Shane Zhang , Michael H Ritzwoller

Subglacial geothermal heat flux affects the dynamics of the Antarctic ice sheet but is poorly known. We estimate heat flux across West Antarctica and the interior of East Antarctica by transporting heat flux observations from the contiguous US and Europe, based on seismic structure with a lateral resolution of about \qty{100}{km}. We transport with three Machine Learning models across a hierarchy of complexity (Linear Regression, Decision Tree, and Random Forest). We have validated the models within the US and Europe and cross-validated them between the two continents. The uppermost mantle shear wavespeed is the primary predictor. The geographical patterns can be reasonably reproduced but the variability tends to be underestimated. Antarctic heat flux estimates are highly consistent between the three Machine Learning models, but we report the Decision Tree results. In West and East Antarctica, respectively, we estimate subcontinental-scale heat flux values of $64\pm \hfu{7}$ (spatial mean $\pm$ standard deviation) and $53 \pm \hfu{3}$. Heat flux varies regionally across West Antarctica from \hfu{50} to \hfu{83} and across East Antarctica from \hfu{50} to \hfu{74}. Particularly high heat flux occurs in the western Transantarctic Mountains and Marie Byrd Land in West Antarctica and the lowest heat flux occurs in Victoria Land, the Wilkes Subglacial Basin, the Vostok Subglacial Highland, and south of Maud Subglacial Basin in East Antarctica. We estimate systematic errors including a constant shift in the continental average (\hfu{2.5}) and a variability (higher highs, lower lows) of $25\%$ from the continental average, along with an average random error of about \hfu{13} in West Antarctica and \hfu{10} in East Antarctica. Our estimates are consistent with a recent seismically-based study by \textcite{Shen_etal_2020} but are significantly lower in West Antarctica than a magnetically-based study by \textcite{Martos_etal_2017}. To resolve such inconsistencies, validation of all proxies against heat flux observations outside Antarctica is needed.