Multisource fingerprinting for region identification of walnuts in Xinjiang combined with chemometrics

Multiple techniques were employed to extract characteristic fingerprint information of walnuts from the different regions of Xinjiang. Analysis was conducted using near‐ and mid‐infrared spectroscopy, fatty acid and element data to establish geographic classification models of walnuts combined with chemometrics. The least squares discriminant analysis (PLSDA) model had a classification accuracy result of 98.8–100% for walnut samples from different geographical origins by NIR spectra. For the same variety of samples, PLSDA model also had high classification accuracy of 91–100% by MIR spectra. However, variance analysis showed that single element or fatty acid could not be differentiated between the three geographical origins of walnuts. In addition, the back propagation neural network classification model was used to analyze the elements found in the walnut specimens provided results with a mean accuracy of 89.7% for validation. This work demonstrated an acceptable feasibility of using multisource fingerprinting for differentiating regional and characteristic agriculture product to prevent frauds. Practical applications Currently, there is no effective, simple and rapid technology to prevent frauds of walnuts in Xinjiang occurring in the market, especially walnuts with adulteration or which has been counterfeited. Hence, in this research, we provided a feasible method to differentiate the walnut from different regions based on multiple fingerprinting and geographic traceability techniques, which could be applied to prevent frauds of walnuts occurring in the market in the future.