Pollutant roses for daily averaged ambient air pollutant concentrations

Pollutant roses are indispensable tools to identify unknown (fugitive) sources of heavy metals at industrial sites whose current impact exceeds the target values imposed for the year 2012 by the European Air Quality Daughter Directive 2004/207/EC. As most of the measured concentrations of heavy metals in ambient air are daily averaged values, a method to obtain high quality pollutant roses from such data is of practical interest for cost-effective air quality management. A computational scheme is presented to obtain, from daily averaged concentrations, 10° angular resolution pollutant roses, called PRP roses, that are in many aspects comparable to pollutant roses made with half-hourly concentrations. The computational scheme is a ridge regression, based on three building blocks: 1. ordinary least squares regression; 2. outlier handling by weighting based on expected values of the higher percentiles in a lognormal distribution; 3. weighted averages whereby observed values, raised to a power m, and daily wind rose frequencies are used as weights. Distance measures are used to find the optimal value for m. The performance of the computational scheme is illustrated by comparing the pollutant roses, constructed with measured half-hourly SO2 data for 10 monitoring sites in the Antwerp harbour, with the PRP roses made with the corresponding daily averaged SO2 concentrations. A miniature dataset, made up of 7 daily concentrations and of half-hourly wind directions assigned to 4 wind sectors, is used to illustrate the formulas and their results.