Periodic Reporting for period 3 - USMILE (Understanding and Modelling the Earth System with Machine Learning)
Reporting period: 2023-09-01 to 2025-02-28
Earth system models are fundamental to understand climate change. Although they have improved significantly, considerable biases and uncertainties in their projections remain. Process parameterisations limit the models’ ability to simulate both global and regional Earth system responses, which are key for assessing climate change and its impacts on ecosystems and society. Reduction in projection uncertainties is still a cornerstone, for which the joint exploitation of higher resolution models, the wealth of observational data, and machine learning promise advances. The overarching goal of USMILE is to develop new data-driven, physics-aware modelling approaches that integrate ML, and deep learning in particular, into Earth System modelling and analysis. ML will allow us to define novel observational constraints on Earth system feedbacks and climate projections. We will (1) develop ML algorithms to enhance Earth observation datasets accounting for spatio-temporal covariations, (2) deploy ML-based parameterisations and sub-models for clouds and land-surface processes that have hindered progress in climate modelling for decades, and (3) detect and understand modes of climate variability, multivariate extremes and uncover dynamical aspects of the Earth system with novel deep learning and causal inference techniques. USMILE will drive a paradigm shift in the current modelling of the Earth system towards a new data-driven physics-aware science and to an unprecedented reduction of uncertainties in projections.
USMILE is now fully underway with many studies being published or under development. During the reporting period, the development of ML-based parametrizations was enhanced to first demonstrations of the potential of hybrid ML models to reduce systematic errors in for example diurnal precipitation, precipitation extremes, and carbon cycle uncertainties. The project has made foundational strides towards addressing two primary challenges: (a) breaking through the Earth system model parameterization deadlock and (b) revolutionizing the analysis and interpretation of Earth system data. These initial advances, though not yet achieving the full scope of the breakthroughs anticipated, represent critical progress.
More specifically, in WP1, we developed machine learning techniques to enhance satellite products and improve understanding of atmospheric and terrestrial interactions, leading to advancements in cloud classification, climatology, and carbon and water flux representation. In WP2 our work on hybrid ML-physical modeling resulted in improved parameterizations for the ICON model, advancements in land process modeling, and the creation of the DifferLand hybrid model for complex land relationships. In WP3 we developed ML techniques to detect and understand climate variability and extreme events. In WP4 we enhanced causal model evaluation, analyzed teleconnections and used ML-based products to constrain drying trends. In WP5, the USMILE PIs published two high-level Perspectives on ML’s role in climate modeling and analysis, advocating for hybrid Earth system models to improve model accuracy and projections also for extreme events, highlighting AI's significance in enhancing climate models and information for mitigation and adaptation.
A substantial portion of USMILE's work has been published in peer-reviewed journals. Publications are accessible through our group’s web pages and generally available in open repositories (e.g. arXiv.org). The code supporting our publications is made publicly available as open-source software. Additionally, we maintain group repositories (e.g. EyringMLClimateGroup GitHub, IPL-UV GitHub) and contribute code to the ESMValTool repository (ESMValTool GitHub).
The USMILE project disseminates its research via its dedicated webpage, https://www.usmile-erc.eu/(opens in new window). Several group members also have Google Scholar and ResearchGate accounts, where ongoing activities and developments are publicized. Members have delivered numerous invited and keynote lectures at geoscience conferences (e.g. EGU, AGU, ELLIS), as well as at machine learning, statistics, and image processing conferences (e.g. NeurIPS), and have given talks at esteemed institutions and summer schools.
More specifically, in WP1, we developed machine learning techniques to enhance satellite products and improve understanding of atmospheric and terrestrial interactions, leading to advancements in cloud classification, climatology, and carbon and water flux representation. In WP2 our work on hybrid ML-physical modeling resulted in improved parameterizations for the ICON model, advancements in land process modeling, and the creation of the DifferLand hybrid model for complex land relationships. In WP3 we developed ML techniques to detect and understand climate variability and extreme events. In WP4 we enhanced causal model evaluation, analyzed teleconnections and used ML-based products to constrain drying trends. In WP5, the USMILE PIs published two high-level Perspectives on ML’s role in climate modeling and analysis, advocating for hybrid Earth system models to improve model accuracy and projections also for extreme events, highlighting AI's significance in enhancing climate models and information for mitigation and adaptation.
A substantial portion of USMILE's work has been published in peer-reviewed journals. Publications are accessible through our group’s web pages and generally available in open repositories (e.g. arXiv.org). The code supporting our publications is made publicly available as open-source software. Additionally, we maintain group repositories (e.g. EyringMLClimateGroup GitHub, IPL-UV GitHub) and contribute code to the ESMValTool repository (ESMValTool GitHub).
The USMILE project disseminates its research via its dedicated webpage, https://www.usmile-erc.eu/(opens in new window). Several group members also have Google Scholar and ResearchGate accounts, where ongoing activities and developments are publicized. Members have delivered numerous invited and keynote lectures at geoscience conferences (e.g. EGU, AGU, ELLIS), as well as at machine learning, statistics, and image processing conferences (e.g. NeurIPS), and have given talks at esteemed institutions and summer schools.
The work in the indiviual WPs if ongoing. The full progress beyond the state of the art as outlined in the proposal will be achieved in the remaining time of the project.