4 Analysis Of Air Pollution Forecasting Success RateAEAT-3844

4.1 Introduction
Air quality in the UK is generally within the VERY GOOD/ LOW or GOOD/ MODERATE air quality bands. Episodes of POOR/HIGH or VERY POOR/VERY HIGH air quality are, however, also experienced. Episodes during the winter are generally associated with periods of poor pollutant dispersion caused by low temperatures and light winds (see for example Bower et al, 1994). Air pollution episodes during the summer are often photochemical in nature and are associated with light winds, high temperatures and strong sunlight.

The forecasts that are issued tend to err on the side of caution and consequently more occurrences of POOR/ HIGH air quality were predicted during 1997 than were measured. This is because it is considered by DETR to be better for public information and health, to predict POOR/ HIGH and be wrong than predict GOOD/ MODERATE and fail to warn the public of potentially POOR/ HIGH air quality.

The forecasting success rates are discussed in two sections: the old bandings in Section 4.2 and the new bandings, including forecasts of CO and particles, in Section 4.3.

4.2 Forecast Analysis For Old Bandings, 1 January 1997 To 18 November 1997
The forecasting success rates for ozone, nitrogen dioxide and sulphur dioxide are presented in Tables 4.1, and respectively. Each table gives an analysis of forecast versus measured POOR air quality during 1997. The air quality band for each region is the band for the highest concentration measured at any monitoring site within that region.

There was one measurement of VERY POOR air quality for SO2 during the period covered by this report. There were no measurements of VERY POOR air quality for O3 or NO2 . For the tables of forecast analysis, VERY POOR air quality has been included in the totals of POOR air quality. The instance of VERY POOR air quality is discussed in the text.

4.2.1 Forecast Analysis for Ozone
Twenty five of the 28 "region-days" with POOR air quality for ozone were correctly forecast (89%, see Table 4.1). All the instances of POOR air quality were correctly forecast for SE England, the Midlands and London, and 5 out of 6 days were correctly predicted in NE England.

A cautious approach by the forecasters resulted in this very high level of achievement, but also led to a number of region-days being forecast which did not occur. A significant number of these were 'near-misses' where the 90 ppb threshold just failed to be breached. There were usually exceedances of EPAQS recommended guideline on these days, with the 8-hour running mean exceeding 50 ppb, the National Air Quality Standard, and this would be considered MODERATE air pollution within the new bandings.

The frequency of days with POOR air quality for ozone during 1997 was lower than in 1996, and much lower than the more photochemically active year of 1995, which had 57 episode days (Stedman and Willis, 1996).

VERY POOR air quality for ozone was neither forecast nor measured during 1997.

Table 4.1 Forecast Analysis for Ozone, 1 January 1997 to 18 November 1997
  NW England NE England Midlands SW England SE England London N Ireland Scotland Wales UK
POOR days measured 1 6 4 1 8 7 0 0 1 27
POOR days forecast 5 19 17 15 28 12 3 5 5 96
forecast & measured 0 5 4 0 8 7 0 0 1 24
forecast & not measured 5 14 13 15 20 5 3 5 4 72
not forecast & measured 1 1 0 1 0 0 0 0 0 3

4.2.2 Forecast Analysis for Nitrogen Dioxide

Forty seven out of a total of 95 "region days" (50%) with POOR air quality for nitrogen dioxide were correctly forecast during this period (see Table 4.2). Fifty four out of a national total of 95 "region days" with POOR air quality for nitrogen dioxide were measured in London, of which 31 days (57%) were correctly forecast. In the Midlands three out four days were correctly forecast while, while in NE England only 5 out of 16 days were forecast. A large proportion of these episodes were recorded at Lincoln Roadside, a new site on a very busy road with a high-sided street canyon. This site appears to show higher concentrations than ever measured before at a non-London roadside site. Consequently a period of "hands-on" experience was required before the forecasters could accurately predict the NO2 levels at this site.

VERY POOR air quality for NO2 was not forecast or measured during this period.



Table 4.2 Forecast Analysis for Nitrogen Dioxide, 1 January 1997 to 18 November 1997
  NW England NE England Midlands SW England SE England London N Ireland Scotland Wales UK
POOR days measured 7 16 4 10 1 54 0 3 0 92
POOR days forecast 20 17 13 25 4 62 3 5 0 141
forecast & measured 4 5 3 4 0 31 0 0 0 47
forecast & not measured 16 12 10 21 4 31 3 5 0 94
not forecast & measured 3 11 1 6 1 23 0 3 0 45
        

4.2.3 Forecast Analysis for Sulphur Dioxide

Table 4.3 Forecast Analysis for Sulphur Dioxide, 1 January 1997 to 18 November 1997
  NW England NE England Midlands SW England SE England London N Ireland Scotland Wales UK
POOR days measured 1 11 1 0 4 2 13 0 0 19
POOR days forecast 1 11 1 0 1 5 37 0 0 19
forecast & measured 0 2 0 0 0 1 6 0 0 3
forecast & not measured 1 9 1 0 1 4 31 0 0 16
not forecast & measured 1 9 1 0 4 1 7 0 0 16
        

VERY POOR air quality for SO2 was measured on 5/11/97, at Middlesbrough, with a maximum hourly average concentration of 450 ppb. This was due to some local residents celebrating Bonfire Night by burning a pile of old tyres next to the monitoring station. The exact location and duration of the event was not predicted by the forecasters! SO2 concentrations were low at all other monitoring sites due to the relatively efficient dispersion of the emissions from bonfires not located directly next to monitoring sites.

