Building the capacity of efficiently determining the provenance of food products represents a crucial step towards the sustainability of the global food system. Whether it is for enforcing existing legislation or providing reliable information to consumers, technologies to verify geographical origin of food are being actively developed. Biological tracers (bio-tracers) such as DNA and stable isotopes have recently demonstrated their potential for determining provenance. Here we show that the data fusion of bio-tracers is a very powerful technique for geographical provenance discrimination. Based on 90 individuals of Sockeye salmon that originate from 3 different areas for which we measured 17 bio-tracers, we demonstrate that increasing the combined bio-tracers results in stronger the discriminatory power. The generality of our results are mathematically demonstrated under simplifying assumptions and numerically confirmed in our case study using three commonly used supervised learning techniques.

Muscle tissue trimmings were donated by Albion Farms & Fisheries Ltd. (now Intercity Packers Ltd.), Richmond, BC, Canada. All samples were derived from fillet trimmings to simulate a likely Quality Assurance/Quality Control scenario. Each muscle trimming was processed to obtain 2 muscle tissue samples for analyzing 17 bio-tracers of two classes: 3 stables isotopes (δ 15 N, δ 13C and δ 34S) and 14 fatty acids (FA). One muscle sample from each fish was delivered frozen to the Lipid Analytical Services at the University of Guelph for fatty acid analysis using a combination of Bligh & Dwyer and Morrison & Smith methods.46,47 Individual FA weights (μg/g) were converted to a % FA composition. The second muscle samples were dried at 70°C for 2 days and ground into a fine powder in preparation for stable isotope analysis. Tissue samples were sent to the University of Windsor GLIER Chemical Tracers Lab for isotopic analysis of δ 15N, δ13 C and δ 34S (Windsor, ON, Canada).

Columns are as follows.

- id: sample identifier

- isotopic ratios
  - cal_d13C: δ13C (Carbon 13)
  - cal_d15N: δ15N (Nitrogen 15)
  - cal_d34S: δ34S (Sulfur 34)

- Composition of fatty acid (%, note that only fatty acids with >1% presence were retained).
  - FA_C16_0: Palmitic acid (C16:0, PubChem CID: 985)
  - FA_C16_1: Palmitoleic acid (C16:1, PubChem CID: 445638)
  - FA_C18_0: Stearic acid (C18:0, PubChem CID: 5281)
  - FA_C18_1: Oleic acid (C18:1, PubChem CID: 445639)
  - FA_C18_2n6: Linoleic acid (C18:2, PubChem CID: 5280450)
  - FA_C18_3n3: α-linolenic acid (C18:3, PubChem CID: 5280934)
  - FA_C18_4n3: Stearidonic acid (C18:4, PubChem CID: 5312508)
  - FA_C20_1: Gondoic acid (C20:1, PubChem CID: 5282768)
  - FA_C20_4n3: Eicosatetraenoic acid (C20:4, PubChem CID: 21863049)
  - FA_C20_5n3: Eicosapentaenoic acid (C20:5, PubChem CID: 446284)
  - FA_C22_1: Erucic acid (C22:1, PubChem CID: 5281116)
  - FA_C22_5n3: Docosapentaenoic acid (C22:5, PubChem CID: 5497182)
  - FA_C22_6n3: Docosahexaenoic acid (C22:6, PubChem CID: 445580)
  - FA_C24_1: Nervonic acid (C24:1, PubChem CID: 5281120)