Study On The Discrimination Of Ephedra Plants With Diffuse Reflectance FT-NIRS And Multivariate Analysis

Ephedra plants (Ephedraceae) are widely distributed in the world. Among them, Ephedra sinica (E. sinica), Ephedra intermedia (E. intermedia) and Ephedra equisetina (E. equisetina) are dominant and economically important species in China and their main habitats are in Inner Mongolia, Xinjiang, Shanxi, Gansu and so on. The secondary metabolites contained in Ephedra plants mainly include alkaloids, essential oils, tannins, flavonoids and organic acids. They are very important for the plants'survivals. On the other hand, plant secondary metabolites are unique sources for pharmaceuticals, food additives, flavors, and other industrial materials. Hence, the identification of Ephedra plants of different species, habitats and picking times could help the research of phytoecology of Ephedra and the utilization of Ephedra plants.In this thesis, the feasibility of the discrimination of Ephedra plants of different species, habitats and picking times is evaluated with diffuse reflectance FT-NIRS. The optimization of the spectral measurement conditions, spectra processing approaches, data pre-processing methods, and data analysis techniques is also discussed.The main research contents and results are as follows.1. Discrimination of Ephedra plants of different species with diffuse reflectance FT-NIRS: The samples of E. sinica, E. intermedia and E. equisetina were discriminated with reference to morphological characters. The herbaceous stems (moisture contents < 8.0%) were pulverized and sieved. The Fourier transform near infrared diffuse reflectance spectra (NIRDRS) were acquired from pulverized samples put in glass vials in the near infrared (NIR) region between 10 000 and 4000cm?1, averaging 64 scans per spectrum at a resolution of 4 cm?1. The NIRDRS were processed first with smoothing techniques. The five-point Norris smoothing was used in the discriminant analysis (DA). The seven-point Savitzky–Golay smoothing was used in the self-organizing map (SOM) and back-propagation artificial neural network (BP-ANN). The methods for subsequent processing were first derivative and multiplicative signal correction (MSC). The dimensions of the spectral data were reduced to 9 principle components (PCs) for the DA and to 10 PCs for the artificial neural network (SOM and BP-ANN). The cumulative contribution rates of 9 and 10 PCs were 98.7% and 99.9% in sequence. Ephedra plants of different species could be discriminated respectively with the DA, SOM and BP-ANN. The performance index of the DA model was 84.2%. The prediction accuracies of both the SOM and the BP-ANN models were 95.0%. The visualization functions were achieved with the DA and the SOM.2. Discrimination of E. sinica from different habitats with diffuse reflectance FT-NIRS: The NIRDRS of the samples of E. sinica from Shanxi and Inner Mongolia were measured using the method used in Section 1. The NIRDRS were processed by the five-point Savitzky–Golay smoothing technique and MSC. The dimensions of the spectral data were reduced to 10 PCs and the cumulative contribution rate was 99.8% for the DA, SOM and BP-ANN. E. sinica from different habitats could be identified respectively with the three data analysis methods. The performance index of the DA model was 91.9%. The prediction accuracies of both the SOM and the BP-ANN models were 100.0%. The visualization functions were achieved with the DA and the SOM.3. Discrimination of E. sinica from Shanxi picked at different times of day with diffuse reflectance FT-NIRS: The NIRDRS of the samples of E. sinica from Shanxi picked between 10:00 and 11:30 am and between 4:30 and 5:00 pm were measured using the method used in Section 1. The techniques of spectra processing and data pre-processing for this task were the same as those used in Section 2. Nevertheless, the cumulative contribution rate of 10 PCs was 99.7% for the three data analysis methods. E. sinica from Shanxi picked at different times of day could be distinguished respectively with the DA, SOM and BP-ANN. The performance index of the DA model was 88.1%. The prediction accuracies of both the SOM and the BP-ANN models were 93.3%. The visualization functions were achieved with the DA and the SOM.The main conclusions are as follows.1. The results obtained in this task suggest that there is significant difference between E. sinica picked in the morning and afternoon. This fact implies that the secondary metabolites of E. sinica are modulated by time of day.2. The experiment has proved that diffuse reflectance FT-NIRS with multivariate analysis techniques could distinguish not only the Ephedra plants of different species and different habitats but also the plants picked at different times of day without special sample treatment and the use of chemical reagents. The approach established is objective, easy-to-use, rapid and pollution-free. It is a useful tool for the research of phytoecology of Ephedra and the utilization of Ephedra plants.