In this study (open access), the authors developed a classification model for shrimp (prawn) origin using fused data from Stable Isotope Ration Analysis (SIRA) and trace elemental profiling. They built the model using reference samples from Ecuador (n = 191), Honduras (n = 118), and Thailand (n = 66). Reference samples were not only shrimp meat, but also telson (part of the shell), pond water, and feed. Reference samples were collected from different pond types in different sub-regional locations, but all over one season (winter 2024/25).
The authors report that random forest models demonstrated high accuracy for country-level classification of reference shrimp (out-of-bag error = 0.47%) and retained strong predictive power at subnational catchment levels for Ecuador and Honduras (OOB = 3.08–5.32%). They subjected the reference shrimp to typical commercial processing (e.g. tumbling with polyphosphates or sulphites) and found that the treated shrimp retained chemical fingerprints comparable to their reference shrimp meat counterparts, achieving a 100% successful assignment to subnational areas. Spearman tests among shrimp meat, telson, feed, and water revealed strong isotopic and elemental correlations. The telson samples were correctly classified to their country of origin when tested against reference models built from shrimp meat data, demonstrating that telson shell chemistry reliably mirrors the geographic signature of the edible tissue.
They applied their model to a survey of retail samples and report that these exhibited low assignment accuracy (16%), suggesting either post-processing alteration or false/fraudulent labeling of origin.
Photo by Fernando Andrade on Unsplash
