| Title: | Forecasting of VOC emissions from traffic and industry using classification and regression multivariate methods |
| Author(s): | Stojic A; Maletic D; Stanisic Stojic S; Mijic Z; Sostaric A; |
| Address: | "Institute of Physics Belgrade, University of Belgrade, Pregrevica 118, 11080 Belgrade, Serbia. Electronic address: andreja.stojic@ipb.ac.rs. Institute of Physics Belgrade, University of Belgrade, Pregrevica 118, 11080 Belgrade, Serbia. Electronic address: dimitrije.maletic@ipb.ac.rs. Singidunum University, Danijelova 32, 11010 Belgrade, Serbia. Electronic address: sstanisic@singidunum.ac.rs. Institute of Physics Belgrade, University of Belgrade, Pregrevica 118, 11080 Belgrade, Serbia. Electronic address: zoran.mijic@ipb.ac.rs. Institute of Public Health Belgrade, Bulevar Despota Stefana 54, 11000 Belgrade, Serbia. Electronic address: andrej.sostaric@zdravlje.org.rs" |
| DOI: | 10.1016/j.scitotenv.2015.03.098 |
| ISSN/ISBN: | 1879-1026 (Electronic) 0048-9697 (Linking) |
| Abstract: | "In this study, advanced multivariate methods were applied for VOC source apportionment and subsequent short-term forecast of industrial- and vehicle exhaust-related contributions in Belgrade urban area (Serbia). The VOC concentrations were measured using PTR-MS, together with inorganic gaseous pollutants (NOx, NO, NO2, SO2, and CO), PM10, and meteorological parameters. US EPA Positive Matrix Factorization and Unmix receptor models were applied to the obtained dataset both resolving six source profiles. For the purpose of forecasting industrial- and vehicle exhaust-related source contributions, different multivariate methods were employed in two separate cases, relying on meteorological data, and on meteorological data and concentrations of inorganic gaseous pollutants, respectively. The results indicate that Boosted Decision Trees and Multi-Layer Perceptrons were the best performing methods. According to the results, forecasting accuracy was high (lowest relative error of only 6%), in particular when the forecast was based on both meteorological parameters and concentrations of inorganic gaseous pollutants" |
| Keywords: | Air Pollutants/*analysis Air Pollution/*statistics & numerical data Environmental Monitoring/*methods Multivariate Analysis Regression Analysis Serbia Vehicle Emissions/analysis Volatile Organic Compounds/*analysis Forecasting Mva Ptr-ms Receptor modeling; |
| Notes: | "MedlineStojic, Andreja Maletic, Dimitrije Stanisic Stojic, Svetlana Mijic, Zoran Sostaric, Andrej eng Research Support, Non-U.S. Gov't Netherlands 2015/04/02 Sci Total Environ. 2015 Jul 15; 521-522:19-26. doi: 10.1016/j.scitotenv.2015.03.098. Epub 2015 Mar 28" |
