Abstract

Carbon monoxide (CO) is a major pollutant, impacting air quality and contributing to the greenhouse effect. Buchholz et al. [1] used a multiple linear regression model to link CO anomalies in the atmosphere to variability in the climate. Measurements of total column CO were retrieved from the Mea... Show moreCarbon monoxide (CO) is a major pollutant, impacting air quality and contributing to the greenhouse effect. Buchholz et al. [1] used a multiple linear regression model to link CO anomalies in the atmosphere to variability in the climate. Measurements of total column CO were retrieved from the Measurements Of Pollution In The Troposphere (MOPITT) instrument onboard the Terra satellite. Anomalies generated by these measurements are used as the response variable in this model. Climate mode indices represent regional variability in the climate, and these indices at various time lags are used as the predictor variables. Buchholz et al. [1] performed this analysis in MATLAB using serial algorithms in non-functionalized scripts. Simonson et al. [2] refactored the MATLAB codebase, improving both code interpretability and generality and adding parallelization options. In this technical note, we present further updates to the codebase used in the atmospheric CO analysis from Buchholz et al. [1]. These updates stem from four primary objectives. First, transfer the codebase into an R package that is available to researchers without a MATLAB license. Second, allow for a single climate index to appear in the model multiple times with different lag values. Third, implement an exhaustive variable selection algorithm. Finally, improve upon the non-exhaustive variable selection algorithm implemented in the MATLAB code. These updates are discussed, and the accuracy and timing of the new variable selection algorithms are compared. Show less