Home About News E3SM Researchers Join Genesis Mission Projects to Advance AI-Enabled Earth System Prediction

E3SM Researchers Join Genesis Mission Projects to Advance AI-Enabled Earth System Prediction

  • September 9, 2026
  • Brief

    • The U.S. Department of Energy announced 278 Phase I projects under the Genesis Mission, selected from over 5,000 submissions. Phase I grants provide 9-month funding between $500,000 and $750,000 to integrate artificial intelligence with scientific research.

    E3SM researchers are contributing to 13 Genesis Mission projects that apply artificial intelligence, machine learning, and advanced Earth system modeling to some of the nation’s most pressing water and energy challenges. The projects span a wide range of scientific areas, including cloud microphysics, precipitation, turbulence, storm prediction, hydrology, groundwater, flooding, and subseasonal-to-seasonal forecasting. Together, they reflect E3SM’s expanding role in developing next-generation tools to improve predictions of water availability and extreme weather.

    E3SM Staff as PIs

    E3SM staff are the principal investigators (PIs) of seven projects.

    One project, “From improved Coupled Ocean-Ice States to Improved Water Prediction: A Generative AI Approach” (LANL description), led by PI Luke Van Roekel (LANL) and coPIs Andrew Roberts (LANL) and Peter Caldwell (LLNL), with Olawale Ikuyajolu (LANL) contributing, is specifically designed to support E3SM. The project develops a computationally efficient generative data assimilation system using diffusion models and graph neural networks to create physically consistent, uncertainty-aware ocean and sea-ice initial conditions directly on E3SM’s unstructured mesh. Through E3SM and emulator hindcasts, it aims to improve weeks-to-years U.S. water-availability forecasts while reducing initialization costs and supporting reusable AI and high-performance computing (HPC) workflows.

    Ordered by number of E3SM contributors, the other projects are:

    Shixuan Zhang (PNNL) is the PI of “Drift-Aware Initialization of Coupled Earth System Models for Scalable Subseasonal-to-Seasonal Prediction Using Agentic AI” (PNNL description) with Ruby Leung (PNNL) and Kai Zhang (PNNL) contributing. The project will use AI to identify and correct initialization problems that can cause coupled model forecasts to lose accuracy.

    Dali Wang (ORNL) is the PI of “Integrated AI-Driven Weeks-to-Years Prediction of Water for Energy in the Tennessee Valley Authority Region” (ORNL description) with Xiaoying Shi (ORNL) and Dan Ricciuto (ORNL) contributing. The project applies AI and Earth system science to deliver actionable weeks-to-years forecasts of water availability, streamflow, river temperatures, and associated risks across the Tennessee Valley Authority region, supporting more resilient energy planning and operations.

    Yan Feng (ANL) is the PI of “Physics-Informed Machine Learning of Cloud Microphysics for High-Resolution Earth System Modeling with Observationally Constrained Online Training” (PAMS abstract, ANL description, LANL description), with Yunpeng Shan (ANL) and Jiwen Fan (ANL) contributing. This project will combine high-fidelity simulations, machine learning, and observations to develop more accurate and stable representations of cloud microphysics across multiple scales.

    Hong-Yi Li (Lehigh University), is the PI of “RIVER-AI: Reservoir-Groundwater Interactions for River Flow Variability, Energy, and Resilience with AI” (PNNL description), with Ruby Leung (PNNL), and Yilin Fang (PNNL) contributing. This project will combine AI, groundwater modeling, reservoir operations, and river-system physics to improve predictions and decision support in the heavily managed Delaware River Basin.

    Qing Zhu (LBNL) is the PI of “WISE: Water Intelligence System for Energy infrastructures” (LBNL description “Meeting Water-Energy Demand”) with Jennifer Holm (LBNL) and former E3SM staff member Kate Calvin contributing. The project develops an AI framework that links U.S. water availability and risks directly to energy needs, helping stakeholders manage cooling demands, data-center growth, power generation, and grid resilience.

