Scale-dependent controls on soil organic carbon in perennial vineyard systems

Understanding how controls on soil organic carbon (SOC) vary with scale is a central challenge for interpreting SOC patterns in perennial agricultural systems. In vineyard landscapes, management effects may be masked by broader biophysical and edaphic gradients. We evaluated scale-dependent SOC controls by integrating predictive modeling and field measurements across seven commercial vineyards spanning a broad spatial gradient and contrasting soil orders in central Chile.

Machine-learning–based predictive models were developed independently for each vineyard and for three soil depths. Predicted SOC stocks for 0–30 cm ranged from 2.0 to 5.8 kg m⁻². Model performance varied strongly among sites, with validation RMSE between 0.3 and 1.9 kg m⁻² and R² from 0.00 to 0.89. Variable importance analyses consistently identified biophysical predictors related to topography, soil moisture, vegetation, and climate as dominant drivers, whereas management variables showed low predictive relevance. Field measurements revealed clear depth effects, with higher SOC in 0–15 cm than in 15–30 cm layers. Inter-row positions tended to show higher surface SOC than vine rows (+14%), but these differences were inconsistent among vineyards and did not emerge as significant effects at the global scale. Row–inter-row differences in 0–30 cm SOC ranged from −0.47 to +0.84 kg m⁻².

Overall, SOC dynamics reflected hierarchical controls dominated by biophysical and edaphic factors at the site scale, illustrating how management effects emerge only at finer spatial scales in perennial systems.