Colloquium: Dr. Charlotte Connolly | Loyola University Maryland

TITLE:  Using Machine Learning to Explore Possible Climate Futures

ABSTRACT: Possible futures where the proposed method of climate intervention, stratospheric aerosol injection (SAI), occur are shaped by social, environmental, and political landscapes. These hypothetical SAI scenarios are often explored by simulating the SAI scenario with Earth system models. However, the number of scenarios simulated is limited by available resources. Here, the applicability and skill of a machine learning emulator designed to explore SAI scenarios is discussed. This machine learning emulator is comprised of multiple interpretable machine learning models that are trained on simulations from the Community Earth System Model version 2 including a novel simulation titled, Climate Response After Stratospheric Sulfur dioxide injection (CRASSULA). In total, this machine learning emulator is comprised of four individual machine learning models that independently predict the internal variability of near surface temperature, temperature responses to the total CO2 in the atmosphere, aerosol optical depth (AOD) response to SAI, and temperature response to changes in AOD. The machine learning emulator can generate the near surface temperature response to a variety of SAI scenarios which can inform the design of the more computational expensive earth system models simulations.