Atmospheric Environment 40 (2006) 3737–3749 Evaluation of the GEM-AQ air quality model during the Que´ bec smoke event of 2002: Analysis of extensive and intensive optical disparities N.T. O’Neill a,Ã , M. Campanelli b , A. Lupu c , S. Thulasiraman a , J.S. Reid d , M. Aube´ a , L. Neary c , J.W. Kaminski c , J.C. McConnell c a CARTEL, De´partement de ge´omatique applique´e, Universite´de Sherbrooke, Sherbrooke, Que´., Canada J1K 2R1 b Institute of Atmospheric Sciences and Climate, Area della Ricerca di Torvergata, Via Fosso del Cavaliere 100, 00133 Rome, Italy c CRESS, Department of Earth and Space Science and Engineering, York University, 4700 Keele Street, Toronto, Ont., Canada M3J 1P3 d Naval Research Laboratory, 7 Grace Hopper Avenue, Stop 2, Monterey 93943-5502, CA, USA Received 31 October 2003; received in revised form 28 February 2006; accepted 4 March 2006 Abstract The root-mean-square (rms) differences between the Canadian air quality model GEM-AQ and measurements for intensive and extensive optical variables (aerosol optical depth or AOD and A ˚ ngstro¨m exponent or a) were investigated using data from the July 2002 Que´bec smoke event. In order to quantify regional differences between model and measurements we employed a three component analysis of rms differences. The behaviour of the two absolute amplitude rms components of AOD (difference of the means and the difference of the standard deviations) enabled us to infer emission properties which would otherwise have been masked by the larger ‘anti-correlation’ component. We found the inferred emission fluxes to be significantly higher than the original geostationary, satellite-derived FLAMBE ´ (fire locating and modelling of burning emissions) emissions flux estimates employed as inputs to the simulations. The model captured the regional decrease of the intensive a exponent (increase of particle size with trajectory time), while the agreement with the extensive AOD parameter was marginal but clearly dependent on the nature of the spatio-temporal statistical tools employed to characterize model performance. In establishing the a versus trajectory time trend, the modelled AOD data was filtered in the same way as the measured data (very large AODs are eliminated). This processing of modelled results was deemed necessary in order to render the a results comparable with the measurements; in the latter case it was difficult, if not impossible, to discriminate between measured a trends due to instrumental artifacts (non-linearities at low signal strength) versus trends due to coagulative effects. r 2006 Elsevier Ltd. All rights reserved. Keywords: Forest fire; Aerosol optical depth; A ˚ ngstro¨m exponent 1. Introduction Model evaluation is a multi-disciplinary research activity which ranges from qualitative ‘eyeball’ comparisons to sophisticated spatio-temporal ARTICLE IN PRESS www.elsevier.com/locate/atmosenv 1352-2310/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2006.03.006 Ã Corresponding author. E-mail address: Norman.T.ONeill@USherbrooke.ca (N.T. O’Neill).