Modeling airborne benzo(a)pyrene concentrations in the Czech Republic Amir Zalel a , Yuval b , Vlasta Svecova c , Radim J. Sram c , Alena Bartonova d , David M. Broday b, * a Craven Works Ltd., Israel b Civil and Environmental Engineering, Technion, Haifa, Israel c Institute of Experimental Medicine, Prague, Czech Republic d Norwegian Institute for Air Research (NILU), Kjeller, Norway highlights  Models for estimation of daily and monthly ambient B[a]P levels were developed.  The concentrations seasonality was addressed by combined CARTeMLR models.  The models reproduced very accurately monthly mean ambient B[a]P concentrations.  Spatial across-site extrapolations revealed reliable models performance.  Temporal extrapolations revealed comparable errors to the spatial extrapolations. article info Article history: Received 2 September 2014 Received in revised form 12 November 2014 Accepted 14 November 2014 Available online 15 November 2014 Keywords: PAHs B[a]P Multivariate linear regression Classification trees Air pollution monitoring abstract Polycyclic aromatic hydrocarbons (PAHs) are complex hazardous organic compounds that are introduced into the atmosphere as by-products of partial combustion processes. For common atmospheric condi- tions, the large molecular weight PAHs, such as benzo(a)pyrene (B[a]P), are found in the particulate phase and are believed to account for a considerable amount of the fine particulate matter toxic po- tential. Nonetheless, unlike meteorological variables and criteria pollutants, PAHs are very rarely monitored on a routine basis in most parts of the world. We present methodology for development and evaluation of a model for estimation of daily and monthly ambient B[a]P concentrations. The model utilizes a very large ambient B[a]P database from three sites in the Czech Republic. The difficulties faced when dealing with ambient PAH data are discussed. Model performance was evaluated by a complete internal-, external-, and temporal cross validations. The models reproduced very accurately monthly mean ambient B[a]P concentrations and provided acceptable daily mean B[a]P concentrations. Spatial extrapolations resulted in small deterioration of the models' performance. The temporal backward extrapolation revealed comparable errors to the spatial extrapolations in spite of the dramatic emissions reduction in the early years of the study period. © 2014 Elsevier Ltd. All rights reserved. 1. Introduction Monitoring of ambient pollutants is required for air resources management and for exploring relationships between air pollution and health outcomes. For example, environmental epidemiology studies normally use air quality monitoring data to derive exposure metrics, and may suffer from considerable uncertainties in sparsely monitored regions. Criteria air pollutants (NO x , NO, NO 2 , CO, SO 2 , O 3 , PM 10 and PM 2.5 ) are monitored on a wide spatiotemporal scale global-wise. Yet, many other pollutants are not monitored regularly at most places. This situation is not expected to change dramatically in the coming years due to technological and mainly financial limitations. In such cases, it is oftentimes required to rely on indi- rect means for estimating ambient concentrations of harmful pollutants. * Corresponding author. E-mail address: dbroday@tx.technion.ac.il (D.M. Broday). Contents lists available at ScienceDirect Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv http://dx.doi.org/10.1016/j.atmosenv.2014.11.031 1352-2310/© 2014 Elsevier Ltd. All rights reserved. Atmospheric Environment 101 (2015) 166e176