A first attempt at a continental scale geothermal heat flow model for Africa

Reliable and direct geothermal heat flow (GHF) measurements in Africa are sparse. It is a challenging task to create a map that reflects the GHF and covers the African continent in in its entirety.

We approached this task by training a random forest regression algorithm. After carefully tuning the algorithm's hyperparameters, the trained model relates the GHF to various geophysical and geological covariates that are considered to be statistically significant for the GHF. The covariates are mainly global datasets and models like Moho depth, Curie depth, gravity anomalies. To improve the predictions, we included some regional datasets. The quality and reliability of the datasets are assessed before the algorithm is trained.

The model's performance is validated against Australia, which has a large database of GHF measurements. The predicted GHF map of Africa shows acceptable performance indicators and is consistent with existing recognized GHF maps of Africa.