Enhancing sub-ice geology in East Antarctica with Self-Organizing maps based on gravity, magnetic and radar data

Sub-ice geology significantly influences the dynamics and future evolution of the Antarctic Ice Sheet, but largely inaccessible for direct sampling. Here, we present an approach, where we use a Self-Organizing Map (SOM) to describe sub-glacial properties. Based on attributes derived from gravity, magnetics and radar data from the NASA Operation Ice Bridge dataset in East Antarctica, we train a SOM, where attributes are selected to best represent sub-glacial conditions. Therefore, we study the trade-offs between these data sets helping to identify for which properties these are most sensitive.
The trained SOM identifies the outlines of the main geological structures beneath the ice and supplements models based on inverse and forward modelling. In contrast to such often regional interpretations, the SOM captures small-scale structures at the ice bed, as we illustrate with case examples, and highlights areas with inconsistencies in existing geological interpretations. The SOM can furthermore be used as input for inverse modelling of the physical properties of the sub-glacial geology in Antarctica.