Abstract:Earth system models (ESMs) serve as vital tools for comprehensively simulating the intricate interplay of physical, chemical, and biological processes across the Earth system's diverse components. Here, we provide a brief overview of the historical development of ESMs and highlight key challenges posed by the intricate feedback mechanisms in the cryosphere, the nonlinear and long-term effects of the lithosphere, and the growing impacts of human activities for modeling Earth system. We then focus on the current opportunities in Earth system modeling, driven by the growing capacity for data-driven approaches such as machine learning (ML) and Artificial Intelligence (AI).The next generation of ESMs should embrace dynamic frameworks that enable more precise representations of physical processes across a range of spatiotemporal scales. Multi-resolution models are pivotal in bridging the gap between global and regional scales, fostering a deeper understanding of local and remote influences. Data-driven methodologies including ML/AI offer promising avenues for advancing ESMs by harnessing a wide array of data sources and surmounting limitations inherent in traditional parameterization techniques. However, the integration of ML/AI into ESMs presents its own set of challenges, including the identification of suitable data sources, the seamless incorporation of ML/AI algorithms into existing modeling infrastructures, and the resolution of issues related to model interpretability and robustness. A harmonious amalgamation of physics-based and data-driven methodologies have the potential to produce ESMs that achieve greater precision and computational efficiency, better capturing the intricate dynamics of Earth system processes.Although ESMs have made substantial progress in simulating the complex dynamics of Earth system's subsystems, there is still considerable work to be done. Prospects in the development of ESMs entail a deepened comprehension of pivotal subsystems, including the anthroposphere, lithosphere, and cryosphere. Adopting innovative technologies and methodologies, such as ML/AI and multi-resolution modeling, holds immense potential to substantially enhance our capability to anticipate and mitigate the consequences of human activities on the Earth system.

Recommendations for Comprehensive and Independent Evaluation of Machine Learning-Based Earth System Models

Next-Generation Earth System Models: Towards Reliable Hybrid Models for Weather and Climate Applications

Machine learning for numerical weather and climate modelling: a review

Evolution and Prospects of Earth System Models: Challenges and Opportunities

Coevolution of machine learning and process‐based modelling to revolutionize Earth and environmental sciences: A perspective

Machine Learning: New Ideas and Tools in Environmental Science and Engineering

Learning earth system models from observations: machine learning or data assimilation?

A Deep Learning Earth System Model for Stable and Efficient Simulation of the Current Climate

Leveraging machine learning to enhance climate models: a review

Machine Learning Emulation of Urban Land Surface Processes

Machine Learning for the Physics of Climate

Applications of physics-informed scientific machine learning in subsurface science: A survey

Machine learning and artificial intelligence to aid climate change research and preparedness

A Quantitative Analysis on the Use of Supervised Machine Learning in Earth Science

Validating Deep-Learning Weather Forecast Models on Recent High-Impact Extreme Events

A Review of Machine Learning Approaches to Soil Temperature Estimation

The promise of implementing machine learning in earthquake engineering: A state-of-the-art review

Machine Learning in Environmental Research: Common Pitfalls and Best Practices.

Iterative Integration of Deep Learning in Hybrid Earth Surface System Modelling

Challenges in data-based geospatial modeling for environmental research and practice