Regional-scale air pollution source identification using backward particle dynamics

Abstract

Air pollution is one of the most harmful consequences of industrialization because of its strong influence on both human health quality and climate in general. Often there appears a need to identify one single strong source of air pollution appearing as a result of an accident. In this paper, we propose a new algorithm for a single pollution source localization. The proposed algorithm uses the source–receptor matrix concept and assumption about the linearity of pollution transport that allows us to use the pollution spread simulations backward in time. In particular realization, we make use of the weather regional forecast model WRF for airflow simulation and of Lagrangian particle dispersion simulation software FLEXPART-WRF for pollution advection simulation both forward and backward in time. As a result, our algorithm produces the semi-empirical heatmap of possible pollution source locations with marked point of the biggest probability and estimative emission intensity at this point as a function of time. The algorithm is tested on several semi-synthetic and practical cases and compared with other solutions in this field. The mean distance between the predicted and the real sources is around 7 km for the Moscow dataset with 1096 experiments and 45 km region size and around 3 km for the Regional dataset with 803 experiments and 30 km region size. We also conduct an experiment on European Tracer Experiment-1 and get a strong performance on it: distance between the real and the predicted sources is around 6 km, which is comparable or superior to other approaches.