Title: CREST: Towards an AI-first framework for assembling Earth System Digital Twins and the deployment of the TERRAHydro terrestrial twin
Presenting Author: Craig Pelissier
Organization: NASA Goddard Space Flight Center
Co-Author(s): Brandon Smith, Grey Nearing, Deepthi Raghunandan, Carlos Cruz, Vanessa Valenti, Mahmoud Saeedimoghaddam

Abstract:
Today, the Earth Science community relies on Earth Systems Models (ESMs) to simulate the complex relationships between environmental variables including their chemical, biological, and physical processes and interactions that can be thought of as an evolution of global climate models. Earth System Digital Twins require integrating ESM (Digital Replica) with many relevant, interconnected models, including anthropogenic models, economic models, infrastructure models, as well as complex data analysis and visualization components into a framework capable of making predictions fast enough to perform near real-time system updates (what-now), forecasting (what-next), and be relevant for hypothesis testing and scenario analysis (what-if). The eventual federation of ESDTs into an EDT will require a community of experts working together that spans across all Earth Science disciplines, and includes contributions from teams in government, industry, and academia. In this talk, we present the Coupled Reusable Earth System Tensor Framework (CREST), an AI-first Python-based framework for assembling community developed federated ESDTs. We will discuss its principles: AI-first, federation, interoperability, and accessibility, and report on progress. In addition, we will discuss the CREST-built TERRAHydro AI-based terrestrial digital twin including our progress, prototype demonstrations, and planned deployment.