Abstract

Seasonal precipitation extremes may potentially be exacerbated by anthropogenic climate change, increasing the risks of drought and flooding and their associated impact. Given data paucity in certain geographies, such as those perhaps more susceptible to the effects of climate change, the development of empirical models that can deliver high performance with minimum calibration and extant, obtainable inputs could be highly beneficial.

In this talk, I will begin with a simple hydrological model alongside a, likewise, simple artificial neural network and how these serve as the rationale for the development of machine learning approaches to the problem; this includes: identification of a suitable feature set and proxy variables, handling extreme values, a comparison of machine learning methodologies and their suitability to the problem, the development of a generalisable machine learning model, and that of a novel architecture(s). I may also discuss my approach to storm prediction, how my lack of progress has failed me thus far, and how I intend to resolve this and fit it into this project as a whole.