Title of Example

Example

Introduction

The dispersion model used by Birmingham City Council in the Review and Assessment of air quality is the INDIC Airviro model, which is licensed by the Swedish Meteorological Office (SMHI).

Airviro is a complete Air Quality Management System. It includes the following functional blocks:

• Emission surveying and modelling within the Emission Database (EDB)
• Dispersion modelling with the Dispersion module
• Monitoring data collection, analysis and presentation with the Indico package.

This example considers the effects that meteorology data can have on the output of the model. The example highlights the importance of considering carefully what meteorological data should be used when conducting dispersion modelling to predict pollution levels for future years, and the impact it can have on final conclusions and often important decisions, for example the declaration of Air Quality Management Areas.

Meteorology Data

The meteorology data input was supplied and validated by SMHI. In addition three – hourly synoptic data interpolated to hourly values was received from Birmingham Airport for the period June 1992 – December 1998. The model was run for 2 years (1996 and 1998) with contrasting meteorology so that the effect of annual variability could be considered. Meteorologically 1996 is considered as ‘stable’, that is to say it is typified by lower wind speeds. Whereas UK weather patterns are typically more blustery like those seen in 1998. Blustery conditions lead to ‘unstable’ conditions and aid pollution dispersion; therefore 1998 was used as ‘typical’ dispersion conditions and 1996 as ‘worst case’ dispersion conditions.

The work carried out allowed for consideration of meteorological influences on the output of the model. The Gauss ‘time series’ simulations for 1996 and 1998 use the same base EDB, therefore the only input variation to the model is meteorology. An investigation into each year’s meteorology finds differences, particularly in the wind speed. Figure 1 illustrates the different wind speeds during 1996 and 1998. During 1996 the hourly wind speed was less than 5m/s for 75% of measurements. The mean and maximum wind speeds recorded were 3.7m/s and 12.5m/s respectively. In contrast 1998 saw only 22% of the hourly wind speed values measure less than 5m/s. The mean and maximum wind speeds in 1998 were 8.3m/s and 30.0m/s respectively.

Figure 1

Atmospheric pollution dispersion is aided by wind, therefore with the higher wind speeds, as in 1998; improved air quality is seen if all other factors remain constant

Modelling results for primary pollutants, CO and NOx show that the model over-predicts the annual average concentrations for 1996 by the same degree it under- predicts in 1998, but the correlation was acceptable. For single hourly values the predictions are less reliable, particularly for high percentile values. As illustrated by Figures 2 to 5 the highest pollutant concentrations are predicted by Airviro at the lowest wind speeds with a uniform reduction as wind speed increases.

Figure 4 and 5 show monitored and modelled NOx concentrations against wind speed at a monitoring station in the Birmingham area. At wind speed 1-2m/s the model over-estimated measured values in both 1996 and 1998. The monitoring station has a sample inlet at 15m whereas the dispersion model calculates concentration at height of 2m. It could be argued that with the measurement made at 15m, the pollutant has time to disperse and dilute, and concentrations will be lower than those modelled closer to the ground, however the effect is seen at other locations where measurements are made at 3m.

Figure 2

Figure 3

Figure 4

Figure 5

Figures 4 and 5 for Birmingham East illustrate that this over estimate also occurs for CO. This over estimate at low wind speeds is typical but it is not seen at all sites.

As Figure 1 shows, 1996 was characterised by low wind speeds, which leads to stable weather conditions and suppresses dispersion. The model anticipates the stable conditions and predicts appropriately, however, as figures 2 to 5 show for several sites, these predictions are over estimates. Therefore running the Gauss model as a ‘time series’ may have possible limitations for any single hours data when the wind speed is low.

Predictions of Future Air Quality

Having completed validation work on the Gauss model and investigated its possible limitations as a predictive tool, it was then possible to make predictions of future air quality. Predictions were run using the Gauss ‘time series’ model, and applying the ‘future edb’ to both 1996 and 1998 meteorogical data.

For example, Tables 1 and 2 below show NO₂ modelling predictions for base and future years with a comparison to any measured data available. Table 1 shows predictions made using the 1996 meteorological data with the 'base edb' to obtain a 1996 prediction and with the 'future edb' to generate a 2005 prediction and is shown against 1996 measured values. Table 2 shows predictions made using the 1998 meteorological data with the 'base edb' to obtain a 1998 prediction and with the 'future edb' to generate a 2005 prediction and is shown against 1998 measured values. Modelling was carried out using the Gauss model and running the time-series set-up, generating hourly NO_X data for the year(s) of interest. Predicted hourly NO₂ values were calculated from the simulated hourly NO_X data using the Derwent-Middleton equation. A correction factor of 0.9 was applied to the calculated NO₂ to compensate for the under-estimation, and then percentile values were obtained.

Table 1: Predictions of NO₂ concentration for 1996 and 2005 modelled using 1996 meteorological data

Table 2: Predictions of NO₂ concentration for 1998 and 2005 modelled using 1998 meteorological data

Automatic monitoring for the above sites in Tables 1 and 2 shows that the 1 hour mean objective of 200mg.m^-3 (105 ppb), not to be exceeded more than 18 times a year, is met at all sites. The predictions for base year show exceedence at several locations in 1996 and at Birmingham Centre in 1998. However no exceedences are predicted for 2005.

