Modelling trends in OH radical concentrations using generalized additive models 1Environment Dept., University of York, York, YO10 5DD, UK
2Institute for Transport Studies, University of Leeds, Leeds, LS2 9JT, UK
*now at: School of Earth and Environment, University of Leeds, Leeds, LS2 9JT, UK
Received: 17 Jun 2008 – Published in Atmos. Chem. Phys. Discuss.: 31 Jul 2008 – Published: 20 Mar 2009Abstract. During the TORCH campaign a zero dimensional box model based on the Master
Chemical Mechanism was used to model concentrations of OH radicals. The model
provided a close overall fit to measured concentrations but with some
significant deviations. In this research, an approach was established for
applying Generalized Additive Models (GAM) to atmospheric concentration data.
Two GAM models were fitted to OH radical concentrations using TORCH data, the
first using measured OH data and the second using MCM model results. GAM
models with five smooth functions provided a close fit to the data with 78%
of the deviance explained for measured OH and 83% for modelled OH. The GAM
model for measured OH produced substantially better predictions of OH
concentrations than the original MCM model results. The diurnal profile of OH
concentration was reproduced and the predicted mean diurnal OH concentration
was only 0.2% less than the measured concentration compared to 16.3%
over-estimation by the MCM model. Photolysis reactions were identified as
most important in explaining concentrations of OH. The GAM models combined
both primary and secondary pollutants and also anthropogenic and biogenic
species to explain changes in OH concentrations. Differences identified in
the dependencies of modelled and measured OH concentrations, particularly for
aromatic and biogenic species, may help to understand why the MCM model
predictions sometimes disagree with measurements of atmospheric species.
Citation: Jackson, L. S., Carslaw, N., Carslaw, D. C., and Emmerson, K. M.: Modelling trends in OH radical concentrations using generalized additive models, Atmos. Chem. Phys., 9, 2021-2033, doi:10.5194/acp-9-2021-2009, 2009.