Impact of assimilated observations on improving tropospheric ozone simulations Palmira Messina a, * , Massimo D’Isidoro b , Alberto Maurizi a , Federico Fierli a a Institute of Atmospheric Sciences and Climate e CNR, Bologna, Italy b National Agency for New Technologies, Energy and the Environment, Bologna, Italy article info Article history: Received 19 April 2011 Received in revised form 17 August 2011 Accepted 19 August 2011 Keywords: Tropospheric ozone Data assimilation Air quality modelling Emission inventories abstract The work aims to evaluate the improvement in the capability of regional models to reproduce the distribution of tropospheric pollutants, using the assimilation of surface chemical observations. In particular, the efficacy in correcting the biases of perturbed emission scenarios was analysed. The study was carried out using the Air Quality Model BOLCHEM coupled with a sequential Optimal Interpolation (OI) routine to perform ozone and nitrogen dioxide assimilation. The OI routine was chosen because it is computationally inexpensive. The work was performed using the Observing System Simulation Exper- iment (OSSE), which allowed the quantification of assimilation impact, through comparison with a reference state. Different sensitivity tests were carried out in order to identify how assimilation can correct perturbations on O 3 , induced by NO x emissions biased in both flux intensity and time. This simple assimilation approach provided a substantial improvement in surface O 3 . It was found to be more effective to assimilate an O 3 precursor, like NO 2 , than O 3 itself, and, in order to obtain a discernible impact on 24-h forecasts, it could be sufficient to assimilate observations when NO x emissions are higher over a 12-h window. It was also found that temporally biased NO x emissions only slightly perturb O 3 . Ó 2011 Elsevier Ltd. All rights reserved. 1. Introduction Assessing and predicting the composition of the lowermost atmosphere are issues of increasing interest to the scientific community. Air Quality Models (AQMs) provide a key instrument for understanding and forecasting chemical pollution (see, for instance, Lawrence et al., 2003; Sportisse, 2007; Zhang, 2008). Nevertheless, AQMs still present discrepancies in predicting chemical fields in comparison with observations, due to uncertainties in input (such as emissions, initial values and boundaries conditions) and gaps in the treatment of chemical and physical processes (Tilmes et al., 2002; Russell and Dennis, 2000). Uncertainties in emissions are an important concern (Sawyer et al., 2000; Simpson et al., 1999; Hanna et al., 1998), since they continue to be high, despite many efforts made to provide a more accurate emissions inventory. Emissions are in general supplied by inventories on annual basis. For air quality modelling purposes, they have to be processed to obtain hourly estimate of the relevant pollutants on a geographical grid. The whole procedure is based on a priori assumptions on daily, weekly and seasonal activity profiles and it can be greatly affected by errors (Tao et al., 2004; Placet et al., 2000; Tilmes et al., 2002; Menut et al., 2000). Hanna et al. (1998) and Mallet and Sportisse (2005) considered that NO x emissions could be affected by an uncertainty of 30%e50%, while Beekmann and Derognat (2003) confirmed this assumption, considering an uncertainty of 40%. Moreover, the impact of the temporal allocation of emissions on pollutant distribution is still under debate: Mallet and Sportisse (2005) found that ozone concentration can be sensitive to it, while Tao et al. (2004) found that daytime O 3 concentrations are slightly dependent on changes in the temporal allocation of NO x emissions. Over the last decade, the great number of observations and advances in spatio-temporal Data Assimilation (DA) methods in atmospheric chemistry have allowed a more efficient use of measurements, thus possibly reducing model uncertainties. Never- theless, the issue of determining an efficient technique for chemical DA remains challenging. Wu et al. (2008) compared different algo- rithms for ozone forecasts (optimal interpolation, reduced-rank square root Kalman Filter, Ensemble Kalman Filter, and 4D-varia- tional assimilation), finding that the Optimal Interpolation (OI) and Ensemble Kalman Filter algorithm (EnKF) have the best perfor- mance, the former during assimilation periods, the latter during forecast. The EnKF is a promising method for chemical DA, (Constantinescu et al., 2007a,b,c; Hanea et al., 2004, 2007). However, since air quality * Corresponding author. E-mail address: p.messina@isac.cnr.it (P. Messina). Contents lists available at SciVerse ScienceDirect Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv 1352-2310/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2011.08.056 Atmospheric Environment 45 (2011) 6674e6681