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

This repo includes the GEOS-Chem simulations and R scripts that are needed to replicate and evaluate the conclusions from Qiu, Zigler, and Selin, ACP, 2022 "Statistical and machine learning methods for evaluating trends in air quality under changing meteorological conditions".

The GEOS-Chem simulations

For the US (2011-2017):

observational_o3_pm_2011_2017_us.rds: the simulated daily PM2.5 and O3 concentrations, and MERRA-2 meteorological features in the observational scenarios (changing meteorology, changing emissions).

counterfactual_o3_pm_2011_2017_us.rds: the simulated daily PM2.5 and O3 concentrations, and MERRA-2 meteorological features in the counterfactual scenarios (constant meteorology, changing emissions).

constant_emis_o3_pm_2012_2017_us.rds: the simulated daily PM2.5 and O3 concentrations in the constant-emission scenarios (constant meteorology, constant emissions).

regional_features_2011_2017_4x5_us.rds: the MERRA-2 meteorological features in the observational scenarios (aggregated to 4x5 degrees), inputs for the "RF-regional" model.

For China (2013-2017):

observational_o3_pm_2013_2017_china.rds: the simulated daily PM2.5 and O3 concentrations, and MERRA-2 meteorological features in the observational scenarios (changing meteorology, changing emissions).

counterfactual_o3_pm_2013_2017_china.rds: the simulated daily PM2.5 and O3 concentrations, and MERRA-2 meteorological features in the counterfactual scenarios (constant meteorology, changing emissions).

constant_emis_o3_pm_2014_2017_china.rds: the simulated daily PM2.5 and O3 concentrations in the constant-emission scenarios (constant meteorology, constant emissions).

regional_features_2013_2017_4x5_china.rds: the MERRA-2 meteorological features in the observational scenarios (aggregated to 4x5 degrees), inputs for the "RF-regional" model.

R scripts:

main.r: the main script to perform statistical correction of meteorological variability.

main.r uses functions from the other R script files (see below) which perform different statistical correction methods, respectively.  

parametric_regression_methods.r: performs meteorological correction with parametric regression methods (MLR, polynomial, spline, GAM)

tune_RF_regional.r and RF_regional.r: perform the meteorological correction with the "RF-regional" model

GEOS_Chem_constant_emis.r: performs the meteorological correction using the simulations from the constant emission scenarios from the GEOS-Chem model