Assessing the continuity of the blue ice climate record at Patriot Hills, Horseshoe Valley, West Antarctica

We use high‐resolution ground‐penetrating radar (GPR) to assess the continuity of the Blue Ice Area (BIA) horizontal climate record at Patriot Hills, Horseshoe Valley, West Antarctica. The sequence contains three pronounced changes in deuterium isotopic values at ~18 cal ka, ~12 cal ka, and ~8 cal ka. GPR surveys along the climate sequence reveal continuous, conformable dipping isochrones, separated by two unconformities in the isochrone layers, which correlate with the two older deuterium shifts. We interpret these unconformities as discontinuities in the sequence, rather than direct measures of climate change. Ice sheet models and Internal Layer Continuity Index plots suggest that the unconformities represent periods of erosion occurring, as the former ice surface was scoured by katabatic winds in front of mountains at the head of Horseshoe Valley. This study demonstrates the importance of high‐resolution GPR surveys for investigating both paleoflow dynamics and interpreting BIA climate records.


Introduction
With a capacity to resolve internal layering within ice, ground-penetrating radar (GPR) has transformed our ability to study and interpret historic changes in ice flow [Paren and Robin, 1975;Daniels et al., 1988;Fujita et al., 1999;Rippin et al., 2003Rippin et al., , 2006Woodward and King, 2009;Sime et al., 2011;Drews et al., 2013]. Despite this, there is limited analysis of the detailed internal structure of Blue Ice Areas (BIAs), which are estimated to cover 120,000 km 2 (~0.8%) of the Antarctic continent [Winther et al., 2001]. This is perhaps a function of the reduced performance of conventional snowmobile towed GPR surveys in these areas [Spaulding et al., 2013;Turney et al., 2013], where the speed of travel results in a reduced scan rate relative to the distance traveled, which reduces the ability to image the detailed internal strata of BIAs. Defined as regions of exposed ice with a relatively low surface albedo [Bintanja, 1999], BIAs typically form on the leeward foreground of mountain ranges, where upward ice flow around the mountains and/or into the mountain front compensates for surface ablation (similar to erosioninduced bedrock uplift in mountains). This allows deeper, older ice to rise toward the surface where it is exposed, typically as a rippled blue ice surface [Bintanja, 1999;Sinisalo and Moore, 2010;Fogwill et al., 2012;Campbell et al., 2013]. This phenomenon enables old ice to be exposed, enabling "ice sequence" climate records to be collected along the BIA surface [Whillans and Cassidy, 1983;Korotkikh et al., 2011;Fogwill et al., 2012;Spaulding et al., 2012;Spaulding et al., 2013;Turney et al., 2013]. So-called "horizontal coring" offers considerable logistical benefits over vertical coring, although such climate records require careful interpretation, as the processes that have brought packages of ice to the surface may impact upon their continuity and therefore their paleo significance.

Ground-Penetrating Radar
A PulseEKKO 1000 GPR system was used to generate a 200 MHz GPR profile along a central BIA transect, extending perpendicular to Patriot Hills for 800 m, along the climate record (Transect A, Figure 1). To obtain a high-resolution GPR profile, we employed continuous step-and-collect mode with a 7000 ns time window and an in-field stack of 8. The GPR data were collected at 0.1 m intervals with copolarized antennae orientated perpendicular to the survey line, with their broadsides parallel to each other. This time-intensive method is described in detail by Woodward et al. [2001]. A further nested grid of high-frequency lines (approximately 7 × 9 km with 1 km × 1.5 km grid cells) extending from the BIA margin ( Figure 1) was also surveyed in 2014 by towing the sledge-mounted PulseEKKO 1000 system by snowmobile at approximately 12 km/h along each transect line with no in-field stacking. This mode of operation is much faster than step-and-collect mode, allowing a larger area to be surveyed, albeit at a reduced resolution. Each line was surveyed for topographic correction using a Trimble differential GPS unit and corrected to decimeter accuracy using a local base station. GPR data were processed in Reflexw [Sandmeier Scientific Software, 2012], version 6.1.1., using standard processing steps [Welch and Jacobel, 2005;Woodward and King, 2009;King, 2011]. These steps include time-zero correction, background removal, high-pass frequency filtering (Dewow), band-pass filtering, and diffraction-stack migration. An energy-decay gain was also applied. For display purposes depth and topographic corrections were applied using an ice velocity of 0.168 m ns À1 . Applying this standard velocity underestimates the depth of firn layers away from the BIA.

