Random Forest-Based Prediction of Geothermal Heat Flow in the Arabian-Nubian Shield: A Multi-Proxy Satellite Geophysical Approach

<title>Abstract</title> <p> Quantifying terrestrial heat flow (THF) is essential for developing geothermal energy, yet direct measurements remain scarce across the Arabian-Nubian Shield (ANS), a tectonically active region with high geothermal promise. To address this data limitation, we construct the first spatially continuous, multi-observable geothermal model for the entire ANS using a Random Forest Regression (RFR) workflow. Sixteen geophysical and geological predictors, including Curie Point depth (CPD), gravity fields, Moho geometry, and lithospheric thickness, are integrated after rigorous outlier removal via an Isolation Forest algorithm. Hyperparameter tuning yields strong predictive performance, with a coefficient of determination (R ²  = 0.92) and a root-mean-square error (RMSe) of 20.4 mW m⁻². Predicted THF ranges from 20 to 200 mW m⁻², defining three thermal domains: (1) high-potential zones (&gt; 80 mW m⁻²) along the Red Sea Rift, (2) moderate-potential zones (60–80 mW m⁻²) under the Precambrian shields, and (3) low-potential zones (&lt; 60 mW m⁻²) beneath the stable Arabian Platform. Variable importance analysis identifies Moho depth and lithosphere–asthenosphere boundary (LAB) depth as the dominant controls, implying strong vertical thermal coupling from the mantle to the surface. By converting sparse borehole measurements into a continuous, physically interpretable heat flow field, our framework reduces exploration risk. It provides actionable targets for high-enthalpy geothermal development, supporting carbon-reduction strategies across the ANS region. </p>