Artificial intelligence: A step change in climate model prediction for climate adaptation

To this day, climate models face the challenge of providing the high-resolution predictions—with quantified errors—needed by a growing number of adaptation planners, from local decision-makers to the private sector, who require detailed assessments of the climate risks they may face. face locally.

This requires a step change in the accuracy and usability of climate predictions, which according to the authors of the paper “Harnessing AI and Computing to Enhance Climate Modeling and Prediction” can be brought about by Artificial Intelligence.

Commentary was published in Nature Climate change by an international team of climate scientists, including CMCC Chief Science Officer Giulio Boccaletti and CMCC President Antonio Navarra.

One proposed approach for a step change in climate modeling is to focus on global models with 1 km horizontal resolution. However, the authors explain, although kilometer-scale models were once considered “digital twins” of the Earth, they still have the same limitations and biases as current models . Furthermore, due to their high computational costs, they impose limitations on the size of simulation ensembles, needed to calibrate the inevitable experimental models of unresolved processes and to calibrate amount of uncertainty.

In general, km-scale models do not yield such stepwise changes in accuracy as to justify accepting the limitations they impose.

Instead of prioritizing kilometer-scale resolution, the authors propose a balanced approach that focuses on generating large ensembles of simulations at moderately high resolution (10–50 km, from about 100 km, which is the standard today) takes advantage of advances in computing and AI to learn. from data.

By increasing moderate global resolution while extensively exploiting observational and simulated data, this approach is more likely to achieve climate modeling goals for risk assessment, including mitigation model error and uncertainty quantification, while allowing for broader application.

1,000 simulations at 10 km resolution cost the same as 1 simulation at 1 km resolution. “While we should push the limits of resolution as computer performance increases, climate modeling over the next decade needs to focus on resolutions in the 10–50 km range,” the authors write. “It is important that climate models be developed so that they can be used and improved through rapid iteration in a globally distributed and comprehensive research program, without centralizing resources into a few monolithic centers is necessary if focusing on the kilometer scale.” global model.”