LEAP RESEARCH GROUPS
Overview
LEAP Research Groups formalize and expand areas of research that have emerged as central to our Center’s mission. They serve as collaborative hubs serving to connect LEAP researchers, NCAR partners, and the broader community as engines of intellectual progress, collaboration, and community within LEAP.
Each Research Group will:
- Advance high-quality research in its domain;
- Contribute to cross-group integration; and
- Strengthen the broader research ecosystem of LEAP and beyond.
Research Groups + Leads
CLIMATE IMPACT RESEARCH
HYBRID MODELING
The Hybrid Modeling Research Group will serve as a collaborative hub connecting high-resolution simulations, satellite and in-situ observations, and model development within CESM. The group focuses on translating process-level insights from convection research into improved parameterizations — including emerging ML-based schemes — and is centrally positioned in efforts to replace traditional parameterizations like the Zhang-McFarlane convection scheme with ML-based alternatives. The group will also support training and mentorship, and help transition new approaches from research into operational Earth System modeling.
PARAMETER ESTIMATION
The Parameter Estimation Research Group focuses on advancing methodologies for robustly inferring physical parameters from sparse, noisy, and heterogeneous data. Priorities include establishing best practices for integrating ML into parameter estimation workflows, leveraging ensemble generation for uncertainty quantification, improving interpretability of inferred parameters, and scaling estimation procedures to large, high-resolution datasets. The group will create a shared platform for methods, lessons, and tools across the CESM community.
MULTISCALE MODELING
The Multiscale Modeling Research Group advances ML as a framework for improving multiscale representation of processes — such as turbulence, convection, and ocean-atmosphere interactions — that span wide ranges of spatial and temporal scales. The research will leverage differentiable Earth system modeling frameworks to design scalable, stable workflows that integrate ML with physics-based models, with a focus on improving fidelity, interpretability, and predictive skill in coupled climate simulations.
SCIENTIFIC / PHYSICS-INFORMED MACHINE LEARNING
The Scientific / Physics-Informed Machine Learning Research Group focuses on advancing physics-informed and hybrid ML methods for cloud microphysical processes. Priority efforts include creating benchmark datasets for ice and warm-rain microphysics to enable community-wide progress, and designing differentiable microphysics schemes for end-to-end training and calibration within ESMs. The group aims to bridge methodological ML advances with domain-specific needs in atmospheric science, positioning LEAP as a leader in AI-enabled, physics-informed Earth System modeling.
DIFFERENTIABLE ESM
The Differentiable ESM Research Group focuses on transitioning from component-level tuning to fully coupled, differentiable ESM calibration — the standard ultimately required by major modeling centers. Work includes jointly optimizing land and atmosphere components within unified differentiable frameworks, enabling next-generation large ensembles that strategically sample key uncertainty sources (cloud feedbacks, carbon cycle responses, transient climate sensitivity). The group emphasizes open, interoperable tools that are broadly adoptable across the Earth system modeling community.
Support LEAP
Help support our grand challenge to develop the next generation of climate model and climate projection for tailored adaptation.
Please contact us to start the conversation.