Remote sensing

The location of macro-relief features (glazed surfaces, transverse dunes, megadunes, glazed-surface sastrugi fields) and their direction and extent were determined by remote-sensing analysis.

In this study we have integrated field surveys with satellite images of Landsat 4 and 5 TM(66/113, 20 February 1992;72/113,13 January 1992;78/113,15 January 1992; 81/113, 20 January 1992; 84/113,17 January 1992), Landsat 7 Enhanced Thematic Mapper Plus (ETM+) (81/114 and 81/115, 2 January 2000) and U.S. National Oceanic and Atmospheric Administration (NOAA) 12–14 AVHRR (November 1994– November 1999). The Landsat TM and ETM+ have a pixel resolution of 30m and seven spectral bands: three in the visible wavelength region (bands 1–3), three in the near-infrared wavelength region (bands 4, 5 and 7) and one region of thermal infrared wavelength (band 6). The Landsat 7 ETM+ has panchromatic band with a higher pixel resolution of 15 m. Landsat images were georeferenced by means of geographical coordinates derived from the satellite ephemeris; the geolocation of ETM+ has a 1σ error of 50m (personal communication from R. Bindschadler, 2000). The AVHRR images were collected by the high-resolution picture transmission (HRPT) receiving station installed at TNB. AVHRR sensors installed on NOAA’s satellite have a pixel resolution of 1.1km at nadir in five spectral bands: visible red (band 1), near-infrared (band 2), middle infrared (band 3) and thermal infrared (bands 4 and 5). More than 200 AVHRR images from NOAA 12–14 meteorological satellites, with 530% cloud cover over the plateau, were analyzed, having been selected from those acquired by the HRPT station from 1994 to 1999. In this study we used the images acquired by NOAA 14 on 16 November 1995 (0505 h GMT; solar azimuth 308°; solar elevation 30°). The ephemeris-based geolocation is accurate to only about ±3 km (Reference Fahnestock, Scambos, Bindschadler and KvaranFahnestock and others, 2000a), so the re-projection was performed using the Antarctic Digital Data Base (BAS and others, 1993) and DEM provided from Reference Rémy, Shaeffer and LegrésyRémy and others (1999). The projection used is polar stereographic from the World Geodetic System (WGS84) ellipsoid, with the plane of projection parallel to 71˚ S.

The satellite images were analyzed, compared and organized into a geographic information system (GIS) using ERDAS, TerraScan and ARCINFO software. Image processing included routine procedures such as radiometric corrections, noise and striping removal, and application of special linear stretches of individual bands after inspection of grey-value histograms to enhance surface features and characteristics.

Field survey

A spatial distribution survey of the micro-relief surface type, size and orientation was conducted along the route during the austral spring–summer (November–December). The spring season provided the best opportunity to investigate micro-relief development and distribution, because the observed micro-relief formed during winter, when most annual precipitation and the strongest katabatic surface winds occur (Reference GoodwinGoodwin, 1990). The roughness of the snow surface, as well as the orientation and size of the surface relief, was measured over 500m × 500m quadrates at intervals of 5 km along the traverse and continuously along some profiles in the megadune area. The micro-relief features observed were classified according to the system described by Reference Fujiwara, Endo and MurayamaFujiwara and Endo (1971) and applied to Japanese and Australian Antarctic Research Expedition (JARE and ANARE) traverses inland in the plateau region of Dronning Maud Land and Wilkes Land (Reference GoodwinGoodwin, 1990). The mean azimuths of the micro-relief were measured by magnetic compass and converted into true bearings by GPS, and the mean height of the micro-relief was recorded using a ruler. The surface micro-relief observed is divided into three types (Reference WatanabeWatanabe, 1978; Reference GoodwinGoodwin, 1990):

In the first 400 km, the 1998/99 data show the same general orientation in sastrugi directions and types that were noted during the 1996 expedition.

Surface elevation profiles and local topography along the traverse TNB–DC was measured by global positioning system (GPS) (Reference Urbini, Gandolfi and VittuariUrbini and other, 2001), whereas regional surface topography was analyzed using a DEM of Antarctica with digitized element resolution of 1km, provided by Reference Rémy, Shaeffer and LegrésyRémy and others (1999). A 363 pixel window (pixel size: 1×1km) was used to calculate the slope of each pixel using the DEM. A GPS survey was performed along the traverse using two master stations located at the beginning (TNB) and at the end of the traverse (DC). During the megadune survey a master station at D6 (site of field camp) was used. Data acquired from master stations were utilized in post-processing to obtain an accurate location of the entire track (Reference Urbini, Gandolfi and VittuariUrbini and others, 2001). The mobile receivers, equipped with geodetic antennas, were installed on vehicles and were used for the kinematic surveys to perform altimetric profiles and to correct ground-penetrating radar (GPR) acquisition as a function of both ellipsoidal height and surface coordinates. The accuracy of the altimetric profile along the traverse is mainly due to the distance between the master and rover stations and varies between <1m and 3 m at the farthest point. However, the accuracy of the altimetric profile performed at the megadune area, using the D6 master station, is up to 10 cm. Vehicle and thus GPS antenna speed oscillated from about 8 to 12 kmh^–1. The sampling rate of the GPS receivers (master and rover) was fixed at 5 s; meaning about one coordinate every 10–15m.

