Thomas et al. 2023 — Nation-wide distribution of cadmium in soil and cacao beans, Peru
Peru is the world’s eighth largest cacao producer and second largest exporter of organic cacao, but cadmium content in some regions substantially exceeds EU and California regulatory thresholds. Drawing on more than 2,000 representative soil and cacao bean samples from 563 farms across 13 Peruvian departments, this study developed multiple national and regional random forest predictive models for soil and bean Cd at approximately 250 m resolution. The most important predictors across models were soil cadmium, geology, rainfall seasonality, soil pH, and geographic position. Elevated Cd concentrations in both soil and beans are largely restricted to the northern departments of Tumbes, Piura, Amazonas, and Loreto, with very localised high-Cd pockets in the central departments of Huanuco and San Martin. At national scale, fewer than 20% of cacao farming households are estimated to be affected by EU or US regulations, but in the most affected department of Piura, up to 89% of farms may be impacted.
Key numbers
Total samples: 1,717 soil and 1,534 bean samples from individual trees on 563 farms (tree-level), plus 334 composite soil and 484 composite bean samples (farm-level). Soil Cd sampling period: April 2018 to September 2019. Bean Cd LOQ: 0.047 mg/kg (individual trees), 0.03 mg/kg (farm composites); LOD: 0.01 mg/kg. ICP-MS (Nexion 2000C, Perkin Elmer), aqua regia digestion (1:3 HCl:HNO3). Certified reference material: OREAS 906 (0.42 ± 0.04 mg/kg Cd). Cross-validation R² for national model: soil Cd 0.77 ± 0.15 (calibration), 0.24 ± 0.11 (spatial block); bean Cd 0.83 ± 0.03 (calibration), 0.48 ± 0.26 (spatial block). Irrigation water Cd (DGT passive membranes) in Las Lomas, Piura: 2.80 ± 0.82 mg/m³ (soil Cd sites), 7.12 ± 2.70 mg/m³ (bean Cd sites); La Quemazón: 0.315 ± 0.267 µg/L to 0.085 ± 0.108 µg/L. Seasonal pattern in irrigation Cd: highest values in Las Lomas during dry season (months not specified in pages 1–5). EU regulatory thresholds for cacao: 0.6 mg/kg (cacao powder), 0.8 mg/kg (cacao beans) per EC 488/2014. Key predictors: geology (continental Quaternary Holocene layer — Qh_c — was the highest-risk geological unit), rainfall seasonality, soil pH, alluvial deposit proximity, and mining upstream in watershed.
Methods (brief)
Farm selection stratified across Peru’s cacao-producing regions; 100–150 farms targeted per region, fewer in smaller regions. GPS coordinates recorded for every tree. Soil samples: 8 subsamples per farm, zig-zag methodology, 70 cm from trunk, 0–20 cm depth, composited. Bean samples: up to 4 cacao pods per tree, 5–10 pods composited to ~1 kg fresh weight, frozen, ground in food mill, 0.5 g digested in 6 mL HNO3 and 2 mL H2O2, ICP-MS. Farm-level bean analysis: SENASA (Peruvian government ISO-certified laboratory). Random forest models: cforest function (R party package), 15-iteration block cross-validation (blockCV). PMF-style predictor variables included 19 bioclimate variables (worldclim), terrain, soil properties (ISRIC World Soil Information), deforestation (Global Forest Watch 2022), and mining inventory (INGEMMET 2019).
Limitations
Tree-level bean Cd measurements reflect individual genotype variation in addition to soil Cd; bean-level results should not be interpreted as farm means without the composite data. Mining variable is a binary (presence/absence) and does not reflect mine type or activity level. Irrigation water Cd measurements are site-specific (two irrigation channels in Piura) and not representative of all irrigated cacao areas. Regional models had lower block cross-validation R² (0.36–0.87) indicating spatial autocorrelation affects model generalisation.
Implications
- Certification: This is the most comprehensive national-scale Cd mapping study for Peruvian cacao to date (n=2,194 samples). The finding that <20% of Peruvian farms are nationally affected but up to 89% in Piura are above EU thresholds is directly actionable for HMT&C sourcing guidance: fine-flavor Peruvian cacao from Piura or Tumbes regions warrants Cd-specific testing, whereas Amazonian Peruvian cacao (Madre de Dios, Ucayali, southern regions) presents lower risk. The geological Qh_c layer finding explains why geographic provenance is more predictive than farm management.
- Courses: Definitive case study for Peru cacao Cd geography; covers the complete chain from geology → soil → irrigation water → bean Cd with quantitative predictors and nation-wide spatial coverage.
- App: This paper provides the primary quantitative basis for a geographic_breakdown entry for Peruvian cacao Cd in cocoa. Piura/Tumbes region: high risk (>EU threshold for up to 89% of farms). Central Amazon (Madre de Dios, Ucayali, southern): low risk. Seasonality: irrigation water Cd showed seasonal pattern; highest during dry season in Las Lomas, Piura.
- Microbiome: Not applicable.