Publications
You can also find my articles on my Google Scholar profile.
-
Run It Back: Autoregression vs. Video Diffusion for End-Conditioned Weather Trajectories
Lin, J., and Barnes, E. A. (2026). Run It Back: Autoregression vs. Video Diffusion for End-Conditioned Weather Trajectories. NeurIPS 2026 Workshop on Tackling Climate Change with Machine Learning. Accepted, peer reviewed.
-
Signoff: Bringing Human Accountability Back to Agent-Powered Research
Lin, J., and Barnes, E. A. (2026). Signoff: Bringing Human Accountability Back to Agent-Powered Research. Bay Area Machine Learning Symposium (BayLearn). Accepted symposium paper, peer reviewed.
-
Extremes on Rewind: Generating 1,000-Member Ensembles Initialized at a Final Condition
Lin, J., Chien, M.-T., Sakarvadia, M., and Barnes, E. A. (2026). Extremes on Rewind: Generating 1,000-Member Ensembles Initialized at a Final Condition. arXiv:2608.19008. Preprint, under review.
Details about Extremes on Rewind: Generating 1,000-Member Ensembles Initialized at a Final Condition
-
Data-driven integration kernels for interpretable nonlocal operator learning
Published as part of the Climate Informatics 2026 proceedings, following the conference’s peer-review process.
Ferretti, S. L., Lin, J., Shamekh, S., Baldwin, J. W., Pritchard, M. S., and Beucler, T. (2026). Data-driven integration kernels for interpretable nonlocal operator learning. Environmental Data Science, 5, e31.
Details about Data-driven integration kernels for interpretable nonlocal operator learning
-
Crowdsourcing the Frontier: Advancing Hybrid Physics-ML Climate Simulation via a $50,000 Kaggle Competition
This paper takes ideas from the winning teams of the 2024 LEAP ClimSim Kaggle competition and systematically tests them offline and online to reveal SOTA performance, universal failure modes, and architecture-dependent idiosyncracies.
Lin, J., Hu, Z., Beucler, T., Frields, K., Christensen, H., Hannah, W., et al. (2026). Crowdsourcing the frontier: Advancing hybrid physics-ML climate simulation via a $50,000 Kaggle competition. Journal of Advances in Modeling Earth Systems, 18, e2025MS005643.
-
ClimSim-Online: A Large Multi-Scale Dataset and Framework for Hybrid Physics-ML Climate Emulation
This paper builds on ClimSim and introduces a containerized version of E3SM-MMF and PyTorch-Fortran binding to allow for standardized online testing across institutions going forward.
-
Stable Machine-Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection-Permitting Simulations
Building off ClimSim, this paper is the first to yield a stable and accurate machine-learning parameterization of subgrid processes (including explicit cloud condensate coupling) in a comprehensive atmospheric model learned from embedded convection-permitting simulations.
-
Navigating the Noise: Bringing Clarity to ML Parameterization with O(100) Ensembles
This paper traverses the offline-online testing gap in hybrid physics-ML climate simulation using statistically meaningful ensembles and identifies the design decisions that durably contribute to better online performance.
Details about Navigating the Noise: Bringing Clarity to ML Parameterization with O(100) Ensembles
-
Climate-invariant machine learning
This paper explores feature transformations that can make machine learned convective parameterizations invariant to distribution-shift in climates.
-
Stress-testing the coupled behavior of hybrid physics-machine learning climate simulations on an unseen, warmer climate
This paper samples the coupled behavior of neural network convective parameterizations at scale on an unseen, warmer climate to see if design decisions conducive to better online performance in-distribution do the same out-of-distribution.
-
ClimSim: A large multi-scale dataset for hybrid physics-ML climate emulation
This paper establishes a comprehensive set of datasets and benchmarks for machine learned subgrid parameterizations of convection and radiation in E3SM-MMF. It received the Outstanding Paper award for the Datasets and Benchmarks track at NeurIPS 2023.
Details about ClimSim: A large multi-scale dataset for hybrid physics-ML climate emulation