GPR along continuous profiles provides detailed information on spatial variability in snow accumulation (Reference Richardson, Aarholt, Hamran, Holmlund and IsakssonRichardson and others, 1997; Reference Richardson and HolmlundRichardson and Holmlund, 1999; Reference Urbini, Gandolfi and VittuariUrbini and others, 2001). Data acquisition was performed along the TNB–DC traverse with a GSSI SIR10B unit equipped with one monostatic antenna with a central frequency of 400 MHz. The mainframe unit was mounted inside the vehicle (Pisten Bully 330D) cabin together with a geodetic GPS instrument, while the antenna was pulled on a small wooden sledge. Principal acquisition parameters were 150 ns for the vertical investigation range and 1–5 scan s^–1 for the acquisition rate (15–20m). Vehicle and thus antenna speed oscillated from about 8 to 12 kmh^–1, meaning about one scan every 2–3m (with the acquisition rate at 1 scan s^–1) and 0.4 to 0.7 m (with the acquisition rate at 5 scan s^–1). For electromagnetic wave-speed calculations the depth–density relation for the snowpack was established using the density profile of 20 firn cores (12–52mdeep), retrieved with an electromechanical drilling system (diameter 100 mm). The integration of GPS and GPR provided the ellipsoidal height of both the topographical surface and firn stratigraphy. Reference Urbini, Gandolfi and VittuariUrbini andothers (2001) describe the GPR methods in detail. In line with Reference Vaughan, Corr, Doake and WaddingtonVaughan and others (1999), we assume that layers that produced a strong radar reflection are isochronous.

Aeolian distribution

Landsat TM image analysis showed a lower albedo or reflectivity of glazed surfaces compared with snow and blue ice (Fig. 2). Firn and ice albedos generally decrease passing from the visible to the near-infrared wavelengths (e.g. Reference WarrenWarren, 1982), whereas glazed surfaces present spectra reflectance intermediate between snow and ice. The spectral difference between snow and glazed surfaces makes it simple to identify these areas on false-colour image composite TM images with bands 2–4 (Fig. 2). Reference Orheim and LucchittaOrheim and Lucchitta (1987) pointed out that ratioing satellite bands is a useful technique for extracting information regarding surface properties because it accentuates the albedo signal by reducing the slope effects. An evaluation of NOAA AVHRR and Landsat TM images shows that the AVHRR band 2/1 ratio provides a good visual separation between glazed surfaces and snow. Figure 3 shows the satisfactory congruence between the spatial distribution along the traverse of erosional features (glazed surfaces) and the slope and the ratio of AVHRR band 2/1.

Fig. 2. (a) Subscene of Landsat TM image in false colour (RGB, 4, 3, 2) showing a wide glazed surface area of the plateau with glazed surfaces (blue in image) and snow (grey), transverse dunes (orange lines) and directions of wind inferred from satellite image (blue line). (b) Slope map of area in (a) produced from DEM. (c) Elevation along A–B profile produced from DEM.

Fig. 3. (a) AVHRR band 2/1 ratio image of the traverse area from TNB to DC and micro-relief surveyed along the traverse. (b)Micro-relief size and distribution along the traverse. (c) Slope produced from DEM and the band 2/1 ratio along the traverse.