Ice Sheet Model Simulations
Preexisting ice sheet model perturbation experiments [Golledge et al., 2012;Fogwill et al., 2014] were used to investigate ice flux and ice flow direction in Horseshoe Valley during the Holocene. The Parallel Ice Sheet Model (PISM) is a three-dimensional, thermomechanical, continental ice sheet model that combines shallow-ice and shallow-shelf approximation equations in order to simulate the dynamic behavior of grounded ice, floating ice, and ice streams. Model runs used proxy-based interpretations of atmospheric [Petit et al., 1999] and oceanic [Lisiecki and Raymo, 2005;Imbrie and McIntyre, 2006] changes during the last glacial cycle and employ boundary conditions from modified Bedmap topography , as well as a spatially varying geothermal heat flux interpolation [Shapiro and Ritzwoller, 2004]. Our perturbation experiments were run at a resolution of 5 km, starting from a Last Glacial Maximum (LGM) (occurring sometime between 29 and 33 ka in West Antarctica) [Clark et al., 2009] configuration [Golledge et al., 2012]. Additional details on the PISM model runs are available in Fogwill et al.

Internal Layer Continuity Index Plots
An Internal Layering Continuity Index (ILCI), derived from airborne radio echo sounding (RES) of the upper Institute Ice Stream catchment [Winter et al., 2015], was employed to characterize the internal stratigraphy of ice within Horseshoe Valley (using 100 trace moving windows), at predefined depth intervals of 0-20% (uppermost ice column), 40-60%, and 80-100% ice thickness. Developed by Karlsson et al. [2012] and recently applied to the Institute Ice Stream catchment by Bingham et al. [2015] and Winter et al. [2015], the ILCI uses relative changes in reflected radar power to assess the continuity of internal layers within the ice; this can provide insight into ice-flow history . Areas of high-reflected radar power, bounded by values of lower reflected relative power are recorded in A-scope plots of each RES trace (where each trace represents a stack of 10 consecutive raw traces to reduce noise) [Karlsson et al., 2012]. This allows areas of continuous internal layering to return a high ILCI (0.06-0.10), while absent and disrupted layers return a low ILCI (>0.06). These low to intermediate ILCI values have been interpreted to represent areas that have previously encountered or are currently experiencing enhanced flow (defined in this region as >30 m a À1 [Winter et al., 2015]). Following Winter et al. [2015], we specify the term "enhanced flow" as distinct from the term "fast flow," as the latter term is often equated with more extreme ice speeds in ice streams.

Ground-Penetrating Radar
GPR identified the following features in Patriot Hills BIA: blue ice with conformable steeply dipping internal stratigraphy; two pronounced divergent isochrones, associated with truncated layers; and blue ice that lacks strong internal stratigraphy at the start of transect A and profiles Y1-Y8 (Figures 2 and 3). The radar grid also shows a variety of features in the firn zone including truncated firn layers, prograding bedding sequences, surfaceconformable stratigraphy, unconformities, firn that exhibits convergence, and surface snow drifts (Figure 3).  (Figure 2c). These discontinuities, associated with the truncation of isochrones, correlate to rapid changes in the trend of the deuterium isotopic record (δD) at approximately 18 cal ka and 12 cal ka [Turney et al., 2013]. B1 marks the transition from a low average δD rate to a rising trend in δD concentrations, where δD increases from À380 to À254‰. B2 marks a very rapid rise in δD concentrations from À300 to À254‰, after which a higher average ratio continues for the remainder of the profile. It has been suggested by Turney et al. [2013] that these changes, highlighted by shaded bands B1 and B2 in Figure 2d, could reflect significant changes in temperature and/or precipitation during both the late Pleistocene and Holocene [Turney et al., 2013]. There is, however, no evidence of divergent or truncated isochrones at any other location along the profile, even at B3 (~8 cal ka), where a depletion in deuterium isotope content is recorded.
Examples from the snowmobile-towed GPR grid, collected for wider analysis of the BIA and firn, are displayed in Figure 3 (inline Transects Y5 and Y7). Unlike GPR in step-and collect mode we detect limited internal features within the BIA using this method. However, numerous internal horizons are identified at the BIA/firn margin where a net upward ice flow component dominates the radargrams, with compressed isochrones inclined to a maximum dip angle of 5°. Each inline profile displays sequences of convergent and prograding isochrones within the firn zone which can be matched laterally between transects. An erosional unconformity is revealed in profile Y7 (Figure 3) where gently sloping (2°apparent dip toward Patriot Hills) internal horizons are overlain by younger, near horizontal firn layers between 2690 m and 3149 m along the transect, which more than double in thickness with increasing distance from Patriot Hills. A shallow snow drift is also visible in profile Y5 (Figure 3), near the BIA/firn margin where the 9 m thick drift extends 440 m along the former, near horizontal firn surface.

