Attribution-Oriented Atmospheric Visibility: Quantifying Meteorological Constraints and Anthropogenic Cycles Leverage
Journal
Environmental Science and Technology Letters
Journal Volume
13
Journal Issue
2
Start Page
241
End Page
247
ISSN
23288930
Date Issued
2026-02-10
Author(s)
Abstract
Visibility is a direct and policy-relevant indicator of air quality, yet its variability arises from intertwined meteorological and anthropogenic influences that complicate attribution. We developed an attribution-oriented framework using aerosol extinction (βExt) observations in Taichung, Taiwan (2020–2022), together with meteorological parameters and temporal markers, in an interpretable three-stage random forest model (R = 0.89, RMSE = 28 Mm–1). Planetary boundary layer height emerged as the dominant predictor, with contributions up to +135 Mm–1 under shallow layers (<200 m). Relative humidity exhibited threshold behavior, rising between 60–85% and decreasing above 90%, consistent with hygroscopic growth and wet scavenging. These patterns demonstrated physical consistency, while temporal markers captured recurrent emission cycles. Meteorology constrained visibility across a wide range (−50 to +150 Mm–1), whereas anthropogenic cycle varied within ±50 Mm–1, with severe degradation when both were positive (about 30% of cases) and improvements when anthropogenic cycle contributions were reduced during unfavorable meteorology or favorable meteorology offset high emissions. These results show that visibility degradation reflects separable yet interacting processes: meteorology sets broad constraints, while emission reductions offer quantifiable leverage. The attribution-oriented framework provides a physically interpretable basis for using visibility as both an atmospheric diagnostic and a policy-relevant target.
Subjects
aerosol extinction (βExt)
deweather method
explainable machine learning
policy relevance
visibility
Publisher
American Chemical Society
Type
journal article