Automated detection and morphometric analysis of antarctic megadunes: quantifying spatial distribution, size, surface shape, and orientation relative to the wind.

Introduction 

In East Antarctica, local surface mass balance is strongly modulated by wind-driven erosion, sublimation, and snow or condensation redistribution processes. Persistent katabatic winds drive the formation of extensive parallel and alternating ridges and troughs that stand, respectively, above and below the regional ice-sheet surface. They present distinctive surface topography and associated spatial variations in snow accumulation [1,2]. They are flat bedforms with typical amplitudes of a few meters, wavelengths of several kilometers, and lengths of tens to hundreds of kilometers [2-5]. These bedforms are the so-called megadunes, which are sedimentary bedforms that migrate upwind [2-8]. Similar bedforms have been also described on the North Polar Cap of Mars [7].

On Earth as well on Mars, they represent geomorphic imprints of katabatic dynamics and snow redistribution process. These bedforms influence the spatial and temporal patterns of mass and heat exchanges within the ice sheet [3-4], in the overlying atmosphere and across the ice-atmosphere interface due to aerodynamic and mass transfer feedbacks. Consequently, their existence might introduce non-climatic signals in the stratigraphic and isotopic record retrieved from ice cores [12]. In addition, spatial variations in topography, texture and metamorphism associated with megadunes can modify the surface radiometric properties, thereby affecting reflection and absorption of solar radiation and surface heat exchanges [7-8]. Variations in topography and aerodynamic roughness also influence the structure and dynamics of the overlying atmospheric boundary layer [6]. A comprehensive characterisation of megadune spatial organisation, size, surface shape, texture, internal architecture, composition, dynamics, and formation conditions is therefore essential for accurately quantifying mass and heat transfers within ice sheets, in the overlying atmosphere, and across their interface.

Methods

To provide maps of megadune fields only, and enable their morphometric analyses, our mapping protocol is based on their characteristic undulating surface shape. This shape may be defined geometrically as a series of alternating convex ridges and concave troughs, defined by their crest lines, trough lines and boundaries. Characterising surface wavelength, amplitude, and shape of megadunes requires detecting these lines, which could in principle be achieved by computing first and second derivatives of surface elevation. In practice, however, detecting local maxima, minima, and inflections by derivation of gridded datasets representing natural surfaces is highly sensitive to noise and data imperfections; in addition, megadunes, like most bedforms, often form complex, branched networks, and their actual surface shape may include local perturbations, which introduce difficulty in the detection of crest lines, trough lines and boundaries through simple derivative-based methods.
We therefore adopt an alternative approach based on methods recently developed for the detection and morphometric analyses of sand dunes [9-10]. We detect crest lines and trough lines by iterative skeletonization of raster representations of composite detectors that combine several geometrical attributes derived from surface elevation. In parallel, we delineate ridges from raster representations of volumetric obscurance, a geometrical parameter that highlights 95 convex surface features while reducing noise in the topographic signal [10]. This delineation is performed using a deep-learning recognition algorithm. Although the resulting outlines do not correspond to surface inflections, they provide a delineation of ridges that is useful to (i) produce continental-scale maps and (ii) remove artifacts from the detected crest lines and trough lines, before these are used in the morphometric analysis. Morphometric parameters derived from these features are then analyzed with respect to the wind speed and direction derived from the ERA5-Land weather reanalysis.

Results and discussion 

We present new observations on the spatial distribution, size, surface shape, and orientation relative to the wind, of megadunes based on REMA surface elevation data at the scale of the entire East-Antarctic Ice Sheet. The detected megadunes cover about 7. 10⁵ km² and include several previously undocumented fields that confirm megadunes occur in regions where the wind speed is comprised between 6 and 10 m.s^-1 and the regional topographic slope between 0.8 and 1.6‰. Their wavelength 2±0.5 km and amplitude 2.4±1.9 m decreases from the central regions of megadune fields towards their margins. They display a statistical tendency for steeper lee sides compared to stoss sides; this asymmetry however disappears once local slope measurements are corrected for the background regional surface slope.

Megadune spatial distribution likely reflects spatial, and possibly temporal, gradients in the environmental conditions governing megadune formation and growth. The relationship to the prevailing wind is quantified by the wind incidence angle (WIA), defined as the horizontal angle between the wind direction and the crest line. They are oriented at 113±25° to the right of the prevailing wind direction. This obliquity likely reflects the confined nature of katabatic flow within the atmospheric boundary layer, and the clockwise tendency may tentatively be attributed to modulation of megadune development by the Coriolis force.

Conclusion and perspectives

The automated deep-learning–based method to the REMA surface elevation dataset enabled the detection of megadunes over approximately 5% of the East Antarctic Ice Sheet. This value encompasses newly detected megadune fields, which morphometric properties that confirm previous statements. In contrast to previous estimates, we find that megadunes are oblique to wind streamlines, with an average angle of 113±25° to the right of the prevailing wind direction.

The resulting dataset provides systematic and statistically significant geometrical constraints for (i) models of megadune dynamics, (ii) assessments of the environmental conditions required for their formation, and (iii) appraisal of their potential influence on both ice sheet dynamics and atmospheric processes.

Acknowledgement

 Part of this research work was done in the framework of the GeoPlaNet Master programme (https://geoplanet-impg.eu/).

References

[1] Dolgushin (1958), Izvestiya Akademii Nauk SSSR, Seriya Geograficheskaya, 28—47.

[2] Black and Bud (1964), Journal of Glaciology, 3–15.

[3] Frezzotti et al. (2002), JGR: Atmospheres, 107.

[4] Anschütz et al. (2006), Geophysical Research Letters, 33.

[5] Alberti and Biscaro (2010), Computers & Geosciences, 36, 1-9.

[6] Dadic et al (2013), JGR: Earth Surface, 118.

[7] Herny et al (2014), Earth and Planetary Science Letters,403, 56-66.

[8] Traversa et al. (2023), The Cryosphere, 17, 427-444.

[9] Shumack et al. (2020), Earth Surface Processes and Landforms, 45, 2417-2431.

[10] Daynac et al. (2024), Geomorphology, 463.