Statistical and machine learning methods for evaluating trends in air quality under changing meteorological conditions

Abstract

Abstract. Evaluating the influence of anthropogenic-emission changes on air quality requires accounting for the influence of meteorological variability. Statistical methods such as multiple linear regression (MLR) models with basic meteorological variables are often used to remove meteorological variability and estimate trends in measured pollutant concentrations attributable to emission changes. However, the ability of these widely used statistical approaches to correct for meteorological variability remains unknown, limiting their usefulness in the real-world policy evaluations. Here, we quantify the performance of MLR and other quantitative methods using simulations from a chemical transport model, GEOS-Chem, as a synthetic dataset. Focusing on the impacts of anthropogenic-emission changes in the US (2011 to 2017) and China (2013 to 2017) on PM2.5 and O3, we show that widely used regression methods do not perform well in correcting for meteorological variability and identifying long-term trends in ambient pollution related to changes in emissions. The estimation errors, characterized as the differences between meteorology-corrected trends and emission-driven trends under constant meteorology scenarios, can be reduced by 30 %–42 % using a random forest model that incorporates both local- and regional-scale meteorological features. We further design a correction method based on GEOS-Chem simulations with constant-emission input and quantify the degree to which anthropogenic emissions and meteorological influences are inseparable, due to their process-based interactions. We conclude by providing recommendations for evaluating the impacts of anthropogenic-emission changes on air quality using statistical approaches.

Daily Local-Level Estimates of Ambient Wildfire Smoke PM2.5 for the Contiguous US

“…As higher spatial and temporal densities of training data were predictive of increased model performance (Figure d), future inclusion of these data sources should improve the accuracy of estimates. Other features and data sources such as smoke plume height, spatial lags of meteorology, and indicators of atmospheric mixing such as air temperature at different vertical heights have been found in other settings to improve total PM 2.5 estimates, predict variation in the relationship between PM 2.5 and AOD, or have the potential to improve the model’s ability to identify when smoke mixes to the surface, something the current model occasionally struggles with, as evidenced by the range of predicted values on days with very low observed smoke PM 2.5 values. − Future advances could also include alternative machine learning models, such as convolutional neural networks, that take advantage of the spatial information instead of features at a single point and have been found to provide good performance on total PM 2.5 . While our estimates rely on plume boundaries drawn by NOAA analysts over the contiguous US, automation of plume identificationa task for which early computer vision work has shown promise − could allow for generalization of this approach to other geographic regions, an effort of increasing importance as wildfires grow in many parts of the world. ,− Finally, uncertainty quantification from machine learning models is an active area of research, and future improvements to these estimates could include more granular quantification of uncertainty.…”

Statistical and machine learning methods for evaluating trends in air quality under changing meteorological conditions

“…Code and data availability. The GEOS-Chem simulation of different scenarios and the R scripts to implement the statistical methods to correct for meteorological variability are available at the following repository: https://doi.org/10.5281/zenodo.6857259 (Qiu et al, 2022). All the other data needed to evaluate the conclusions in the paper are present in the paper.…”

Predicting Atmospheric Water-Soluble Organic Mass Reversibly Partitioned to Aerosol Liquid Water in the Eastern United States

“…However, due to the complex nature of atmospheric processes and their nonlinear behavior, these regressions often fail to capture the spatial and temporal variations of these processes with sufficient accuracy. 22 Applying machine learning models in the field of atmospheric science could remedy such drawbacks.…”

“…The use of multivariate linear regressions is deployed extensively in the field of aerosol science because it is a useful and simple tool that provides insight into different factors that may be simultaneously affecting one dependent parameter in the atmosphere (e.g., refs − ). However, due to the complex nature of atmospheric processes and their nonlinear behavior, these regressions often fail to capture the spatial and temporal variations of these processes with sufficient accuracy . Applying machine learning models in the field of atmospheric science could remedy such drawbacks.…”