The paper presents the results of investigation on classification of the ambient air samples collected in a vicinity of the LOTOS Group S.A. petroleum plant with respect to odour. The investigations were carried out with an electronic nose prototype and a commercial electronic nose of Fast GC-type – HERACLES II. The prototype was equipped with a set of six semiconductor sensors by FIGARO Co. and a photoionization sensor of PID-type. Classification of the air samples depending on the sampling point with respect to the petroleum plant was performed with linear discriminant function supported with cross-validation method. Analyses with the HERACLES II yielded over 94.2% of correct classifications of the ambient air samples. In case of the electronic nose prototype correct classification was at the level of 67.5%.