Cyclone and extreme weather risks for hospitals across Vanuatu, Solomon Islands, Kiribati and Tonga are being modelled in a study led out of Waipapa Taumata Rau, University of Auckland.

Climate models with a high level of local detail are being combined with models of tropical cyclones to estimate risk levels through 2100 in the machine-learning project.

A decade ago, Cyclone Ian ripped part of the roof off Niu'ui Hospital in the northeast Tongan islands of Ha'apai, leading to the hospital being relocated on a higher site seen as less vulnerable to disaster.

Pacific Island communities are especially vulnerable to climate change because so many people live close to the coast and not far above sea level.

Project leader, climate scientist Dr Michelle McCrystall, said that while Pacific Island nations have contributed little to the emissions causing global warming they face some of the harshest consequences.

"These nations don't have unlimited funds to cope with disaster and rebuild," she said.

"Hopefully, these risk projections will aid planning for the future, giving indications of the sites most at risk from the storms, and limiting losses."

The one-year project, due to conclude at the end of 2024, was possible due to a US$139,000 (AUD 210,000) grant from Climate Change AI (CCAI), a global not-for-profit organisation.

Findings for about 30 hospital sites will be shared on a website to be accessible and useful.

"We're building up relationships with hospitals and health facilities and welcome more of this contact – it's a very practical project which can provide some really useful information," Dr McCrystall said.

Collaborators include Dr Berlin Kafoa, the director of the public health division of the international organisation Pacific Communities.