Retrieval of air quality annual statistics from a limited number of profiles

Abstrakt

Annual air quality statistics in an urban environment are usually computed from data collected from a sparsely distributed network of monitoring stations. Complex numeric models may provide air quality simulations at the street level, these simulations, however, are not computationally feasible for large time periods like a year. We propose a method for identification of a limited number of 'typical' days, which, if simulated in microscale, guarantee a reasonable coverage of different scenarios during the year. Annual statistics then can then be estimated on street level from simulated fields. The identification method is based on k-medoids clustering. We also develop a means for validating the approach so as to add confidence in estimates derived in this manner.