Ice Sheet Model Simulations
Simulated regional ice flux models show the initial response of the LGM ice sheet to ocean and atmospheric forcing where high discharge rates are simulated through all the major troughs, although no major ice flux or flow direction change is modeled in Horseshoe Valley (Figure 4a). With a rapid increase in ice flux in response to ocean

Discussion
Our GPR transects, ice sheet model simulations, and ILCI analyses each contribute to our understanding of ice sheet flow history in Horseshoe Valley and help to constrain the evolution of Patriot Hills BIA. Analysis of highresolution GPR-detected internal stratigraphy reveals largely conformable isochrones which are inclined toward Patriot Hills BIA surface. Minor changes in the dip angle of the predominantly parallel internal horizons within the BIA do occur and are expected as a result of differential snow deposition, burial, and subsequent ice flow over time, but the pronounced changes in dip angles at D1 and D2 (Figure 2) represent larger scale change. These discontinuities correspond to abrupt shifts in the local climate record between~18 cal ka (B1) and~12 cal ka (B2) ( Figure 2) and therefore represent breaks in an otherwise largely unbroken 30,000 year climate record. These breaks, given new context by the unconformities in GPR Transect A could have formed by one of two mechanisms: (1) changes in ice flowline trajectory or (2) local interaction of topography, snow accumulation, and wind.
Ice sheet model simulations and ILCI analysis suggest that ice in Horseshoe Valley has not experienced directional change (Figure 4) and has remained slow flowing ( Figure 5) since the mid-Holocene. These findings eliminate the possibility that discontinuities D1 and D2 were formed by changes in ice flow-line trajectory but do not rule out significant periods of erosion. Periods of erosion could have resulted from the interaction of topography, snow accumulation, and wind as the ice flowed from the head of Horseshoe Valley toward Patriot Hills (Figure 6). We therefore expect that discontinuities D1 and D2, corresponding to changes in deuterium isotope concentrations at B1 and B2, were created by localized katabatic wind scour of the former snow and ice surface, as ice flowed through BIAs in front of Liberty and Marble Hills ( Figure 6). Consequently, it seems probable that B1 and B2 do not directly represent abrupt climatic changes. As no other erosional events are found in the GPR record, it is assumed that other inferred depletions in the deuterium isotopes, such as that at B3, could reflect direct climatic changes during the early Holocene and indeed may correlate with changes in other ice cores [Turney et al., 2013].
Our findings from the extended radar grid are in close agreement with the high-resolution BIA transect. Here the inline profiles show more recent periods of BIA stability and instability, reflected by convergent and prograding isochrones in the firn zone. Prograding isochrones in the GPR record ( Figure 3) can be attributed to increased katabatic wind scour, and subsequent BIA expansion since the LGM. This is likely the result of surface lowering in Horseshoe Valley of up to~480 m since the LGM [Bentley et al., 2010], which would have revealed more of the nunataks in the Southern Heritage Range, capable of promoting stronger katabatic wind scour. In contrast, younger convergent isochrones in the GPR record ( Figure 3) represent more stable meteorological conditions, where katabatic winds of consistent velocity and direction have produced a transition zone between all annual snowfalls to no snowfall scoured. If these transition zones are in the same location annually, convergent layering will result. This also requires slow and stable ice sheet flow. These sequences of BIA growth and stabilization combine to identify an evolving BIA over the past~1000 years, which is consistent with the previously analyzed 30,000 year ice flow records. The unconformable surface firn in profile Y7 and the snow drift in profile Y5 ( Figure 3) have anthropogenic origins which are attributed to the recent movement of snow to create Patriot Hills Antarctic Logistics and Expeditions Base Camp (seasonally occupied between 1987 and 2010).

Conclusions
Radar-detected stratigraphic relationships analyzed in conjunction with deuterium isotope records, ice sheet model simulations, and internal layer continuity analysis at the Patriot Hills Blue Ice Area (BIA), West Antarctica, indicate the following: (1) stable periods of snow accumulation and ice flow have been interrupted by episodes of significant erosion, which have resulted in unconformities within the otherwise conformable stratigraphic record and (2) the current trajectory of ice flowing toward Patriot Hills BIA is, in essence, unchanged over the historical record. We conclude that deuterium isotope records from Patriot Hills BIA reflect conditions in Horseshoe Valley (and the West Antarctic Ice Sheet) over at least the last 30,000 years, though due consideration must be taken around the two periods of differential wind scour.
Importantly, this study also demonstrates the considerable value of using GPR in step-and-collect-mode to interpret ice sheet history from BIAs, as conventional snowmobile towed GPR cannot resolve the detailed internal structure of these ice features. This finding is particularly relevant to the climate community, as low-cost and portable GPR surveys in step-and-collect mode can greatly improve the reliability of relatively easily accessible horizontal climate records.