An important observation from the micro-relief survey is that whilst the redistribution and erosional features are spatially continuous up to 1000 km, depositional micro-reliefs only extensively occur close to the David Glacier ice divide at 810 km and in the last 200 km (Fig. 3). Between Midpoint (MdPt) and DC, the micro-reliefs are mostly of the redistribution and depositional type. The erosional features, between the David Glacier ice divide and DC, are present only on the leeward side of megadunes. The ice divide between David Glacier and the Ross Ice Shelf basins (Fig.1), along the traverse, is a flat area (810 km from GPS1), and the difference in elevation between the two basins is <10m, characterized by a low slope (50.1%) similar to the DC area. Along the traverse TNB–DC, erosional features (glazed surfaces) constitute 31%, redistribution features (sastrugi) 59% and depositional features only 10% of micro-reliefs. The average height of micro-relief distribution shows very high variability, including an increase in height from 400 km to 1000 km from GPS1 (Fig. 3). Slope and AVHRR band 2/1 ratio profiles and micro-relief surveys show very high variability along the coastal and plateau areas, and homogeneous characteristics in the dome area. The lower value of AVHRR band 2/1ratio is present in the first 300 km of traverse where the katabatic wind model of Reference Bromwich, Parish and ZormanBromwich and others (1990) shows speeds of 8–16ms^–1. The intensity and persistence of predominantly wintertime katabatic winds has created an extensive blue-ice area along the valleys and on the leeward sides of nunataks (Reference FrezzottiFrezzotti, 1998) in the mountainous area of TNB.

Remote-sensing and field data along the traverse TNB–DC indicate four sectors (Fig. 3):

Remote-sensing analysis of the plateau area integrated with field data allowed the survey of the following macromorphology: drift plumes and snowdrifts, wide glazed surfaces, sastrugi glazed surface fields, transversal dunes, and megadunes.

Drift-plume, snowdrift and blue-ice areas have been surveyed in the mountain area up to 2200 ma.s.l.; these features have been extensively studied by Reference Bromwich, Parish and ZormanBromwich and others (1990) and Reference Frezzotti, Flora and UrbiniFrezzotti (1998). In this paper we investigate the unexplored part of the plateau.

A wide glazed surface is an extensive area (several km²) where the surface is characterized by glazing (Fig. 2). Awide glazed surface presents cracks (up to 2 cm), with patterns in polygonal form, that could be correlated to a long-term hiatus in snow accumulation (Reference WatanabeWatanabe, 1978). Glazed surfaces were one of the common features observed along the traverse and consist of a single snow-grain thickness layer cemented by thin (0.1–2mm) films of regelated ice. Trenches cut in mound slopes of wide glazed surface show the depth-hoar layer (up to 2 m) penetrated by iced crust with a very coarse snow grain-size (up to 2 mm). Under strongly developed glazed surfaces the depth-hoar layer clearly indicates prolonged sublimation due to a hiatus in accumulation and therefore a long, multi-annual, steep temperature-gradient metamorphism (Reference GowGow, 1965). Long-term hiatus forms do not allow the burial of snow layers by accumulation in subsequent years. The sublimation and upward transport of water vapour belongingto the subsurface snow layer causes the condensation of vapour (recrystallization) on the lower part of the ice crust (Reference Fujii and KusunokiFujii and Kusunoki, 1982). The glazed surface forms on the surface following the kinetic heating of saltant drift snow under constant katabatic wind flow (Reference GoodwinGoodwin, 1990) and the condensation–sublimation process on both sides of the crust (Reference Fujii and KusunokiFujii and Kusunoki, 1982). Reference Fujii and KusunokiFujii and Kusunoki (1982) determined a density value of 0.69 gcm^–1 from the glazed surface at Mizuho station. GPS and GPR profiles of a wide glazed surface, located on the traverse between GPS2 and 31Dpt (Fig. 1), show excellent resolution of sedimentary structures that could be correlated with long-term hiatus forms or with low or lack of accumulation processes (Fig. 4). Themass balance of this wide glazed surface is nil or slightly negative. Integration of satellite images with slope and field-survey data indicated that the glazed surfaces are present in a wide area of the plateau where the slope is higher than 0.25˚ (0.4% or 4m km^–1). Figures 2 and 4 show slope and elevation profiles of the wide glazed surface, where the correlation between the presence of wide glazed surface and the increase of slope in the downwind part is obvious.

Fig. 4. GPS and GPR profiles of wide glazed surface show the reduction of thickness and disappearance of the snow layer in the wide glazed surface that could be correlated with longterm hiatus in accumulation.

The drainage basin of the glaciers that drain into Terra Nova Bay is about 235 000 km² in area (Reference Frezzotti, Tabacco and ZirizzottiFrezzotti and others, 2000), about 30% of which presents a slope of 40.4%. Stake measurements performed in the traverse from GPS1 to 31Dpt showed a large spatial variability, with ablation values up to 13 kgm^–2 a^–1 in the wide glazed surface area. Reference Frezzotti, Tabacco and ZirizzottiFrezzotti and others (2000) pointed out that the ice discharge of glaciers draining into TNB is less than half that required for a zero net surface mass balance, according to the inputs given by the accumulation estimates widely adopted at present. The explanation for this large apparent imbalance is probably an overestimation of snow accumulation using historical data from snow-pit stratigraphy, as pointed out by Reference StenniStenni and others (in press), and the extensive presence of wide glazed surface in the basin with nil or slightly negative snow accumulation.

