T010-0001
A Machine Learning Heat Flow Model for Antarctica
A Machine Learning Heat Flow Model for Antarctica
Tuesday, 8 December 2020
Poster
Abstract:
Ice sheet modelling and glacial isostatic adjustment in Antarctica are considerably affected by geothermal heat flow. It influences the ice rheology and can lead to basal melting, thereby promoting ice flow. Direct observations are sparse and models inferred from e.g. magnetic or seismological data differ immensely. While these different approaches are generally justified, they go along with strong simplifications and great uncertainties.
We adopt a machine learning approach to overcome such shortcomings by assuming that heat flow is substantially related to its geodynamic setting. More specifically, we establish a Gradient Boosted Regression Tree model, in order to find an optimal predictor for locations with sparse direct heat flow observations. With this technique, a complex relationship between geothermal heat flow and relevant geophysical features (e.g. gravity field, magnetic anomaly, crustal and lithospheric thickness) is generated. As a result, we can produce a map of predicted heat flow beneath the Antarctic ice sheet.
However, this approach largely relies on global data sets, which are notoriously unreliable in Antarctica. Therefore, validity and quality of the data sets is reviewed. Using regional and more detailed data sets of Antarctica’s tectonic neighbors changes the predictions significantly. Finally, we present a new geothermal heat flow model and discuss differences to previous predictions.
We adopt a machine learning approach to overcome such shortcomings by assuming that heat flow is substantially related to its geodynamic setting. More specifically, we establish a Gradient Boosted Regression Tree model, in order to find an optimal predictor for locations with sparse direct heat flow observations. With this technique, a complex relationship between geothermal heat flow and relevant geophysical features (e.g. gravity field, magnetic anomaly, crustal and lithospheric thickness) is generated. As a result, we can produce a map of predicted heat flow beneath the Antarctic ice sheet.
However, this approach largely relies on global data sets, which are notoriously unreliable in Antarctica. Therefore, validity and quality of the data sets is reviewed. Using regional and more detailed data sets of Antarctica’s tectonic neighbors changes the predictions significantly. Finally, we present a new geothermal heat flow model and discuss differences to previous predictions.