calciumcontaining minerals such as calcite and Ca-rich plagioclase (Radhi et al., 2011;Zárate, 2003), and thus the reactions (2)-(5) 255 could occur in the atmosphere. Furthermore, the reactions (2)-(5) accelerate when the relative humidity increases, encouraging preferentially Ca(NO 3 ) 2 formation because of the hygroscopic nature of Ca(NO 3 ) 2 (Krueger et al., 2004;Mahalinganathan and Thamban, 2016;Vlasenko et al., 2006). Due to Antarctic's characteristic low relative humidity and shallow atmospheric boundary layer, it is less likely for neutralization reactions to occur locally (Mahalinganathan and Thamban, 2016).
Investigation of nssCa 2+ / NO 3 association in the snow pit samples of coastal and inland East Antarctica suggests that binding 260 https://doi.org/10.5194/cp-2020-151 Preprint. Discussion started: 10 December 2020 c Author(s) 2020. CC BY 4.0 License. reactions most likely occur over the Southern Ocean during the long-range transport of nssCa 2+ due to a higher relative humidity over the region (Mahalinganathan and Thamban, 2016). During the glacial period, a time when relative humidity in the Antarctic atmosphere is significantly decreased compared to present day, we propose that neutralization reactions between mineral dust and NO 3 most likely occurred over the Southern Ocean. This suggests a possible link to long-range transport of nssCa 2+ and NO 3 by SHWW and also implies that the NO 3 signal preserved through the glacial period is the dust-bound NO 3 -265 dispersed by the westerly winds to the Antarctic atmosphere.
The amount of extra-Antarctic dust reaching Antarctica from the mid-latitude source regions during AIM events is primarily controlled by the zonal position of SHWW (Lambert et al., 2012). As NO 3 is dust bound and the variability of the dust concentrations across AIM events depends on the latitudinal movement of SHWW, it implies that NO 3 concentration is indirectly regulated by SHWW. This opens-up the possibility of utilizing NO 3 as a westerly wind proxy for the AIM events. 270 During AIM events, NO 3 exhibits a reduction in concentration similar to that of nssCa 2+ (Fig. 2b, c). In addition to the congruent response we observe in the time series, the wavelet analysis also reveals statistically significant coherence among these species in the periodicity of AIM events (~2000 years) (Fig. 3a, b). To our observation into wider context, we assessed NO 3 and nssCa 2+ records from EDC and Dome F across AIM events (numbered in Fig. 4). The identical response of both species is evident in both the cores during large events (AIM 4 and AIM 8) (Fig. 4c-f). The relationship can also be identified 275 for smaller events (AIM 3, 5.1, 5.2, 6,7) in EDC (Fig. 4c, d). However, at Dome F, lower temporal resolution (~200 years) act as a hindrance to identify the smaller events (duration ~1000 years). Moreover, the wavelet analysis also shows significant coherence between NO 3 and nssCa 2+ in the frequency domain of AIM events at both the locations (Fig. 5a, b). This shows that dust-NO 3 coupling identified during AIM events can also equally be observed in other regions of Antarctica which are located distantly from RICE (e.g. EDC, Dome F). This indicates a continent-wide validity for this relationship, which also strengthens 280 our argument of using NO 3 as a westerly wind proxy during AIM events.
To investigate the phasing between nssCa 2+ and NO 3 records at different sites, we performed WCS analysis. In RICE, the nssCa 2+ shows a lead of ~100 -200 years in the frequency window of AIM events (~2000 years) over the whole period (~28 -39 ka BP) (Fig. 3b). Whereas at Dome F, the records are in-phase from ~39 -34 ka BP (Fig. 5b). While, starting from ~34 ka BP, NO 3 shows a lead of ~100 -250 years. On the other hand, at EDC, both the records show an in-phase relationship 285 over the whole interval (~26 -39 ka BP) (Fig. 5a). This observation of spatially-varying phase relationship is intriguing.
However, resolving it would require more detailed investigation perhaps on the of site characteristics, such as distance between the site and dust source and therefore the transport time for the aerosols or difference in the residence time along the transport route.

Conclusion
The correlation between NO 3 and nssCa 2+ in the RICE record provides additional evidence of the role of calcium-containing mineral dust on the scavenging of NO 3 through the formation of Ca(NO 3 ) 2 . This process is particularly important during the glacial period when the dust load is higher by a factor of up to ~25 than today. This association has been observed previously https://doi.org/10.5194/cp-2020-151 Preprint. Discussion started: 10 December 2020 c Author(s) 2020. CC BY 4.0 License.
in the east Antarctic cores of EDC (Rothlisberger et al., 2000), Vostok (Legrand et al., 1999) and central Dome Fuji (Watanabe 295 et al., 1999). Here, we show the same relationship for the first time for a coastal site in West Antarctica, thereby showing it is an Antarctic-wide phenomenon during the glacial. We conclude that NO 3 scavenging by mineral dust in the atmosphere is the dominant mechanism for the NO 3 -/nssCa 2+ correlation observed in ice cores and binding reactions most likely occur during the long-range transport from mid-latitudes to Antarctica by SHWW. During AIM events, the NO 3 variability depends on the level of mineral dust mobilized from mid-latitude source regions, which inturn influenced by the latitudinal position of SHWW. 300 For this reason, we also suggest that NO 3 may be a useful paleo-westerly proxy for the AIM events of the glacial period.

Data Availability
The following data are available in the supplementary material of the manuscript. RICE Ca 2+ (CFA), NO 3 -(IC), nssCa 2+ (IC) and the other Antarctic ice core data sets used. The RICE data will also be made available at PANGEA Data Publisher 305 (ww.pangea.de) upon the acceptance for publication.    https://doi.org/10.5194/cp-2020-151 Preprint. Discussion started: 10 December 2020 c Author(s) 2020. CC BY 4.0 License.