2019/06/15 by Tom Beucler, Stephan Rasp, Beucler, Tom +6 · 105 citations
Computer Science · Earth and Planetary Sciences · Engineering · Environmental Science · Mathematics · Physics and Astronomy · #Artificial intelligence #Artificial neural network #Atmospheric and Oceanic Physics (physics.ao-ph) #Climate change #Climate model #Climate variability and models #Cloud computing #Computational Physics (physics.comp-ph) #Computer science #Computer security #Conservation law #Conservation of energy #Ecology #Emulation #Energy conservation #Engineering #FOS: Computer and information sciences #FOS: Physical sciences #Geography #Machine Learning (cs.LG) #Mathematics #Meteorological Phenomena and Simulations #Network architecture #Obstacle #Set (abstract data type) #Solar Radiation and Photovoltaics #cs.LG #physics.ao-ph #physics.comp-ph
paper · pdf · doi:10.48550/arxiv.1906.06622
published in arXiv (Cornell University) (Cornell University) · ICML 2019 Workshop. Climate Change: How Can AI Help? 3 pages, 3 figures, 1 table
arxiv created 2019/06/15 · openalex publication_date 2019/06/15 · arxiv updated 2019/06/18 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Artificial neural-networks have the potential to emulate cloud processes with higher accuracy than the semi-empirical emulators currently used in climate models. However, neural-network models do not intrinsically conserve energy and mass, which is an obstacle to using them for long-term climate predictions. Here, we propose two methods to enforce linear conservation laws in neural-network emulators of physical models: Constraining (1) the loss function or (2) the architecture of the network itself. Applied to the emulation of explicitly-resolved cloud processes in a prototype multi-scale climate model, we show that architecture constraints can enforce conservation laws to satisfactory numerical precision, while all constraints help the neural-network better generalize to conditions outside of its training set, such as global warming.