Earth System Response to Climate Mitigation and Geoengineering Intervention | Observations, ERA5 Reanalysis, CMIP6 (GeoMIP/CDRMIP), and AI
In the first iteration of this pursuit, we evaluate the potential implications of mitigation and intervention strategies with a set of experiments utilizing historical reanalysis data (e.g., ERA5) and scenario-based model simulations (e.g., GeoMIP, CDRMIP) to examine the global response to deploying these strategies. The following iteration will integrate a reduced complexity model (i.e., OSCAR v3.3) and artificial intelligence (AI) into the framework to better understand the disparities among the permafrost carbon feedback (PCF) as well as optimizing the decision-process governing mitigation and geoengineering intervention. GeoEngAI is a hybridized ensemble learning framework composed of stacked convolutional layers and long short-term memory-encoded bidirectional recurrent neural networks. This multimodal deep learning architecture simultaneously ingests and analyzes reanalysis observations (ERA5) and process-based CMIP6 modeling ensemble outputs (i.e., 2-m temperature, total precipitation, atmospheric methane concentration).
Bradley A. Gay, PhD | NASA Postdoctoral Program Fellow Jet Propulsion Laboratory, California Institute of Technology