Sastrugi glazed surface fields have the typical texture of a seasonal feature, with a relatively flat horizontal plane and a thin and soft glazed surface. Sastrugi glazed surface fields have been frequently surveyed along the traverse in the plateau part (300–900km), and they are the most common macro-morphological structure observed along the traverse and surveyed from satellite images. The surface conditions are characterized by the alternating occurrence of wide, smooth surfaces (glazed surfaces) and wide, rough surfaces (sastrugi zones) with a distinct boundary (Fig. 5) and a nearly uniform width for considerable distances. Sastrugi glazed surface fields are characterized by the alternation of sastrugi fields with sporadic longitudinal dunes (10–20m long anda few metres wide), measuring up to 1m in height, and flat glazed surfaces with sporadic sastrugi. The fields boundaries are roughly linear, with the line of elongation parallel to sastrugi direction and therefore to the prevailing wind direction. Sastrugi glazed surface fields are typically several kilometres long and 100–200m wide, cover several hundred km² and are similar to longitudinal sand dunes in satellite images (Fig. 5). The regional slope of a sastrugi glazed surface field area is generally 50.3%. One of the most important pieces of positive feedback is the smoothness and low albedo of the glazed surface compared to the roughness and albedo of the surrounding sastrugi field, causing the wind to be stronger over the glazed surface than over the sastrugi (Reference Fujii and KusunokiFujii and Kusunoki, 1982; Reference Van den Broeke and BintanjaVan den Broeke and Bintanja, 1995). As hypothesized for longitudinal sand dunes (Reference HouboltHoubolt, 1968), pressure exists between the axes of the inter-dunes (glazed surfaces) and the crest of the dunes (sastrugi fields); these pressure gradients are caused by the sastrugis resistance to the wind. This results in the formation of long vortex airflows in the glazed areas, which give the wind an overall spiral motion directed outward at ground level towards the sastrugi, with transportation of snow from the glazed surfaces to the sastrugi fields. Vortices may also control the migrations of sastrugi fields along the wind direction.

Fig. 5. Subscene of Landsat TM image in false colour (RGB, 4, 3, 2) showing an area of plateau with a sastrugi glazed surface field; the arrow shows the prevailing wind direction.

Transverse dunes have been surveyed in satellite images along the traverse between 2200ma.s.l. (74°30 S, 153°20 E) and 3150ma.s.l. (75°15 S, 127°20 E). Transverse dunes are present in sporadic areas between 2200 and 3000ma.s.l., mainly downwind of the wide glazed surface (Fig. 2), where the surface decreases in slope and the atmosphere above the surface has a wave motion. Transverse dunes are perpendicular to the slope and similar to that reported by Reference Black and BuddBlack and Budd (1964) in Wilkes Land. Extensive presence of transverse dunes (4100 000 km²), defined as megadunes, occurs along the traverse route between 75°27 S, 129°56 E and 75°25’ S, 129°14’E (Fig. 6) and has been recognized by Fahne-stock and others (2000b) and described in Reference Frezzotti, Gandolfi and UrbiniFrezzotti and others (in press). Megadune crests are found to be perpendicular to the prevailing katabatic wind direction but present an angle in the direction of the general surface slope at regional scale turning to the left under the action of the Southern Hemisphere Coriolis force (Reference Frezzotti, Gandolfi and UrbiniFrezzotti and others, in press).

Fig. 6. Subscene of Landsat ETM+ in false colour (RGB, 4, 3, 2) showing an area of plateau with megadune; the arrow shows the prevailing wind direction.