The results of automatic monitoring in the West Midlands have shown a progressive reduction in the annual average level of NO₂ between 1993 and 1999. Thus, whilst the annual average exceeded the objective, of 40mg.m^-3 (21ppb), at four of the six operational stations in 1996, the objective was met at all eleven operational stations in 1998. The model also predicts exceedence during 1996 and achievement in 1998. For future predictions the model predicts exceedence of the annual average at several locations when using the 1996 meteorological data set but compliance when using the 1998 meteorological data set, as shown in Table 3. Those sites that are predicted to exceed or fall close to the objective, Birmingham Centre, Birmingham West, Walsall Alumwell are highlighted in Table 3 in red text.

Table 3: Comparison of NO₂ Annual Mean concentrations for 1996 and 1998

The modelled results obtained are greatly dependent on the meteorological data set used. Predictions for 1998 are closer to the 1998 measured values than the 1996 predictions are to the 1996 measured values. When using the 1996 meteorology with the base EDB the model over-predicts the measured values at all six stations, for which monitored data exists.

Shown in Table 3 is a column titled "% Reduction in Prediction". For the annual mean based on 1996 meteorological data, this value is the reduction in NO₂ between 1996 and 2005 shown as a percentage of the 1996 prediction. A similar calculation was carried out for the annual mean based on the 1998 meteorological data. This predicted reduction in pollution levels is a result of differences in the base and future EDB's.

Treatment of Uncertainty

There are many sources of uncertainty when using computer modelling. The output of any modelling study is dependant on the input data. The greatest uncertainty is with modelling of future events where it is not possible to foresee all statutory, regulatory, economical and technological factors and their impact on air quality. Therefore, it is generally the case that when compiling emission databases for future years the ‘worst case’ options should be selected. As it is impossible to predict the weather conditions for any future year, to err on the side of caution a ‘worst case’ meteorological data set was used in the modelling as well as a ‘typical’ data set.

Discussion and Conclusions

· Birmingham City Council used the Airviro ‘Gauss’ model, to calculate and predict for both the present time and for 2005 levels of various pollutants. The model has been used to calculate two sets of predictions. Firstly, a ‘time series’ of hourly values over a full calendar year, from which average and high percentile values could be derived for specific points, and secondly, annual ‘scenario’ calculations.

· From the 1998 emissions database ‘time series’ modelling was used to find the 18^th highest hourly and annual average concentrations for two sets of meteorological data;1996 and 1998. These years represent the ‘worst’ and ‘best’ case weather conditions, with respect to air pollution. The weather in 1996 was characterised by a number of periods of high pressure with low winds and stable conditions, which favour the accumulation of air pollutants, whilst the weather in 1998 was characterised by winder conditions, which favour the dispersion of air pollutants.

· For nitrogen dioxide the 2005 emissions database was then used to calculate the 18^th highest hourly and annual average concentrations, using the 1996 meteorological data, representing the ‘worst case’ for pollution values. These predictions were used to assess whether the respective nitrogen dioxide objectives were likely to be met in 2005.

· The modelled predictions for the 18^th highest hourly concentrations, from the time series calculations, are presented in Table 4, below, together with the comparable values from two roadside monitoring stations. The predictions for the 1998 EDB agree very well with the measured values at the Birmingham East AURN station, when using the 1998 meteorological data. They are approximately twice the values at the Birmingham Centre AURN station, when using either the 1996 or the 1998 meteorological data. But they over read the measured values threefold at the Birmingham East AURN station, when using the 1996 meteorological data.

Note 1: Air quality objective: 18^th highest hourly value not to exceed 200g/m³ in 2005.

Note 2: the measured values refer to either 1996 or 1998 depending on the met year used

Table 4: Comparison of 18^th Highest Hourly Nitrogen Dioxide Values

(all values quoted in µg/m³)

· The predictions for both Birmingham Centre and Birmingham East stations are considerably higher than the respective measured values due largely to difficulties the model has in simulating dispersion for single hours, at low wind speeds. The effect is particularly noticeable at the Birmingham East station, which may be because of the nature of the residential area.

· The modelled predictions for the annual average concentrations, from the time series calculations are presented in Table 5, below.

· The predictions for the 1998 EDB agree very well with the measured values at both Birmingham Centre and Birmingham East AURN Stations, when using the 1998 meteorological data. However, the predictions are approximately 40% higher than the measured values at the Birmingham Centre AURN station, and 30% higher than the measured values at the Birmingham East AURN station when using the 1996 meteorological data.

Note: Air quality objective: annual average value not to exceed 40mg/m³ in 2005.

Note 2: the measured values refer to either 1996 or 1998 depending on the met year used.

Table 5: Comparison of Annual Average Nitrogen Dioxide Values

(all values quoted in µg/m³)

The work carried out by Birmingham City Council to investigate how different metrological data influences the output of the model is demonstrated above and shows how it can have an impact when making final and often important decisions. For example, if the model had only been run using 1996 meteorological data then in some instances the model indicates that objectives would be exceeded whereas when using the 1998 meteorological data, objectives were predicted to be achieved. It is therefore important to know whether the meteorological data used in the model is ‘best case’ or ‘worse case’ and whether it is representative of a typical year.

It is good practice to run any model for the ‘worst case’ scenario, which in this case was 1996 meteorological data, (even though it was not classed a ‘typical’ year), if objectives are predicted to be achieved by the model in the worse case then it is highly likely that they will be achieved in the ‘best case’ scenario.

Last Updated

13th January 2005