    Kyle Pressel (PNNL) is the PI of “EARL: Earth-Atmosphere Agentic Research and Learning — Physics-Constrained AI Closure Development for Cloud Microphysics and Turbulence” (PAMS abstract, PNNL description). The project will develop a physics-constrained machine-learning closure and a human-supervised agentic workflow to accelerate model development and experimentation.

    E3SM Staff as contributors

    E3SM staff are contributing to six further projects. Ordered by number of E3SM contributors, they are:

    Hassan Beydoun (LLNL), Peter Bogenschutz (LLNL), Jiwen Fan (ANL), Yunpeng Shan (ANL), and Xue Zheng (LLNL) are contributing to “Scalable Twin for Intelligent Turbulence and Cloud Heuristics,” or STITCH (PAMS abstract, ANL description). The project will develop process emulators and regional model capabilities using large-eddy simulations, SCREAM, and observations from the ARM facility. The long-term goal is to establish the foundation for an ARM digital twin that can support the discovery of new cloud and turbulence physics.

    Susannah Burrows (PNNL), Meng Huang (PNNL), Johannes Mülmenstadt (PNNL), and Mingxuan Wu (PNNL), are contributing to “DL4MCS: Deep Learning Methods to Enhance Subseasonal Predictions of Mesoscale Convective Systems by Physics-based Systems” (PAMS abstract, PNNL description).The project will combine physics-based models with generative AI and observational data to improve predictions of damaging storm systems weeks in advance.

    Bryce Harrop (PNNL), Ruby Leung (PNNL), and Naser Mahfouz (PNNL) are working on “GENESIS-AR: GENErative Models and New observationS for Improved S2S Prediction of Atmospheric Rivers” (PAMS abstract, PNNL description). The project combines physics-based modeling, generative AI, and century-scale microseism observations to improve calibrated sub-seasonal forecasts of atmospheric rivers, supporting more reliable hydropower planning and reduced flood and infrastructure risks along the U.S. West Coast.

    Ruby Leung (PNNL) is also contributing to “An Agentic AI Framework for Seasonal-to-Interannual U.S. Water Prediction: Coupling Large AI-Based Prediction Models and Wavelet Diffusion with E3SM.”

    Jiwen Fan (ANL) is contributing to “Advancing Multimodal and Multidimensional AI to Discover Connections between Cloud Microphysics and Precipitation” (PAMS abstract, ANL description).The project will use multimodal AI to analyze numerical observations, radar imagery, and other Atmospheric Radiation Measurement (ARM) facility data to identify connections between cloud processes and rainfall.

    Jiwen Fan (ANL) is also contributing to “Multi-Fidelity AI Foundation Model for Coupled Surface-Groundwater Predictions of Water Availability and Flood Hazards Across CONUS” (PAMS abstract, ANL description). The project will develop an AI foundation model that integrates physics-based simulations and observations to produce rapid, uncertainty-aware predictions of groundwater levels, soil moisture, streamflow, and flooding.

    Through these projects, E3SM staff are helping advance AI-enabled Earth system science from individual cloud droplets and storm systems to regional watersheds and continental-scale water prediction. The work supports the Genesis Mission’s broader goal of creating faster, more capable, and more actionable scientific tools for strengthening U.S. energy, water, and infrastructure resilience.

    Reference

    List of Awardees: GM RFA Awards List

    For some projects, further information, including abstracts, can be found on the PAMS Award Search site. Use Advanced Search Parameters: Solicitation Number: DE-FOA-0003612 for Genesis Mission, or search by PI’s name. Direct abstract links have been included for several projects above.

    Lab specific announcements:

    Contact

    • Ryan Forsyth, Lawrence Livermore National Laboratory
     
     

    This article is a part of the E3SM “Floating Points” Newsletter, to read the full Newsletter check:

     
    Send this to a friend