Assessment of near-infrared reflectance spectroscopy for monitoring soil carbon, nitrogen and organic matter in South African forestry plantations

Abstract

Soil organic matter, carbon and nitrogen contribute to soil quality, which influences plant growth. In forestry planta- tions soil carbon, nitrogen and organic matter significantly influence site productivity. In South Africa, plantation forestry occurs across a diverse landscape requiring assessment methodologies capable of high throughput and cost-effective screening to accommodate the scale at which plantations are established and monitored. Therefore, near-infrared reflectance spectroscopy (NIRS) was evaluated as a potential technology for this application. NIR spectra (3 600–12 500 cm⁻¹) of 403 soil samples, selected by parent material representative of major plantation sites or for improving model calibration ranges, were modelled against loss-on-ignition organic matter content (LOI OM) and combustion analysis for total carbon (%) and nitrogen (%) using partial-least squares (PLS) regression. The resulting calibration models for LOI OM (R ² = 0.891, RMSEE = 1.47%), total carbon (R ² = 0.906, RMSEE = 0.528%) and nitrogen (R ² = 0.893, RMSEE = 0.034%) all showed good predictability. An independent model evaluation using 297 soil samples showed a marginal increase in prediction error for LOI OM (R ² = 0.853, RMSEP = 1.678%), total carbon (R ² = 0.864, RMSEP = 0.627) and nitrogen (R ² = 0.425, RMSEP = 0.083%). The prediction error of these samples was less than 1.70% compared with the analytical reference values, indicating that NIR has the potential to reliably predict forestry SOM properties. Further model development is required to incorporate the soil spectral variation across less typical or less frequently represented plantation sites to expand the suitability of models for routine laboratory analysis. This study demonstrates that NIRS is an attractive technology for the rapid, non-destructive and cost-effective analysis of forestry soils for site quality evaluations and precision forestry.