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 : The 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 is unknown.

Data and Method

• Radon-222 and Soil gas permeability per measuring location are used to calculate the Geogenic radonpotential (GRP) using the (Neznal et al. 2004) equation.

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

• Use spatial cross validation for feature selection.

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

• Choose the best performing models for spatial prediction.

Deliverables

The code and workflow will be available here after publication of the research