[Paper Review] Physics-aware Machine Learning Revolutionizes Scientific Paradigm for Machine Learning and Process-based Hydrology

Qingsong Xu, Yilei Shi+4 more|arXiv (Cornell University)|Oct 8, 2023

Hydrological Forecasting Using AI15 citations

TL;DR

This paper conceptualizes and reviews physics-aware ML (PaML) for hydrology, categorizing methods (physical data-guided ML, physics-informed ML, physics-embedded ML, and physics-aware hybrid learning), and introducing HydroPML as an open-source platform to bridge ML with process-based hydrology.

ABSTRACT

Accurate hydrological understanding and water cycle prediction are crucial for addressing scientific and societal challenges associated with the management of water resources, particularly under the dynamic influence of anthropogenic climate change. Existing reviews predominantly concentrate on the development of machine learning (ML) in this field, yet there is a clear distinction between hydrology and ML as separate paradigms. Here, we introduce physics-aware ML as a transformative approach to overcome the perceived barrier and revolutionize both fields. Specifically, we present a comprehensive review of the physics-aware ML methods, building a structured community (PaML) of existing methodologies that integrate prior physical knowledge or physics-based modeling into ML. We systematically analyze these PaML methodologies with respect to four aspects: physical data-guided ML, physics-informed ML, physics-embedded ML, and physics-aware hybrid learning. PaML facilitates ML-aided hypotheses, accelerating insights from big data and fostering scientific discoveries. We first conduct a systematic review of hydrology in PaML, including rainfall-runoff hydrological processes and hydrodynamic processes, and highlight the most promising and challenging directions for different objectives and PaML methods. Finally, a new PaML-based hydrology platform, termed HydroPML, is released as a foundation for hydrological applications. HydroPML enhances the explainability and causality of ML and lays the groundwork for the digital water cycle's realization. The HydroPML platform is publicly available at https://hydropml.github.io/.

Motivation & Objective

• Bridge the knowledge gap between physics-based hydrology and machine learning by structuring PaML methodologies.
• Systematically analyze PaML approaches across four categories in the hydrology context.
• Assess PaML applications in hydrodynamics and rainfall-runoff processes.
• Propose HydroPML as an open-source platform to enable PaML-based hydrology applications.

Proposed method

• Define PaML and categorize approaches into physical data-guided ML, physics-informed ML, physics-embedded ML, and physics-aware hybrid learning.
• Provide a systematic review of PaML methods with examples spanning neural networks, deep operator networks, and physics-discovery models.
• Analyze PaML applications in hydrodynamic and rainfall-runoff processes to classify by objectives and methods.
• Highlight promising directions and challenges for PaML in process-based hydrology.
• Introduce HydroPML as a foundation for PaML-based hydrological applications and real-time flood forecasting.

Experimental results

Research questions

• RQ1 What are the main PaML paradigms used to integrate physics with ML for hydrology?
• RQ2 How do PaML methods perform across hydrodynamic and rainfall-runoff processes?
• RQ3 What are the promising directions and remaining challenges for PaML in process-based hydrology?
• RQ4 How can an open-source platform (HydroPML) facilitate PaML adoption in hydrology?

Key findings

• PaML is categorized into four groups: physical data-guided ML, physics-informed ML, physics-embedded ML, and physics-aware hybrid learning.
• A systematic review is conducted for PaML in hydrology, focusing on hydrodynamic and rainfall-runoff processes.
• HydroPML is released as an open-source hydrology platform to enhance explainability, causality, and real-time forecasting in water systems.
• The paper emphasizes that PaML can accelerate scientific insights by enabling ML-aided hypotheses and better exploitation of big data in hydrology.
• The study outlines promising directions and challenges for different application objectives within PaML methods.