The performance evaluation of WinOSPM model for urban street canyons of Nantes in France

Abstract

Air quality modelling is primarily the quantative approach. It is more difficult as it demands input data accuracy, uncertainties and the efficient methodologies to judge the extent of models accuracy. As a result, model validation has to be regarded as an integral part of the modelling process. Furthermore, models are often validated on a limited number of testcases therefore, appropriate evaluation procedure must be implemented to ensure these models will be applicable for various conditions. The study presented here was carried out to evaluate the WinOSPM (Preliminary version of windows based Operational Street Pollution Model) for air pollutants viz. CO, NO, NO2, NOx and C6H6 for three street canyons of Nantes (France) and for the three base years 1999, 2000, and 2001. Each street canyon selected for this study has typical and unidentical features. The rue de Strasbourg and Boulevard Victor Hugo have many building exceptions whereas rue Crébillon has not any. Application of the model above to the three street canyons revealed that WinOSPM could be used in the case when measurements are not available. This was justified from the results at rue Crébillon. The special interest was in the benzene modelled values as its content in fuel has been targeted to reduce to 1% for the years 2000 and onwards (from its 5% until the year 1999). The 50 to 70% reduction in the benzene concentrations is found for both the years i.e. in 2000 and 2001. This has further justified that air quality models are useful and interesting tools in optimising emission reduction strategies. Moreover, it is also the new pollutant added to the measurement campaign of Air Pays de la Loire (APL) for the city of Nantes. For benzene weekly averages are estimated from the hourly-modelled values for all the streets and compared with that of measurements. They are found in excellent agreement with each other's. For other pollutants annual means and percentiles were compared. The statistical analysis was done to evaluate the models performance using index of agreement and correlation coefficient. The index of agreement (d) and correlation coefficient (r) for all the streets show that estimated concentration levels are in good agreement with that of measurements. From the index of agreements, it can be inferred that model has very less potential for errors. The model's sensitivity to building-exceptions was also tested for the rue de Strasbourg. Results did not reflect this feature very well. It is perceived that the influence of this feature might have been suppressed in averaging the annual hourly values. This influence is apparently seen in hourly average time series variations. Finally, WinOSPM model was found a simple but very useful model. It could very well represent the detailed flow and dispersion conditions in urban streets.