Assessing the performance of standard methods to predict the standard uncertainty of air quality data having incomplete time coverage†
Abstract
As a result of the complex nature of operating multi-station national air quality networks it is rare that complete data sets are produced from these networks. The reliance of most air quality legislation on the assessment of measured annual average concentrations against target or limit concentrations necessitates the use of methods to calculate an annual average value and the uncertainty in this value in the absence of a complete data set for the year in question. Standard procedures exist for performing these calculations, but it is not clear how effective these are when data having low time resolution are collected and missing data accounts for large periods of the year. This paper investigates the influence of these deficiencies using data from UK air quality networks in the form of monthly average concentrations for polycyclic aromatic hydrocarbons and for metals in the PM10 phase of ambient air. Whilst the standard methods currently employed produce good results on average, for individual cases the uncertainty in the annual average calculated when data is missing may be appreciably different from that obtained when full knowledge of the distribution of the data is known. These effects become more apparent as the quantity of missing data increases.