Uncertainty Quantification for Surface Ozone Emulators using Deep Learning

2025-08-06

AI Summary

To address the limited accuracy of physics-based models and the lack of uncertainty quantification in deep learning approaches for surface ozone simulation, this paper proposes an uncertainty-aware U-Net architecture that integrates multi-source chemical transport model outputs and land-use data to predict spatial biases of the MOMO-Chem model. Methodologically, we innovatively couple a Bayesian neural network with quantile regression to enable calibrated uncertainty estimation. Furthermore, we leverage the resulting uncertainty distribution to identify optimal and suboptimal ground monitoring stations, thereby enhancing the interpretability of bias correction. Empirical evaluation over North America and Europe in June 2019 demonstrates significant improvements in regional bias estimation accuracy. The results reveal land use as a key driver of residual spatial heterogeneity, offering policy-relevant, interpretable, and probabilistically grounded support for ozone pollution mitigation strategies and public health decision-making.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Uncertainty Representations
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Air pollution is a global hazard, and as of 2023, 94% of the world's population is exposed to unsafe pollution levels. Surface Ozone (O3), an important pollutant, and the drivers of its trends are difficult to model, and traditional physics-based models fall short in their practical use for scales relevant to human-health impacts. Deep Learning-based emulators have shown promise in capturing complex climate patterns, but overall lack the interpretability necessary to support critical decision making for policy changes and public health measures. We implement an uncertainty-aware U-Net architecture to predict the Multi-mOdel Multi-cOnstituent Chemical data assimilation (MOMO-Chem) model's surface ozone residuals (bias) using Bayesian and quantile regression methods. We demonstrate the capability of our techniques in regional estimation of bias in North America and Europe for June 2019. We highlight the uncertainty quantification (UQ) scores between our two UQ methodologies and discern which ground stations are optimal and sub-optimal candidates for MOMO-Chem bias correction, and evaluate the impact of land-use information in surface ozone residual modeling.

Problem

• Quantify uncertainty in deep learning ozone emulators for health policy
• Improve interpretability of ozone models using Bayesian and quantile regression
• Identify optimal ground stations for ozone bias correction

Innovation

• Uncertainty-aware U-Net for ozone prediction
• Bayesian and quantile regression methods
• Evaluates land-use impact on ozone modeling