Transverse Aeolian Ridges in Oxia Planum, Mars: Classifying landforms using machine learning

Post contributed by Dr. Alexander Barrett, School of Physical Sciences, The Open University, UK.

Transverse Aeolian Ridges (TARs) are a distinctive aeolian landform, commonly found on the surface of Mars (Balme et al., 2008; Bourke et al., 2006). They consist of metre to decimetre scale granular ripples, which form transverse (i.e. perpendicular) to the direction of the prevailing wind. Similar to terrestrial megaripples, TARs are believed to form through surface creep of coarse particles. Understanding the distribution of TARs is important for understanding the aeolian environment in which they formed (Favaro et al., 2021), and for planning rover missions (Balme et al., 2018). TARs present a potential hazard when they occur in rover landing sites. Consequently, a team at the Open University has trained a machine learning model called NOAH-H (developed by SciSys Ltd. for the European Space Agency; Woods et al. 2020) to identify these features, and distinguish them from other common surface textures.

Image 1 shows bedforms in Oxia Planum. The size and prevalence of bedforms in this area is important, since it is the landing site for the ExoMars Rosalind Franklin Rover (Vago et al., 2017), which will arrive in 2023 to search for signs of past and present life in this once water-rich area (Quantin-Nataf et al., 2021). Large bedforms, or continuous regions of loose aeolian sand, could form navigational hazards, limiting the directions in which the rover can progress, or the science targets that it can access. We need to have a good understanding of the distribution of TARs in the immediate area where the rover will land, however this is not precisely known in advance and could be anywhere within a large landing ellipse. Mapping all of the TARs in the potential landing area by hand is impossible in the time available, even for a large team.

Image 1: Large TARs at the planned ExoMars Rosalind Franklin rover landing site at Oxia Planum, Mars (High Resolution Imaging Science Experiment (HiRISE) Image ESP_048292-1985. a) translucent layer showing Machine Learning produced terrain classification, overlying the red-band HiRISE image from which the classification was made. The area is dominated by rugged and fractured bedrock, while topographic lows contain non-bedrock material. This has been shaped into TARs of various sizes and degrees of continuity. The divide between bedrock and non-bedrock textures is well defined, as are the perimeters of patches of TARS. The transitions between textured, fractured and rugged bedrock are less well identified, as these features form a continuum, and so some areas have characteristics of multiple classes. HiRISE Image Credit NASA/JPL/University of Arizona.


Sandstone outcrops seen with the ExoMars PanCam emulator

Post by Dr. Peter Fawdon, (@DrPfawdon) School of Physical Sciences, The Open University, United Kingdom

PanCam (Coates et al., 2017) is the imaging instrument on the 2020 ExoMars rover and consists of two wide angle cameras; (WAC’s) and a High Resolution Camera (HRC). PanCam will be used to lead the geological characterisation of the local area outcrops. It will be used to establishing the geological setting of outcrops and identify targets for subsurface sampling and analysis with the ExoMars drill and suite of analytical instruments (Vago et al., 2017).

An emulator for the ExoMars PanCam instrument has been used in rover operation field trials in southern Spain. The aim of these trials has been to explore how scientists will use the instruments in rover missions. These images, taken by the emulator, are examples of what PanCam data might look like and show how the PanCam images will be used (e.g., Harris et al., 2015).


Image 1: PanCam Multi-spectral images: (A) A colour composite made from the red, green and blue filters shows a ridge named ‘Glengoyne’ at approximately 20 m distance from the rover. (B) A Multi spectral image using the geology filters stretched to emphasise the variation in the scene.


  • Enter your email address to follow this blog and receive notifications of new posts by email.

  • Io

  • Blog Stats

    • 143,891 hits