Martian Pit Shadow Extractor (MAPS): Determining the Apparent Depths of Martian Pits from the Morphology of their Shadows

This presentation for the ICG2022-25: Planetary Geomorphology session of the 10th International Conference of the International Association of Geomorphologists (IAG) aims to describe the purpose, methodology, and testing results of the Martian Pit Shadow extractor (MAPS) tool. The challenges that were faced in its development, particularly in relation to the geomorphology of these features, will also be discussed. MAPS itself is an automated Python framework which employs K-Means clustering to extract the shadow from a cropped red-band Mars Reconnaissance Orbiter HiRISE image of a Martian pit and measure its width.  MAPS also exploits the sensing information provided for each HiRISE image in order to determine the apparent depth (the depth of the pit at the edge of its shadow).  MAPS is intended to be used after a machine learning model, or another automated method, has detected Martian pits in HiRISE imagery, such that it can be applied to cropped images without the need for manual labelling. While MAPS has currently only been applied to Martian pits imaged by the HiRISE camera, its overall method would be highly applicable to data taken by other sensors, or data taken of other planetary surfaces.

Martian pits are circular-to-elliptical geological depressions commonly seen on the surface of Mars, as well as on other terrestrial planetary surfaces. Their morphologies are most distinguishable from impact craters by the lack of an elevated rim. They are generally formed by gravitational collapse into an underground void, or by the evacuation or removal of subsurface material.  Pits are heavily studied for both their geological significance, but also because of their implications on future space exploration. Current methods can calculate the apparent depths of Martian pits by considering the Sun’s position in the sky and manually measuring the widths of their shadows in remote-sensing imagery. However, an automated method of shadow extraction and depth calculation allows for analysis of far larger numbers of Martian pits, which could be used for further research or engineering use.

There is scope for future work in calibrating the parameters of the MAPS tool, such as the number of clusters that the K-Means algorithm segments images into, as well as the maximum emission angle that is allowed for an input image. The latter was shown in testing to lead to the same features exhibiting significantly different morphologies, despite similar lighting conditions. This may be possible by testing the tool upon artificially created elevation data and images of pits. Other improvements are foreseen by considering the geomorphology of the surrounding Martian surface as a means of calibrating the distribution of the non-shadow pixel values and of controlling their false discovery rates.