Nine out of a total of 34 "region-days" (26%) with POOR air quality for sulphur dioxide were correctly forecast (see Table 4.3). This is lower than the number of POOR days recorded during the whole of 1996. More than a third of POOR days measured were recorded in Northern Ireland: 13 days with POOR air quality for SO2 were measured, of which 6 were forecast. In NE England 11 episode days were recorded of which only 2 was forecast (18%, or 20% if we exclude the VERY POOR), while in SE England none of the 4 episode days measured at Rochester were forecast. This site tends to be affected by a number of nearby point sources, such as refineries and power stations, whose effects are harder to predict. Research work is currently under way to attempt to improve our methods for forecasting point source related SO2 episodes (see Section 6).

4.3 Forecast analysis for NEW bandings 19 november 1997 to 31 december
The forecasting success rates for nitrogen dioxide, sulphur dioxide, fine particles and carbon monoxide are presented in Tables 4.4, and 4.7 respectively. Each table gives an analysis of forecast versus measured HIGH air pollution during the last part of 1997. The air pollution band for each region is the band for the highest concentration measured at any monitoring site within that region.

There were no measurements of VERY HIGH air pollution for any of the pollutants during the period covered by the new set of bandings. There were no measurements of HIGH air pollution for ozone during this winter period.

4.3.1 Forecast Analysis for Nitrogen Dioxide
Neither VERY HIGH nor HIGH air pollution for NO2 was measured or forecast during the last part of 1997.

Table 4.4 Forecast Analysis for Nitrogen Dioxide, 19 November 1997 to 31 December 1997
  NW England NE England Midlands SW England SE England London N Ireland Scotland Wales UK
HIGH days measured 0 0 0 0 0 0 0 0 0 0
HIGH days forecast 0 0 0 0 0 0 0 0 0 0
forecast & measured 0 0 0 0 0 0 0 0 0 0
forecast & not measured 0 0 0 0 0 0 0 0 0 0
not forecast & measured 0 0 0 0 0 0 0 0 0 0

4.3.2 Forecast Analysis for Sulphur Dioxide

VERY HIGH air pollution for SO2 was not measured in the last part of 1997 and HIGH was recorded once in Northern Ireland, which was not correctly forecast.

Table 4.5 Forecast Analysis for Sulphur Dioxide, 19 November 1997 to 31 December 1997
  NW England NE England Midlands SW England SE England London N Ireland Scotland Wales UK
HIGH days measured 0 0 0 0 0 0 1 0 0 0
HIGH days forecast 0 0 0 0 0 0 0 0 0 0
forecast & measured 0 0 0 0 0 0 0 0 0 0
forecast & not measured 0 0 0 0 0 0 0 0 0 0
not forecast & measured 0 0 0 0 0 0 1 0 0 0
        

4.3.3 Forecast Analysis for Particles (PM10)

VERY HIGH air pollution for particles was not measured during the last part of 1997. Of the six episode days measured across the country two were correctly forecast (33% success rate), one each in Scotland and Northern Ireland. The two days of HIGH air pollution in Wales were both recorded at Port Talbot, which is situated close to a steel works and therefore susceptible to local pollution which it is difficult to forecast.

Table 4.6 Forecast Analysis for Fine Particles, 19 November 1997 to 31 December 1997
  NW England NE England Midlands SW England SE England London N Ireland Scotland Wales UK
HIGH days measured 0 0 0 0 0 0 2 2 2 0
HIGH days forecast 0 0 0 0 0 0 1 1 0 0
forecast & measured 0 0 0 0 0 0 1 1 0 0
forecast & not measured 0 0 0 0 0 0 0 0 0 0
not forecast & measured 0 0 0 0 0 0 1 1 2 0
        

4.3.4 Forecast Analysis for Carbon Monoxide

Table 4.7 Forecast Analysis for Carbon Monoxide, 19 November 1997 to 31 December 1997
  NW England NE England Midlands SW England SE England London N Ireland Scotland Wales UK
HIGH days measured 0 0 0 0 0 0 0 0 0 0
HIGH days forecast 0 0 0 0 0 0 0 0 0 0
forecast & measured 0 0 0 0 0 0 0 0 0 0
forecast & not measured 0 0 0 0 0 0 0 0 0 0
not forecast & measured 0 0 0 0 0 0 0 0 0 0
        
Neither VERY HIGH nor HIGH air pollution for carbon monoxide was measured or forecast during the last part of 1997.

4.4 Comparison With Previous Years>
Table 4.8 shows the forecasting success rate for the whole UK over the last four years. This is the percentage of episode days which were correctly forecast The number of days with POOR air quality for ozone that were correctly forecast was 89%, a small improvement on 1995 (82%), though not as good as 1996 (100%). The success rate for forecasting of POOR NO2 was lower than in 1996, though about the same in London (57%). This may be due, in part, to an increase in the number of monitoring sites. The success rate for SO2 in 1997 was lower than in 1996, both in Northern Ireland and across the country as a whole.

Table 4.8 Forecasting Success Rates for the Whole of the UK
  93/94 95 96 97 (old) 97 (new)
Ozone 69% 82% 100% 89% *
Nitrogen Dioxide 32% 70% 52% 49% *
Sulphur Dioxide 27% 42% 38% 28% 0%
Carbon Monoxide - - - - *
Particles - - - - 33%
* No HIGH or VERY HIGH days recorded.

The number of HIGH days recorded in the last part of 1997 indicates the drop in episode days following the move to the new banding structure. In particular nitrogen dioxide episodes were very rare (none in period included) and sulphur dioxide episodes were also less frequent (only one in this period).
  INDEX  
To other DETR publications