Geophysical Research Letters · 2020 · 366 citations · 45 references
ComputingComputer ScienceEarth SciencesArtificial IntelligenceHydrologic ForecastingMachine LearningDeep LearningHydrological ModelingGeoscience Conceptual UnderstandingModel TransferabilityRecurrent Neural NetworksTransferable Representations
Modeling dynamic geophysical phenomena is central to Earth studies, yet the community largely relies on physical models and the integration of AI remains an open challenge. The study proposes a general framework that embeds temporal dynamic geoscientific models as recurrent neural layers to enhance AI awareness. The framework incorporates physical models as recurrent neural layers within a deep learning architecture. In a runoff modeling case across the United States, the physics‑aware DL model achieved higher prediction accuracy, robust transferability, and effective inference of unobserved processes, marking progress toward physics‑AI integration for Earth system challenges.
Abstract Modeling dynamic geophysical phenomena is at the core of Earth and environmental studies. The geoscientific community relying mainly on physical representations may want to consider much deeper adoption of artificial intelligence (AI) instruments in the context of AI's global success and emergence of big Earth data. A new perspective of using hybrid physics‐AI approaches is a grand vision, but actualizing such approaches remains an open question in geoscience. This study develops a general approach to improving AI geoscientific awareness, wherein physical approaches such as temporal dynamic geoscientific models are included as special recurrent neural layers in a deep learning architecture. The illustrative case of runoff modeling across the conterminous United States demonstrates that the physics‐aware DL model has enhanced prediction accuracy, robust transferability, and good intelligence for inferring unobserved processes. This study represents a firm step toward realizing the vision of tackling Earth system challenges by physics‐AI integration.
45
Learning representations by back-propagating errors
David E. Rumelhart, Geoffrey E. Hinton, Ronald J. Williams · Nature · 1986 · 30.9K citations