The Statistics of Phytotoxic Air Pollutants

The usual analysis of data from monitoring pollutant gas concentrations in the rural atmosphere assumes that the concentration frequencies for the whole data set are log‐normally distributed. The estimated parameters from these distributions are then used to define the treatment levels applied to plants in experiments which are set up to assess the likely damage to vegetation caused by various pollutant gases. Data for sulphur dioxide and ozone concentrations at a rural site are used to show that this approach is not appropriate for studies on air pollutant effects on plants. Over Europe, phytotoxic concentrations of pollutant gases occur infrequently and their frequency is often underestimated. Extreme value statistics provide much improved estimates of the frequency of peak, potentially damaging concentrations. Consideration of the processes which lead to the production of high concentrations of a gas also indicates that analysis of the whole data set may be less informative than analysis of subsets of the data. Because ozone reacts readily with natural surfaces in conditions of poor atmospheric mixing, as on a calm night, concentrations are strongly influenced by local (1–100 m) site characteristics. For studies of larger‐scale patterns (20–100 km) in concentrations, it is therefore desirable to eliminate these effects. This has proved possible through analysis of subsets of the data.