Efficient Sensitivity Computations In 3d Air Quality Models
Kioutsioukis I., D. Melas, C. Zerefos and I. Ziomas
[abstract] Photochemical Air Quality Models (PAQM) predict species concentrations based on a series of processes (dynamical, physical, chemical and radiative) and data describing the status of the domain (like the boundary conditions and the emissions). The overall modelling procedure has an inherent stochasticity arising from either the simplification of the various processes or the data estimates. However, comprehensive sensitivity analysis of the results of 3-Dimensional (3D) PAQM calculations is not, in general, possible - mainly because of the associated computational cost, human effort and data needs. This study utilises a general approach, based on perturbation theory, to facilitate the sensitivity analysis of a PAQM. Specifically, we apply an automatic differentiation tool (ADIFOR) in the 3D PAQM CAMx in order to calculate, in only one run, the gradient of the solution vector with respect to its input parameters. To demonstrate the applicability of the methodology, we address the sensitivity of the modelled concentrations to perturbations at the boundary conditions and the emissions, for three European Metropolises with fundamentally different characteristics (Athens, Milan, and London). Our results for the urban centres demonstrate that the maximum change in O3 (around midday) arises mainly from perturbations at the boundary conditions of O3 and the area emissions of NOX.
Ελληνική έκδοση
back to publications list
