Machine Learning for Modelling Cyclones
JBPacific is partnering with Griffith University WIL program to develop statistical methods for estimating the size of tropical cyclones.
Tropical cyclones are a frequent hazard along the Australian coastline, bringing intense winds, waves, and storm surge to coastal communities. The ability to predict cyclone size and intensity is crucial to determining hazards from waves, erosion and inundation. JBPacific’s coastal engineers are using numerical models to simulate extreme cyclonic events and estimate potential hazards
One of the key parameters in accurately modelling a tropical cyclone system is its absolute size, or extent of influence, often associated with the “radius to outermost closed isobar” or ROCI. This dimension is difficult to quantify and can vary significantly between cyclones of similar intensity, as well as during the life of a single cyclone.
This joint project is analysing historic BOM cyclone data as well as international weather databases with the goal of developing a statistical relationship for predicting cyclone size. This project will use Python programming machine learning techniques together with Linear and Tree-based regression methods to seek to develop a statistical model for making predictions of cyclone size for use in cyclone modelling.