Recommendations for Comprehensive and Independent Evaluation of Machine Learning‐Based Earth System Models DOI Creative Commons
Paul Ullrich, Elizabeth A. Barnes, William D. Collins

и другие.

Journal of Geophysical Research Machine Learning and Computation, Год журнала: 2025, Номер 2(1)

Опубликована: Март 1, 2025

Abstract Machine learning (ML) is a revolutionary technology with demonstrable applications across multiple disciplines. Within the Earth science community, ML has been most visible for weather forecasting, producing forecasts that rival modern physics‐based models. Given importance of deepening our understanding and improving predictions system on all time scales, efforts are now underway to develop Earth‐system models (ESMs) capable representing components coupled (or their aggregated behavior) response external changes over long timescales. Building trust in ESMs much more difficult problem than forecast models, not least because model must represent alternate (e.g., future or paleoclimatic) states which there no direct observations. physical principles enable about often explicitly coded these ML‐based demonstrating credibility thus requires us build evidence consistency system. To this end, paper puts forward five recommendations enhance comprehensive, standardized, independent evaluation strengthen promote wider use.

Язык: Английский

Recommendations for Comprehensive and Independent Evaluation of Machine Learning‐Based Earth System Models DOI Creative Commons
Paul Ullrich, Elizabeth A. Barnes, William D. Collins

и другие.

Journal of Geophysical Research Machine Learning and Computation, Год журнала: 2025, Номер 2(1)

Опубликована: Март 1, 2025

Abstract Machine learning (ML) is a revolutionary technology with demonstrable applications across multiple disciplines. Within the Earth science community, ML has been most visible for weather forecasting, producing forecasts that rival modern physics‐based models. Given importance of deepening our understanding and improving predictions system on all time scales, efforts are now underway to develop Earth‐system models (ESMs) capable representing components coupled (or their aggregated behavior) response external changes over long timescales. Building trust in ESMs much more difficult problem than forecast models, not least because model must represent alternate (e.g., future or paleoclimatic) states which there no direct observations. physical principles enable about often explicitly coded these ML‐based demonstrating credibility thus requires us build evidence consistency system. To this end, paper puts forward five recommendations enhance comprehensive, standardized, independent evaluation strengthen promote wider use.

Язык: Английский

Процитировано

0