Comparative analysis of Vis-NIR and MIR spectroscopy for predicting soil properties and identifying minerals at smallholder cocoa farms across Papua New Guinea

Cocoa ranks as Papua New Guinea’s (PNG) third most important agricultural export, providing livelihoods for ~2 million smallholder farmers. Recently, the condition and capacity of cocoa soils have deteriorated from inadequate nutrient management under intensified cocoa farming—with little soil monitoring and nutrient replenishment—due to resource-constraints. To rapidly and cost-effectively evaluate the condition of cocoa soils in PNG, this study presents the first comparison of visible and near-infrared (Vis-NIR) and mid-infrared (MIR) spectroscopy. We developed a spectral library by measuring 17 soil properties and spectral features from 507 soil samples collected at four smallholder cocoa farms across major cocoa production regions of PNG. Results showed that both Vis-NIR and MIR spectral features predicted over half of the soil properties with relatively high accuracy (Lin’s concordance correlation coefficient ≥ 0.80), including: sand, clay, SOC, available N, available K (Vis-NIR only), exchangeable Ca, exchangeable Mg, Fe, Cu and Mn. MIR outperformed Vis-NIR in predicting soil texture and most chemical properties, while Vis-NIR provided slightly better predictions for certain nutrients—notably available K. The modelling results showed that between Cubist and partial-least-square regression (PLSR) models, Cubist models outperformed PLSR models for both Vis-NIR and MIR. Finally, MIR outperformed Vis-NIR in identifying soil minerals—such as kaolinite, gibbsite, halloysite, and calcite—using wavenumbers, making MIR the preferred spectroscopy method only when accurate mineral identification is required. These findings demonstrate Vis-NIR spectroscopy as the preferred and more practical method to effectively assess key soil properties for cocoa production—serving as a starting point for soil monitoring and nutrient management across cocoa farms in PNG—with broad applicability for soil monitoring and nutrient management in other resource constrained environments. • Vis-NIR and MIR spectroscopy techniques are compared to assess soil condition at four smallholder cocoa farms in Papua New Guinea (PNG). • Both Vis-NIR and MIR spectroscopy predicted over half of the 17 soil properties with high accuracy (LCCC ≥ 0.80). • MIR outperformed Vis-NIR spectroscopy in predicting nearly all soil properties. • Vis-NIR outperformed MIR spectroscopy in predicting some soil properties—notably available K. • MIR outperformed Vis-NIR spectroscopy in identifying soil minerals through wavenumbers. • Vis-NIR spectroscopy is the preferred method to effectively assess key soil properties for cocoa production—serving as a starting point for soil monitoring and nutrient management.