The megadune surveyed along the traverse has a wavelength of about 3 km and an amplitude of 2–4m, in agreement with the results of Reference Fahnestock, Scambos, Shuman, Arthern, Winebrenner and KwokFahnestock and others’ (2000b) remote-sensing analysis. Glazed surfaces are located on the leeward slope of the megadune, while severe sastrugi (up to 1.5 m high) are located on the uphill slope (windward). Alternating sastrugi (up to 40 cm) and glazed surfaces are located at the bottom of the inter-dune area (Fig. 7). A buried megadune sedimentary structure (Fig. 7) has been surveyed by GPR–GPS (Reference Frezzotti, Gandolfi and UrbiniFrezzotti and others, in press). The presence of long-term hiatus surfaces (glazed surfaces) on the downwind faces of megadunes, and of accumulation–redistribution forms (severe sastrugi) on the uphill faces, suggests the surface migration of megadunes uphill (to windward). The uphill migration could be related to increased extension and uphill migration of severe sastrugi burying the slip face (long-term hiatus surfaces) of previous windward megadunes. Migration and ice-sheet surface flow (Reference Bamber, Vaughan and JoughinBamber and others, 2000) show a close module (with the difference being due to the Coriolis force) but moving in the opposite direction and therefore having little or no absolute movement (Reference Frezzotti, Gandolfi and UrbiniFrezzotti and others, in press). This result supports the satellite observation by Reference Fahnestock, Scambos, Shuman, Arthern, Winebrenner and KwokFahnestock and others (2000b) that shows an identical pattern of branching and megadune expansion in images acquired 34 years apart with a maximum possible shift of 60 ma^–1. Analysis of surface slope and satellite images indicated that the slope along prevalent wind direction and glazed surfaces is a crucial factor in the genesis of megadunes. Megadune formation could be explained by feedback between the megadune surface and the atmosphere. To explain the cyclic variation of the redistribution and erosion process of snow along a slope by an atmospheric wave, the atmospheric wavelength must be the same order of magnitude as the megadune (3–4 km). A gravity–inertia wave disturbing the geostrophic equilibrium of katabatic wind must be formed with weak amplitude (3–4 m), and the triggering occurs at the break in slope along the prevalent wind direction (Reference Frezzotti, Gandolfi and UrbiniFrezzotti and others, in press). Reference Pettré, Pinglot, Pourchet and ReynaudPettré and others (1986) hypothesized the formation of a gravity– inertia wave of 40 km, disturbing the geostrophic equilibrium, to explain periodic oscillations in accumulation at the break in slope 200 km from the Adélie Coast. Classification of the glazed surface and snow using Landsat ETM+ images indicated that the glazed surface covers about 20% of the megadune area.

Fig. 7. Surface elevation, micro-relief morphology and internal layering along the megadune profile redrawn from Reference Frezzotti, Gandolfi and UrbiniFrezzotti and others (in press).

These feedback systems between the cryosphere and atmosphere have important implications for the choice of sites for ice coring, because orographic variations of even a few metres per kilometre have a significant impact on the snow-accumulation process. The extensive presence of glazed surface caused by a long-term hiatus in accumulation, with a zero or slightly negative accumulation rate, has a significant impact on the surface mass balance of a wide area of the interior of East Antarctica.

Aeolian direction

The micro-relief directions measured on the field are parallel to the directions detected from the macro- morphology survey inferred from satellite images (Fig. 8). However, the field-inferred directions show the surface wind pattern in much greater detail and reveal areas of diffluence, divergence and confluence as a result of mesoscale topography. Satellite image data provide a large spatial coverage and are more representative at the regional scale. Directions detected in the field and by satellite data generally follow the predicted katabatic wind-field surface (Reference Parish and BromwichParish and Bromwich, 1991) from MdPt to DC (higher plateau); more differences are present in the confluence zone from MdPt to GPS1 over relatively steep slopes and mainly between M2 and 31Dpt. The difference between the observational and model (Reference Parish and BromwichParish and Bromwich, 1991) results could be related to the coarse spatial resolution of the model (50 km) and the inaccuracies of the surface topography (Reference DrewryDrewry, 1983) used for modelling (Reference Bamber and HuybrechtsBamber and Huybrechts, 1996). Sastrugi and macro-morphology surveys show an angle between the wind direction and the direction of general surface slope at regional scale (Fig. 8). The difference between wind direction and slope decreases from the interior of the plateau (DC) to the confluence area of TNB. Wind direction turns to the left under the action of the Coriolis force in the high plateau, and becomes directed in a more downslope direction near the steep coastal ice slopes in response to the sudden enhancement in the katabatic acceleration (Reference Bromwich, Parish and ZormanBromwich and others, 1990). The surface relief in the region from GPS1 to MdPt is oriented in two main directions. Micro-relief of a redistribution type is oriented parallel to the maximum slope, while micro- relief of the depositional type is oriented 30–70˚ off maximum slope. It is highly probable that the surface relief of the depositional type is formed by cyclonic storms, while that of the redistribution type is formed by katabatic winds. Near the dome area (TD and DC) the snow surface becomes smooth, with micro-reliefs oriented in two main directions.

Fig. 8. Directions of sastrugi surveyed during the traverses; wind direction inferred from satellite images, satellite image location, streamlines of wind-field model simulation (Reference Parish and BromwichParish and Bromwich, 1991) and drainage basins.