SINTEZA 2020 INTERNATIONAL SCIENTIFIC CONFERENCE ON INFORMATION TECHNOLOGY AND DATA RELATED RESEARCH 8 Mirjana Perišić 1,2 *, Gordana Jovanović 1,2 , Ana Vranić 1 , Svetlana Stanišić 2 1 Institute of Physics Belgrade, National Institute of the Republic of Serbia, University of Belgrade, Belgrade, Serbia 2 Environment and Sustainable Development, Singidunum University, Belgrade, Serbia Correspondence: Mirjana Perišić e-mail: mirjana.perisic@ipb.ac.rs BENZENE SOURCE APPORTIONMENT USING BIVARIATE CORRELATION AND REGRESSION ANALYSES Abstract: The aim of this study was to identify and characterize the individual sources of benzene in an urban area of Belgrade based on bivariate polar plot con- centration, correlation, and regression analyses. The presented benzene behavior modeling relied on a weighted Pearson correlation coefficient, linear regression slope, and Gaussian kernel locally weighted by wind speed direction surface. The data, including the concentrations of volatile organic compounds (VOCs), inorganic gaseous pollutants and meteorological parameters, were obtained from a measurement campaign conducted at the Singidunum University (Belgrade, Serbia) during and after the heating season 2016. The results indicate the dominance of benzene, toluene, and NO local emission sources, as well as the significant impact of remote NO 2 sources located in the SW and SE. The strong interrelations between VOCs indicate the common origin of these compounds. High toluene to benzene ratio (>2) was almost independent of wind speed and direction, indicating that the entire area was severely exposed to fresh vehicular emissions. The absence of relationships between benzene and fossil fuel combustion gaseous pollut- ants including NO x and O 3 suggests that evaporations from small chemical industrial complex situated in the S direction from the study site might be the main benzene and toluene emission source in the area. As regards inorganic gaseous pollutants, the relationship between benzene and NO in the N and NE might be related to the intensive anthropogenic activities in the central urban area of Belgrade and petrochemical industry in Pančevo. As presented herein, a combination of bivariate polar plot concentration, correlation, and regression analyses offers unique insight into the individual sources of air pollutants and their concentration dynamics. Keywords: benzene, volatile organic compounds, bivariate polar plot analysis. 1. INTRODUCTION In the recent years, volatile organic compounds (VOCs) became an important environmental issue because of their harmful impacts on human health and the environment. Tey originate from numerous natural and anthropogenic sources and are involved in a wide range of chemi- cal reactions in the atmosphere including the formation of secondary Sinteza 2020 submit your manuscript | sinteza.singidunum.ac.rs DOI: 10.15308/Sinteza-2020-8-13 ARTIFICIAL INTELLIGENCE ATLAS SESSION