Skip to main content

Semi-automated Design of Artificial Intelligence Earth Science Models...

by Philipe Ambrozio Dias, Henry Medeiros, Wadzanai D Lunga, Nagendra Singh, Ranjeet Devarakonda
Publication Type
Conference Paper
Book Title
Artificial Intelligence for Earth System Predictability
Publication Date
Conference Name
DOE White Papers to Advance an Integrative Artificial Intelligence Framework for Earth System Predictability
Conference Location
TBA, Tennessee, United States of America
Conference Sponsor
Conference Date

Prediction and observation of water cycles involve not only patterns isolated in space and time, but rather modeling complex spatio-temporal relationships across multiple sources of data and domains. For instance, Evapotranspiration (ET) and Leaf Area Indexes (LAI) are two critical components in DOE’s Energy Exascale Earth System Model (E3SM). Accurate assessments of ET and LAI are critical for understanding hydrological processes, deforestation, crop yield, and irrigation impacts. However, current ET estimates for global simulations are available at very coarse spatial resolution. They are usually derived from satellite data based on broad plant functional types (PFT), which fail to capture the fine-scale variations due to change in vegetation type across the globe. Within this context and in light of the data-model integration challenges highlighted in the EESSD Strategic Plan, the new era of AI model development for geosciences calls for data-driven methods that provide domain scientists with estimations of parameters such as PFT and LAI in an efficient, interpretable, and easy-to-operate manner.