New AI-powered climate models can now predict regional climate impacts — including flooding, drought, and heat wave patterns — at resolutions of 1 kilometre, compared to the 100-kilometre resolution of traditional models.
The improvement enables city-level adaptation planning for the first time. Governments can now see which specific neighbourhoods will be most affected by flooding, where heat islands will intensify, and which agricultural regions face the greatest drought risk.
The models, developed by teams at Google DeepMind, Microsoft Research, and ECMWF, process satellite data, weather station readings, and historical climate records to produce predictions for the next 50 years.
Several countries have already incorporated AI climate predictions into national adaptation plans. The Netherlands is using them to prioritise flood defence investments, and Australia is using them to identify regions where agriculture will need to transition to more resilient crop varieties.