About the project
Coastal flooding is the second largest non-malicious risk to the UK. Accurate coastal forecast information is critical to enabling EA Incident Managers to assess coastal flood risk in real time and take appropriate mitigating actions.
Coastal water level forecast errors are currently twice the target accuracy in key locations due to complex local coastal processes. Forecast uncertainty at Thames Barrier, for example, forces operational teams to be overly cautious and more frequently close the barrier. This increases the probability of closures (together with sea level rise) exceeding that which is feasible before early 2030s, at which point a multi-billion pound upgrade will be needed.
EA water level forecasts are based on a dynamical ocean model. This approach is skillful away from the coast but is fundamentally limited in effectiveness at the coastline, where processes are complex. Remnant errors/model bias have necessitated post-processing of raw model outputs. Current post processing techniques are simplistic and focus on correcting long-term bias. By contrast, machine learning techniques better represent error from complex or poorly understood processes and constrain model trajectories to real time observations. Machine learning techniques have shown promising potential to make a significant contribution to surge forecast skill but have yet to be explored in an operational context by the EA.
As well as Professor Ivan Haigh, you will also receive supervision from Dr Jeff Polton (lead supervisor) at the National Oceanographic Centre (NOC), Southampton.
Please contact the lead supervisor if you require further information about the project.