Accurate prediction of extreme weather events are crucial from a societal point of view, where the consequences of said events can have major financial and demographic impacts upon society.
Extreme Value Theory (EVT) provides a statistical framework for the modelling of such extreme events.
On the other hand, Bayesian Neural Networks (BNNs) extend traditional neural networks by incorporating Bayesian inference, which provides a probabilistic approach to learning and prediction in any given regression task.
In this thesis, we extend the methodology of a recently introduced BNN and integrate it with EVT to be able to infer the parameters of Generalised Extreme Value (GEV) distributions.
We then apply our methodology to annual maximal rainfall in Eastern Canada, where we infer and interpolate GEV parameter estimates across the interpolation region.
The obtained results demonstrate that our approach outperforms Polynomial Regression and Inverse Distance Weighting methods in predicting extreme rainfall events.