The Method And Standard Establishment Of Total Nitrogen Determination In Green Tea By Near Infrared Spectroscopy

Total nitrogen content is an important quality index of green tea. The traditional measurementmethod （Kjeldahl determination） produces substances that are detrimental to the environment and theprocess is time-consuming and complex. A new method to determine total nitrogen in green tea by usingNear Infrared （NIR） Spectroscopy was researched in this study. The sample preparation condition andinstrument parameters were also determined. This detection method adopted a NIR model that can giverobust and precise results, and proposed several sub models for different varieties and tenderness, whichfurther improved the predictive performance. The total nitrogen detection in green tea by Near InfraredSpectroscopy can achieve fast and efficient nitrogen detection and quality evaluation, and imposesminimal effects to environment. This technique has academic and realistic significance in developingthe rapid measurement methods and pushing the frontier of quality testing in tea. The summary of theresults are as follows:Sample preparation and loading conditions were determined. Small particle size and evennesscould be helpful for improving the spectrogram stability. The RMSEP and R of near-infrared nitrogencontent analysis model were0.115and0.9935respectively when the particle of green tea was0.5mm.The prediction performance met the detection requirements. The spectroscopy would have more solidperformance when samples were concealed from light and demonstrated homogeneous granularity. Theresults would be better off if the samples were prepared in20mL and fully compacted beforeimplementing spectroscopy.The spectral acquisition method was determined. The results are optimal when instrumentresolution was4cm^-1and scanned64times. Model results were insignificantly different when theresolution was set to be8cm^-1and scanning were implemented for32,48,64,64times; the result wasthe best when resolution was4cm^-1and scan64times. Analyzing this model, the RMSEP of externalvalidation set was0.0546which was significantly lower than others and the value of R was0.9982. Itsprediction accuracy was high. STDEV and RSD were0.020and0.334respectively.A calibration model of total nitrogen content in the green tea was established based on partial leastsquare.325representative green teas were collected for creating the calibration model. Its performancewas verified by100selective samples. The model demonstrated great predictive performance after thefirst derivative processing and smoothing, in spectrum range of4691³959cm^-1and5126⁴848cm^-1.The RMSEC and the value of R about calibration set were0.0884%and0.9948when the factors of themodel were9. The correlation coefficient of predicted and chemical measurement value about thevalidation set was0.9939and the RMSEP was0.0925%.Sub models of total nitrogen in green tea were established according to different varieties andtenderness. With the sub models, the prediction performance was improved and the mean square errorbetween measured and predicted value was reduced. Compared with the total model, the RMSEP of the models （the small-medium leaves variety, the big leaves variety, tenderness1, tenderness3, andtenderness5） were slightly lower. The prediction performance of the tenderness3model was the best.The RMSEP and R about predicted and chemical measurement value were0.00379%and0.9961whenthe factors of the model were10. At this time, the prediction error of the calibration model was between0.1912%and0.0001%, and the average prediction error was0.0793%.