The project will forecast extreme weather events using multimodal and probabilistic machine learning/deep learning methods to contribute in improvements in flood/drought management. That is to reduce the human and social impacts of flood/drought events caused by weather phenomena such as storms and hurricanes and to support more sustainable development in vulnerable areas. The current numerical weather models face challenges in terms of the complexity of extreme weather systems and the non-linear dynamics of meteorological and topographical predictors. Also, data limitations, such as uncertainties in sensor measurements, and missing data, can distort prediction quality. Additionally, concerns have been expressed for their interpretability along with the computational cost of deep learning models. This project will address these challenges by developing systems that combine the diversity of data sources and incorporate uncertainty quantification. This allows us to take advantage of the multiple, complementary forms in the data and to account for uncertainty, which will improve the quality of the predictions and enable better risk assessment. Moreover, the exploration of explainable artificial intelligence (XAI) methods will enable public trust by increasing the transparency in the model predictions.
Machine learning, Deep learning, Deep Multimodal fusion, Temporal fusion, Attention Mechanisms, Time-series forecasting, Hydrological forecasting, Hydrological-Hydraulic modelling, Atmospheric circulation factors, Signal processing, Flood/drought risk management, Probabilistic forecasting, Uncertainty quantification, Extreme predictions, Extreme distributions, Explainable AI (XAI), Explainable Anomaly Detection (XAD)
Worked 3 years as an Analytics and Software Developer, Arcadis, UK
MSc in Data Science, Newcastle University, UK
BSc in Physics, Aristotle University of Thessaloniki, Greece
Dr Andy Barnes
Prof Mike Tipping
Dr Thomas Kjeldsen
