Predicted Antarctic Heat Flow and Uncertainties using Machine Learning

We predicted Antarctic Geothermal Heat Flow (GHF) using a machine learning approach. The adopted approach estimates GHF from multiple geophysical and geological data sets, assuming that GHF is substantially related to the geodynamic setting of the plates. We applied a Gradient Boosted Regression Tree algorithm to find an optimal prediction model relating GHF to the observables. In Antarctica, only a sparse number of direct GHF measurements are available, and therefore, in addition to the global models, we explore the use of regional data sets of Antarctica as well as its tectonic Gondwana neighbors to refine the predictions. We hereby demonstrated the need for adding reliable data to the machine learning approach. Here, we present a new geothermal heat flow map, which exhibits intermediate values compared to previous models, ranging from 35 to 156 mW/m2 and showing visible connections to the conjugate margins in Australia, Africa, and India. Also, the data set contains minimum and maximum heat flow values and maximum absolute differences, resulting from calculating three additional heat flow models with different feature set-ups to assess the direct uncertainties.