Abstract: In the paper we aimed to provide a rapid, simple and accurate model for estimating of pericarp saponin concent in Sapindus mukorossi, based on the near infrared reflectance spectroscopy (NIRs) and high performance liquid chromatography (HPLC). And 145 samples were collected from Tiantai County of Zhejiang Province. Based on the analysis data of HPLC and NIRs, two models of saponin concent determination were established, and the accuracy of the models were evaluated. The spectroscopic data of group A (the whole fruit) and group B (the powder pericarp) were obtained from the pretreatments of "SNV+1^st derivative" and "MSC+1^st derivative". The results showed that the spectral models based on this data and established by PLS, were the best and the most reliable. They also had prediction precision. The model by group B was better than the model by group A. The models with groups A and B could reach 0.654 and 0.993 of the correlation coefficient between the prediction and the HPLC measured values. The standard deviations of prediction were 0.982 and 0.294, respectively. Thus, the model with the whole pericarp could be used to measure low quantity and precious samples with the relative low accuracy. The model with powder pericarp was very proper and could directly replace HPLC method.

Key words: Sapindus mukorossi Gaertn., near infrared spectroscopy, pericarp, saponin concent, PLS