Abstract: This study aimed to improve the stability and accuracy of mineral element fingerprinting for the geographical traceability of soybean. In this experiment, the contents of mineral elements in 168 soybean samples and the corresponding soil samples from Nenjiang county and Bei’an county in Heilongjiang province were determined by inductively coupled plasma mass spectrometry (ICP-MS). The contents of protein, fat, total soluble sugar and ash were determined as well. The results showed that the discriminant model established based on 10 characteristic indicators (seven mineral elements plus three organic compounds) selected by the stepwise selection method has an overall accuracy rate of 96.4% for the training set. Besides, the discrimination accuracy for soybean samples from Nenjiang county and Bei’an county were 98.1% and 95%, respectively. The overall discrimination accuracy for the validation set was 98.2% in the back substitution test, and the discrimination accuracy for soybean samples from Nenjiang county and Bei’an county were 100% and 96.7%, respectively. The overall discrimination accuracy for the validation set were higher than the test set (98.2% > 96.4%) indicating that all ten characteristic indicators carry sufficient information about geographical traceability.

Key words: soybean, mineral elements, nutrients, characteristic indexes, geographical traceability