This paper investigates the applicability of the Artificial Neural Network (ANN) learning paradigm in deriving sea ice thickness in the Antarctic using a combination of physical measurements, Landsat-8 satellite imagery, and Synthetic Aperture Radar (SAR) imagery as training input, validated against physical thickness measurements. The paper discusses the complexities of sea ice and the difficulties faced in calculating thickness before detailing the current literature applying ANNs to predicting sea ice characteristics. Methodology is described in terms of an overview of ANNs and the optimisations applied in this study, as well as the geospatial processing and data manipulation undertaken to produce the training and test data sets from sparse point measurements. Results indicate that physical field measurements are the principle contributor to model performance, outperforming the Landsat-8 model and increasing performance in a combined approach. Error for thickness prediction varied between models: ±14cm (physical measurements), ±23cm (Landsat-8), and ±21cm (combined Landsat-8 and physical measurements). The paper concludes with a number of suggested improvements to the model that future studies should consider as well as calling for further validation of the model through additional, actual field measurements over manufactured data points produced through interpolation.