A temperature-based approach to predicting lost data from highly seasonal pollutant data sets.

A new technique to predict concentrations of benzo[a]pyrene (BaP) in ambient air during periods of lost data has been developed and tested. This new technique is based on the relationship between ambient temperature and BaP concentration observed at individual monitoring stations over many years. The technique has been tested on monthly data of BaP concentrations in PM10 at individual monitoring stations on the UK PAH Monitoring Network. The annual average concentration values produced with and without the use of predicted data have been compared to the actual annual averages in the absence of any data loss. The use of predicted data is a significant improvement when compared with the averages produced in the absence of any data prediction and outperforms previous prediction strategies associated with intra-year trends. Furthermore the technique is suitable for the prediction of long periods of missing data, which other prediction techniques have not been able to deal with satisfactorily.