ThousandWorlds: A benchmark for climate emulation of potentially habitable exoplanets
Abstract
The search for life beyond Earth will depend on detecting faint signatures in the atmospheres of potentially habitable exoplanets. Interpreting those signatures requires understanding the host planet's climate: the same molecule may signal life on one planet and abiotic chemistry on another. Global climate models (GCMs) provide this understanding, but individual runs can require up to millions of core-hours and substantial domain expert time. Machine-learning emulators could remove this bottleneck, but progress has been limited by the absence of a curated, multi-model exoclimate dataset. We introduce ThousandWorlds, an ML-ready benchmark for exoclimate emulation and for the broader regime of low-data, multi-simulator, parameter-to-field regression. The dataset contains approximately 1,700 simulations from five GCMs, mapping eight planet parameters to 3D atmospheric fields including temperature, humidity, winds, clouds, and radiation. Three nested subsets define progressively harder challenges: single-simulator regression, multi-simulator regression with complete observations, and multi-simulator regression with structured missingness. We propose two evaluation protocols: one for ranking methods, and one that measures performance relative to the disagreement between GCMs themselves. We evaluate ten baselines spanning simple methods, trees, deep learning, and Gaussian processes. GP-based methods perform best, suggesting that ThousandWorlds exposes a regime where off-the-shelf deep learning does not yet succeed. Data: https://doi.org/10.57967/hf/8695. Code: https://github.com/edstevenson/ThousandWorlds.
Related
- Detecting transiting exoplanets in simulated PLATO data. A comparison of light curve filter and transit search algorithms
- Magnetic field strengths of hot giant exoplanets consistent with Solar System values
- Characterizing the oxidation state of rocky exoplanets with the Large Interferometer for Exoplanets (LIFE)
- How dust particles orbiting white dwarfs can reveal undetected co-orbiting exoplanets II. Exoplanets on elliptic orbits
- Life in the Cosmic Neighbourhood: Galactic Habitable Zones in the EAGLE Simulations
BibTeX
@misc{stevenson2026thousandworlds,
title = {{ThousandWorlds: A benchmark for climate emulation of potentially habitable exoplanets}},
author = {Edward T. Stevenson and Mei Ting Mak and Eric Wolf and Denis E. Sergeev and Tobi Hammond and N. J. Mayne and Miles Cranmer},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2606.18338},
eprint = {2606.18338},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.18338},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - ThousandWorlds: A benchmark for climate emulation of potentially habitable exoplanets
AU - Edward T. Stevenson
AU - Mei Ting Mak
AU - Eric Wolf
AU - Denis E. Sergeev
AU - Tobi Hammond
AU - N. J. Mayne
AU - Miles Cranmer
PY - 2026
DA - 2026/09/25
PB - arXiv
DO - 10.48550/arxiv.2606.18338
UR - https://arxiv.org/abs/2606.18338
AB - The search for life beyond Earth will depend on detecting faint signatures in the atmospheres of potentially habitable exoplanets. Interpreting those signatures requires understanding the host planet's climate: the same molecule may signal life on one planet and abiotic chemistry on another. Global climate models (GCMs) provide this understanding, but individual runs can require up to millions of core-hours and substantial domain expert time. Machine-learning emulators could remove this bottleneck, but progress has been limited by the absence of a curated, multi-model exoclimate dataset. We introduce ThousandWorlds, an ML-ready benchmark for exoclimate emulation and for the broader regime of low-data, multi-simulator, parameter-to-field regression. The dataset contains approximately 1,700 simulations from five GCMs, mapping eight planet parameters to 3D atmospheric fields including temperature, humidity, winds, clouds, and radiation. Three nested subsets define progressively harder challenges: single-simulator regression, multi-simulator regression with complete observations, and multi-simulator regression with structured missingness. We propose two evaluation protocols: one for ranking methods, and one that measures performance relative to the disagreement between GCMs themselves. We evaluate ten baselines spanning simple methods, trees, deep learning, and Gaussian processes. GP-based methods perform best, suggesting that ThousandWorlds exposes a regime where off-the-shelf deep learning does not yet succeed. Data: https://doi.org/10.57967/hf/8695. Code: https://github.com/edstevenson/ThousandWorlds.
ER -