How reliable are national digital soil maps for farm-scale baseline assessment? A multi-scale assessment using dairy farm data

ABSTRACT Digital soil mapping (DSM) provides a scalable approach for monitoring soil health indicators to support sustainable land management and precision agriculture. However, uncertainties associated with input data and fine-scale management factors including land-use type, fertiliser history, and grazing intensity can limit predictive reliability at operational scales. Evaluating the performance of DSM products under real farm conditions is therefore critical to determine their suitability for baseline assessment. This study assessed the accuracy of the Soil and Landscape Grid of Australia (SLGA), a national-scale DSM product, in predicting topsoil (0–30 cm) soil organic carbon (SOC), total nitrogen (TN), and total phosphorus (TP) using more than 800 independently collected topsoil samples from nine Australian dairy farms. The SLGA prediction performance was evaluated at sub-farm (individual Carbon Estimation Areas, CEAs, i.e., paddocks or groups of paddocks with similar land use and management within a farm) and farm scales (aggregated across all CEAs within a farm) using R², Lin’s concordance correlation coefficient (CCC), mean absolute error (MAE), and mean bias error (MBE). In addition, linear mixed-effects models tested SLGA's ability to detect differences due to land use and farming systems (pasture-based vs. confinement). The SLGA showed moderate accuracy for SOC and TN at the farm scale (SOC: R² = 0.79, CCC = 0.88; TN: R² = 0.87, CCC = 0.79), but substantially weaker agreement at the sub-farm scale, and very poor accuracy for TP across both scales (R² ≤ 0.01; CCC ≤ 0.02). Standardized errors were moderate for SOC and TN, and highest for TP. The SLGA successfully detected differences in SOC and TN between pasture-based and confinement systems, demonstrating value for broad farm-scale baseline assessments, but showed limited sensitivity to within-farm land-use variation. Overall, SLGA provides useful farm-scale baseline assessments of SOC and TN in dairy systems; however, improvements are required to enhance sub-farm precision, TP prediction, and the representation of local management effects.