An experimental study of the dispersion of the emissions from chimneys in reading—III: The investigation of dispersion calculations

Abstract

Six-hr concentrations of sulphur dioxide were measured at 40 sites in Reading during a 15-month period, together with the relevant meteorological data. An inventory of the sources of sulphur dioxide was compiled and the data have been used to test dispersion models for calculating pollution. These models were based upon the dispersion equation for a single chimney. The criterion used for testing the models was the root mean square error (R.M.S.E.) of the differences between pairs of observed and calculated concentrations. This indicates how well the models predict the observed concentrations, but gives no indication of systematic errors in the models. The high ratio of the R.M.S.E. to the mean in all the models made it difficult to investigate systematic errors and it was not possible to distinguish between systematic errors due to the data and those due to the equations in the models. The investigation finally concentrated on examining empirical models containing only the meteorological parameters which correlated with the observed concentrations. In the first series of tests, using pollution data from 6 sites, it was shown that there was no improvement in the estimates of pollution when hourly meteorological data were used instead of 6-hr average data. These tests also compared the effect of using diffusion coefficients from turbulence measurements with that of using values published by Pasquill and the German VDI; the error in the estimates increased in this order. A second model was tested in which housing areas were treated as crosswind line sources and individual sources were treated as single chimneys. This was compared with the previous model, which treated both areas and individual sources as single chimneys. The test was made using only measuring periods in which the wind direction was reasonably steady in order to sharpen the comparison, but there was nothing to choose between the models. A repeat test using periods of less steady wind directions confirmed that, although the errors were greater, they did not differ between the models. The application of these models to all 40 sites using the selected meteorological data showed that the root mean square error of the estimates was roughly equal to the mean of the observed concentrations. In all these tests, regression equations were calculated from the same data, comparing concentration measurements with a parameter derived from air temperature and wind speed. The root mean square errors from these were consistently lower than those from the dispersion models. This suggested formulating empirical equations based upon a similar parameter derived from air temperature and wind speed, but also including various functions of the distribution of sources around each measuring site. These were tested with the same data used for the dispersion models and were found to give better estimates of the observed concentrations. It is concluded that the dispersion equations do not provide a satisfactory basis for calculating the pollution from extended sources such as housing areas. The better precision given by the empirical equations suggests that a model for extended areas in a town should be developed on the same lines, while retaining the dispersion equation for the tall chimneys. In densely populated countries, such as England, any pollution models will have to take account of pollution arriving from other towns.