Comprehensive gas chromatography (GCxGC) appears as the preferred technique for the characterization of volatile compounds in tobacco samples. Indeed, GCxGC offers an enhanced separation power which results from the combination of two columns of different selectivity - the first column is usually non-polar and the second polar. The main advantage of GCxGC is to reduce the number of co-elutions and hence to provide a way to increase the capacity to discriminate a maximum of compounds. However, the huge number of peaks, and consequently of information, contained into the 2D chromatograms induces the necessity to develop a strategy to process the data automatically.As shown in our previous work, any global comparison of GCxGC chromatograms and discrimination of tobacco classes requires several preprocessing steps such as background correction, intensity normalization, and especially time alignment, along the first or even both dimensions. After preprocessing, a correlation study enables us to locate the pixels that are discriminant for each class of samples, and to elaborate a characteristic map for each class. These maps can be seen as templates allowing the identification of areas of interest, i.e. areas corresponding to compounds over expressed or under expressed in one class as compared to all others.Henceforth, this first step is carried out by an automatic process embedded in a user-friendly graphical interface (G.U.I.), which facilitates data visualizations and analyses. The present study focuses on the map characteristics and how they may be affected by the various sources of variability of the input data: differences of tobacco origin, sample to sample variability, extraction to extraction variability, and finally injection to injection variability. The G.U.I. provides a basis for statistical analysis useful to determine the main sources of data variation, which demonstrate its fit for purpose.Ultimately, the variability analysis based on correlation maps, combined with mass spectrometry information, will provide a new tool to identify systematically the compounds responsible for the differences observed between different classes of samples and thus identify the corresponding chemical markers.