Assessing air-quality forecasting for London
An assessment of local and regional air-quality forecasts for London, UK is reported in the International Journal of Environment and Pollution. In their evaluation, Amy Stidworthy, Mark Jackson, Kate Johnson, David Carruthers, and Jenny Stocker, of Cambridge Environmental Research Consultants Ltd., explain how levels of nitrogen oxides, ozone, and sooty particulates (PM2.5 and PM10), can be predicted quite accurately but the origin and how the various pollutants are dispersed must be taken into account in forecasts to ensure accuracy.
Atmospheric dispersions models are critical to accurate, long-range predictions of air quality. Such forecasts inform the public and others of how pollution levels are shifting over the coming days and can be used to advise people with reduced lung function and other medical conditions sensitive to high pollution levels on whether or not to avoid certain areas at certain times or even to stay indoors entirely.
The researchers add that "Forecasting systems must account for long-range transport of pollutants in addition to local emissions, chemical processes and urban morphology; thus it is common practice to couple local air dispersion models with regional models to account for pollutant emissions, transport and chemistry at a range of scales." As such, their evaluation of the commonly used system of London's airTEXT, which uses CAMS regional ensemble air quality forecast data, and the ADMS-Urban dispersal model to make accurate predictions is important both for ensuring that guidance is appropriate.
The team showed that this system performs better than the regional-scale CAMS forecasts for all pollutants considered, with the exception of PM2.5. However, there were no major air-quality incidents during the study period so the absolute predictive power could not be determined. Prediction of atmospheric nitrogen dioxide levels was much better in urban areas with this system as one might expect given that the main source is road traffic. Ozone levels are a secondary pollutant and so levels depend heavily on dispersal.
The team adds that "conservative" and "cautious" alerts should be considered and data points on the threshold might be removed to avoid bias and so improve accuracy still further, although this would only be sensible with very long data sets covering a significant time period.