A Double Machine Learning Approach for Policy Evaluation Using Meteorologically Fixed Air Pollutant Time Series as an Emission Proxy
Year: 2026
Type: article
Abstract: Air pollution results from the time-dependent combined effects of source emissions and meteorological conditions. Effective air quality management relies on understanding changes in the emission strength of air pollutants, yet this remains technically challenging due to the unobservable nature of emission sources. Various statistical methods have been developed to estimate emission changes using air quality monitoring data. In Recent years, machine learning-based meteorological normalization techinques have become popular in decoupling meteorological effects from air quality time series. These techiniques typically work by normalizing air pollutant concentrations to averaged meteorological conditions. However, exisiting methods have inherent limitations in accurately tracking emission changes, largely due to inconsistencies in the normalized meteorological conditions across observations. This study introduces MetFixed, an improved approach that normalizes air quality time series under consistent, counterfactual meteorological scenarios using predictions from a machine learning model. Applying MetFixed to observational time series of NO2 - a short-lived primary pollutant - at a typical roadside monitoring site, we demonstrated improved alignment with traffic-related emission proxy and counterfactual emission scenarios, compared to widely used normalization technique. The approach also significantly reduces computational cost. Additionally, an auto-MetFixed model is developed to assist local authorities in implementing near real-time assessments of emission abatement during air pollution emergencies.
Source: ACS ES&T Air
Authors: Qili Dai, Yufen Zhang, Xiaohui Bi, Yinchang Feng
Institutions: Ministry of Ecology and Environment, Nankai University
Language: English
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
SDG: Climate action
Open Access status: hybrid
Estimated APC paid: $4,500
OpenAlex ID: W7154691080