Strategies for soil quality assessment using VNIR gyperspectral spectroscopy in a western Kenya Chronosequence

Visible and near-infrared reflectance spectroscopy (VNIRS) is a rapid and nondestructive method that can predict multiple soil properties simultaneously, but its application in multidimensional soil quality (SQ) assessment in the tropics still needs to be further assessed. In this study, VNIRS (350–2500 nm) was employed to analyze 227 air-dried soil samples of Ultisols from a soil chronosequence in western Kenya and assess 16 SQ indicators. Partial least squares regression (PLSR) was validated using the full-site cross-validation method by grouping samples from each farm or forest site. Most suitable models successfully predicted SQ indicators (R² ≥ 0.80; ratio of performance to deviation [RPD] ≥ 2.00) including soil organic matter (OM_LOI), active C, Ca, cation exchange capacity (CEC), and clay. Moderately-well predicted indicators (0.50 ≤ R² < 0.80; 1.40 ≤ RPD < 2.00) were water stable aggregation (WSA), Cu, silt, Mg, pH, sand, water content at permanent wilting point (Θ_pwp), and field capacity (Θ_fc). Poorly predicted indicators (R² < 0.50; RPD < 1.40) were EC, S, P, available water capacity (AWC), K, Zn, and penetration resistance. Combining VNIRS with selected field- and laboratory-measured SQ indicator values increased predictability. Furthermore, VNIRS showed moderate to substantial agreement in predicting interpretive SQ scores and a composite soil quality index (CSQI) especially when combined with directly measured SQ indicator values. In conclusion, VNIRS has good potential for low cost, rapid assessment of physical and biological SQ indicators but conventional soil chemical tests may need to be retained to provide comprehensive SQ assessments.