Martian ice caps are made of CO₂ and water ice interlayered with dusty horizons. Previous investigations resorted to radar data for the characterization of the inner layering of the northern and southern caps, although in optimal circumstances the layering is well visible also from optical imagery (e.g. Hirise/CTX and Cassis for ice at lower latitudes).

The definition of an automated ML-based pipeline for the rapid identification of the potential outcropping layers will allow to perform direct correlations between surface and subsurface data, improving three dimensional models of the inner ice-caps structure.

The analysis is incremental:

1. Identification and definition of features suitable for describing layers in optical imagery
2. Scaling of the system for producing an HPC pipeline for automated processing of imagery
3. Large scale application on a selection of images for the rapid evaluation of layers presence

A bonus result would be reaching a full digitalization of layer's horizon on imagery in vector form and application or similar pipeline for SHARAD / MARSIS radargrams.