Routine soil tests support screening of high aluminium saturation in acidic cacao soils

High aluminium saturation is a frequent but unevenly monitored constraint in acidic tropical soils, where routine soil testing is often more accessible than complete exchangeable-acidity determinations. This study developed and internally validated a screening model for high or very high aluminium saturation in cacao soil samples from the central Peruvian Amazon. Of 1842 anonymized monitoring records, 1539 acidic samples (pH < 5.5) formed the analytical cohort; the outcome was laboratory-reported aluminium saturation ≥ 20%. Six logistic-regression models compared routine soil tests with gridded PISCOp and SoilGrids covariates. All models used the same cohort, and preprocessing and threshold selection were nested within fivefold grouped cross-validation based on surrogate environmental signatures. The routine model achieved pooled balanced accuracy of 0.876 (fold standard deviation 0.022; group-bootstrap 95% confidence interval 0.856-0.896), macro-F1 of 0.851, sensitivity of 0.868, specificity of 0.885 and positive predictive value of 0.952. Areas under the receiver-operating-characteristic and precision-recall curves were 0.936 and 0.968, and the Brier score was 0.082. Relative to the routine model, the balanced-accuracy differences were 0.000 (95% confidence interval -0.010 to 0.010) with PISCOp and -0.007 (-0.020 to 0.005) with PISCOp plus SoilGrids, providing no evidence of a stable improvement within the available spatial support. The model is intended to prioritize confirmatory aluminium-saturation analysis in comparable acidic cacao soils. It does not measure plant toxicity, replace laboratory diagnosis or support continuous zoning of unsampled areas.