Evaluation of a model for hourly spatial concentration distributions

Abstract

A time-dependent finite difference model in three levels combined with a puff model to account for subgrid effects close to single sources was used to calculate hour-to-hour concentration distributions. Measurements from several selected stations were used to account from time variation in background concentrations. For each hour, weight was given to observed values in areas that were not influenced by local sources.

Results of concentration calculations based on hourly data on emission and dispersion are validated by measured time series of SO₂ and NO_x at five stations. A combination of hourly nephelometer readings and 12-h measurements of small particles at three stations are compared with calculated values.

Hourly observed and calculated values from two periods (3 January–15 March 1988 and 18 April–24 June 1988) were used for the evaluation of the model for calculating hourly pollution concentrations in each square kilometre. The results showed that prediction of short-term average concentrations (e.g. hourly data) are usually poorly correlated with observations at the same time and location. Slight displacement errors may cause point-to-point correlation to be poor as a result of errors in input data.

The pattern of NO_x concentration variation with time was reproduced well at all stations. A subgrid model taking into account the influence of nearby roads would probably improved the model for NO_x at some stations. For SO₂ and small particles, industrial sources have the dominant influence and the correspondence between observed and calculated values were improved by taking into account spatial uncertainty and an hourly variation in background concentrations.