Spatially Structured Prediction of Soil Penetration Resistance Reveals Confounded Associations with Cotton Traits Under Long-Term Furrow Irrigation

Soil penetration resistance (PR) is widely used to assess compaction, although water content and field position can confound its relationships with crops. This study characterised depth-specific PR, identified its principal soil predictors, and evaluated its associations with cotton traits under long-term furrow irrigation. Penetration resistance, soil properties, plant height, SPAD, and boll number were measured at 40 georeferenced locations across four transects. Partial least squares (PLS) and multiple linear regression models were compared using spatially structured cross-validation. Mean PR increased from 1.079 MPa at 0–10 cm to 1.540 MPa at 20–30 cm. Along the longitudinal field gradient, soil water content increased downslope (ρ = 0.973), whereas bulk density (ρ = −0.936) and mean PR (ρ = −0.973) decreased. Under contiguous-block validation, PLS slightly outperformed linear regression for mean PR (RMSE = 0.0516 MPa, predictive R2 = 0.968, RPD = 5.68), with soil water content and bulk density as the dominant predictors among the four retained variables. Strong raw correlations between PR and cotton traits became nonsignificant after adjustment for longitudinal position and transect, with 77.1–99.6% attenuation. EC increased downslope, whereas SAR and ESP decreased. Thus, the apparent relationships between PR and cotton traits largely reflected a shared longitudinal field gradient rather than an independently identifiable effect of PR.