Artificial Intelligence for Climate Action

Incorporating physics constraints into machine learning approaches can ensure that algorithms do not violate physical laws, as well as learn with limited data. ISC researchers are building models of sea-ice drift with physics-inspired neural networks and modeling the impact of hurricanes on wave height using neural operators.

Traditional climate models often estimate climate variables at coarse spatial resolution for global coverage. We are applying machine learning super-resolution methods to enhance the spatial resolution of climate data, particularly focused on coastal regions prone to extreme temperatures. This will enhance the precision and usability of climate variables for localized climate impact studies.

As the climate changes faster than emissions can be mitigated, we are faced with the need for climate adaptation. The best way to adapt given resource constraints is a complex problem. ISC researchers are developing reinforcement learning methods to enable climate-informed decision-making. In particular, given potential solutions to address climate threats, our researchers optimize which solutions to deploy, and where, toward a comprehensive coastal defense-in-depth approach.

For media inquiries, please contact the APL Public Affairs office.