How good is the Nowcast?

The Nowcast combines annual modelled pollution concentrations with current pollution measurements to provide a visualisation of air pollution across London each hour. The performance of the NowCast varies each hour and for each pollutant reflecting changes in pollution sources. The current performance of the Nowcast is shown in the table below:

Pollutant	N	R2	Confidence interval µg/m3	Index interval µg/m3	Banding interval µg/m3
Nitrogen Dioxide	44	0.44	+/-21	66	199
PM10 Particulate	27	0	+/-7	10	33

Key and explanation

• N is the number monitoring sites used to make the current Nowcast.
• R² indicates the how well the current Nowcast map captures the variation in concentrations across all the monitoring sites. 1 is a perfect match and zero is no match at all. Values greater than 0.5 are considered good. This number is often lower if the current measurements are low, due to analyser rounding.
• Confidence interval – we are 95% confident that the actual concentration is within this range of our current Nowcast prediction. This is given in terms of µg m-3 concentration units. It is calculated from a match between the Nowcast and measurement at each monitoring site and has been estimated for the mid-range (median) of pollution concentrations in the Nowcast. Measurements aren’t perfect so we’ve included our confidence in the measurements too so that the confidence limit in our Nowcast isn’t penalised by these.
• The index interval is the average difference between air quality index values(1 to 10) for this pollutant. You can use this number to judge how well our 95% confidence interval compares to the difference between the coloured index values on the Nowcast.
• The banding interval is the average difference between air quality bands (low, moderate, high and very high) for this pollutant. These bands determine the health advice given. You can use this number to judge how well our 95% confidence interval compares to the difference between the different categories of health advice.
• Results are not shown of PM2.5 particulates at the moment. This is due to considerable uncertainty in this calculation. We currently have too few PM2.5 FDMS sites, which are the only ones that can be used with the Defra air quality index. We hope to rectify this by getting more PM2.5 sites on the London network, and by getting a working volatile correction model for PM2.5 so that TEOM instruments can also be used.
• Results are also not shown for Ozone (O3). This is due to considerable uncertainty in this calculation. We think that the roadside sites are causing problems with the nowcast ozone caculation and have removed them. Currently have too few background ozone sites to do a linear regression, so we are using a simpler offset model that we think will perform better.