Abstract: In order to construct the pattern recognition model of tobacco producing areas based on chemical measurements, 21 chemical measurements from 1262 samples of flue-cured tobacco were collected from 2003 to 2007, and were used to construct a OPLS-DA model. The results showed that the OPLS-DA model was highly significant with the following model performance statistics (R²X=0.429, R²Y=0.702, Q²=0.627). The OPLS-DA of the training set gave a model with two Y-predictive components, and gave the correct rate of 76.67%, 93.33% and 93.33% to distinguish samples from Sichuan, Yunnan and Fujian tobacco producing areas. The OPLS-DA model based on chemical measurements can recognize tobacco samples from different producing areas, and can be used in screening for biomarkers for different areas and providing interpretations for regional difference of tobacco samples.