Use this URL to cite or link to this record in EThOS:
Title: Sensitivity and uncertainty analysis of episodic ozone predictions from the community multiscale air quality model
Author: Beddows, Andrew Victor
ISNI:       0000 0004 5917 9228
Awarding Body: King's College London
Current Institution: King's College London (University of London)
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Access from Institution:
The first global variance-based sensitivity analysis of ozone and NO2 concentrations produced by the CMAQ model during the July 2006 ozone pollution episode over the UK has been performed. Gaussian process emulation methods have been employed to overcome the problems caused by long model run times which have previously prevented such analyses being undertaken. The computationally efficient Morris' method was used to rank the effect of perturbations in 223 model input variables, including all of the gas-phase species in the model domain chemical boundary conditions and emissions, and all of the reaction rates in the carbon bond five core chemical mechanism. The 30 most influential variables were combined with ozone deposition velocity to emulate the effects of perturbations in 31 input variables on the modelled concentrations of ozone and NO2. These emulators were then used in place of CMAQ in Fourier amplitude sensitivity tests, which decompose the variance induced in model output when all of the inputs are perturbed together into contributions from each of those inputs. These tests were performed for every hour of a 21 day time series spanning the episode and several days either side, for a number of locations around the UK. The results reveal a complex spatio-temporal pattern of model sensitivities, with NO and isoprene emissions, NO2 photolysis and ozone deposition velocity and boundary conditions being amongst the most inuential input uncertainties. The same emulators were used in a Monte Carlo uncertainty analysis of modelled ozone concentrations. The results of this analysis were used with a simple Bayesian weighting procedure to calibrate the model inputs, which led to a significant improvement in peak afternoon ozone predictions. Calibrated UK and EU NO and NO2 emissions were between 1.27 and 1.38 times the baseline values, suggesting that official NOx emissions totals may be substantially underestimated.
Supervisor: Beevers, Sean David Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available