Geothermal heat flow (GHF) is essential for evaluating the thermal states and energy balances of the lithosphere, playing a crucial role in geophysics and geothermal energy research. In this study, we initiate our analysis by exploring lateral variations in unknown thermal parameters across Germany. We apply a Bayesian Markov Chain Monte Carlo approach, using established data on surface heat flow, surface temperatures, and the temperatures and thicknesses at the lithosphere-asthenosphere boundary. Our investigation focuses on assessing the lateral variations in crustal and lithospheric mantle thermal conductivities, crustal heat production, and mantle heat flow.
To address the limitations posed by the sparse and uneven distribution of direct borehole data, which consists of only 595 heat flow points, our study integrates a broad spectrum of geophysical and geological datasets. These include gravity, magnetics, seismic data, topography, proximity to faults, and volcanoes, and compositional data within a machine-learning framework. This comprehensive approach not only surpasses traditional Curie depth estimations in accuracy but also robustly tackles the issue of data scarcity.
We employ quantile regression forests to clustering to integrate the datasets in a geothermal heat flow model. This probabilistic, multi-geophysical inversion method leads to a detailed quantification of uncertainties, offering a refined understanding of Germany’s geothermal potential.