Abstract: In order to investigate the feasibility for identifying the geographical origin of navel orange by near infrared reflectance spectroscopy (NIRS), three kinds of samples obtained from Jiangxi，Chongqing and Hunan province were tested in this paper. Three models of navel orange traceability were developed using the first derivative (9 points smoothing) of spectra at 1140-1170 nm combined with Soft Independent Modeling of Class Analogy (SIMCA). Under 5% significance level, the identification rates of three models for the calibration set of samples were all 100%, and the rejection rates were 85.7%, 83.3% and 100%, respectively. The identification rates for the validation set of samples were all 100%, and the rejection rates were 100%, 89.5% and 100%, respectively. Give values 0, 1, -1 as category variables for origins of Jiangxi, Chongqing and Hunan, partial least squares-discrimination analysis models (PLS-DA) were established to determine geographical origins of navel orange. The best models were developed using raw spectra at 950-1650 nm when the number of principal components were 13. The results showed that the correlation between the predicted category variable and the measured category variable was significant with high correlation coefficient R2 (0.973), low root mean square error of calibration RMSEC (0.110) and root mean square error of prediction RMSEP (0.159). The identification rates were both 100% by PLS-DA model based on the calibration set and validation set of samples. These indicated that NIRS coupled with SIMCA and PLS-DA methods can be used for quickly and accurately discriminating geographical origin of navel orange samples.