Galway and Limerick cities would have experienced extensive flooding in 2025 if Storm Éowyn had struck at a high or spring tide, the lead researchers on a project to build an AI-powered flood forecasting system for Irish cities have found.
Indiana Olbert, an engineer, and economist Thomas McDermott from the University of Galway, lead StopFloods4.ie, which is developing a model to help emergency managers predict where floods will hit and which responses to prioritise in the lead-in to extreme weather events.
Olbert analysed 20 years of tide gauge records to understand why Éowyn, which brought a very rare red rainfall warning to the entire country in January 2025, caused little flooding on the west coast.
She found the storm surge was “exceptionally high”, something she said had never been seen in those records. The surge is the extra water pushed on to the coast by high winds and very low atmospheric pressure, on top of the predictable tide. At Galway Port the tide surge peaked at a level of 2.6m at about 6am on January 24th, 2005, which was unprecedented in the records.
However, it coincided with a neap tide, and a tide that was going out. The total water level rose to the equivalent of a high tide, which Galway’s defences are designed to handle.
“If that occurred on a spring tide at high water level, we would be dealing with nearly six metres of water,” Olbert said.
Olbert said if water levels had risen an additional 2.6m at spring tide, areas including Flood Street, High Street, Lower Shop Street and the Claddagh could have been “substantially inundated”. She said Salthill promenade, 1.5km to the west, was most vulnerable to a southwesterly wind and might have escaped the worst of flooding during a high tide. A more westerly wind pushed more water into Galway Port at the mouth of Galway Bay.
Limerick city would have suffered a similar flood event because of the prevailing southwesterly winds, with a similar storm surge and extensive flooding around the Shannon.
The other big population centre on the Atlantic coast, Cork, would have escaped the worst of the flooding during Storm Éowyn, said the researchers. That was because the wind was not blowing directly into the city and it would therefore not have been in the direct line of the rising waters caused by the storm surge.
Olbert said the storm was only partly predicted. The meteorological forecasts were for gusts of up to 130km/h, but more than 180km/h was recorded at four locations, including Mace Head in Co Galway.
McDermott said the challenge for extreme weather managers in local authorities is that flooding on the ground depends on a number of factors combining. Good forecasts exist for the individual components, including atmosphere and storm surges, but not collectively. The research project aims to put them together so that, in the days and hours leading up to a weather event, authorities can see which specific streets, areas, and properties are most likely to flood.
At present, he said, a local authority facing a warning may have dozens of browser tabs open, pulling information from Met Éireann, international weather models and the Marine Institute, then relying on local knowledge to interpret it. That makes the system vulnerable if staff move on or if climate change shifts the risk profile. Councils must also decide where to send crews and whether to close roads, which can lead to difficult trade-offs.
The project has received funding from Research Ireland and has incorporated internet-of-things sensors, hydrodynamic modelling, statistics and machine learning, and is being prototyped with decision makers at city levels.
McDermott said earlier work by the two teams on Cork showed how costs could escalate. The 2009 flood in Cork caused damages of more than €100 million. It was modelled as roughly a one-in-160-year event historically, but under future climate conditions such damage could occur much more frequently, he said. Under a high-emissions scenario, annual flood costs in Cork could rise tenfold by 2050, from tens of millions of euro to hundreds of millions.
McDermott also said his team used machine learning on the national planning applications database and found about 10,000 housing units were granted permission in flood risk zones between 2016 and 2022, despite the OPW’s flood maps and planning guidelines. The maps reflect historical risk, he said. He said that, in future, flood plains would get bigger, and areas that already flood would flood more often, thus putting an even greater stock of housing at risk of flood events.
