Master's Thesis

Thesis Title: Geogenic radon mapping of Hessen using Machine Learning Techniques

Overview

This research focused on spatial modelling of Geogenic radonpotential (GRP) in Hessen District using machine learning techniques, and environmental covariables.

Geogenic radonpotential (GRP) is defined as the portion of radon emanation, that is predominantly associated with natural factors.

Hypothesis : Spatial variability of GRP is affected by environmental parameters related to soil, geology, meterology etc. Therefore, by modelling a relationship between known sampling points and their environmental co-variables, we can predict the radon potential in areas where the environmental co-variables are present but radon potentials are unknown.

Data and Method

• geogenic radonpotential sampling data.

• 38 covariates related to geology, soil, climate, Uranium, DEM etc.

• Use spatial cross validation for feature selection.

• Develop models such as Random Forest, xGBoost, Suport Vector Regressor, Multi-Layer-Perceptron Regressor

Deliverables