An Error Budget for Mapping Field‐Scale Soil Salinity at Various Depths using Different Sources of Ancillary Data

To manage soil salinity, farmers need to map its variation, often quantified as the electrical conductivity of a saturated soil‐paste extract (EC e , dS m ‐1 ). However, EC e determination is time‐consuming and expensive. Previous studies have evaluated the use of digital elevation models (DEMs, i.e., elevation), airborne γ‐ray (γ‐ray) spectrometry (i.e., K, U, and Th) and electromagnetic (EM, i.e., EM38 and EM34) data to map EC e at the district scale. Herein we use similar ancillary data set and empirical best linear unbiased prediction (E‐BLUP) to make maps of EC e at different depth intervals (0–0.25, 0.25–0.50, and 0.50–0.75 m) at the field scale. The ancillary data was collected using a ground‐based mobile sensing system which included; a GPS which provided spatial coordinates (Easting, Northing), a RS700 (γ‐ray) mobile spectrometer, and a DUALEM‐1. An error budget procedure was conducted to quantify the model, input, and individual covariate errors of EC e . Results show that while none of the γ‐ray data were significant, scaled Easting, elevation, and 1‐m horizontal coplanar (1mHcon) were optimal for mapping EC e at 0 to 0.25 m, while 1mHcon was optimal at 0.25 to 0.50 m and 0.50 to 0.75 m. Among all the individual covariate errors for mapping 0‐ to 0.25‐m EC e , elevation (0.40 dS m ‐1 ) was smallest, followed by 1mHcon (2.23 dS m ‐1 ). To reduce error, additional soil samples are necessary to prevent the edge effect of the kriging process. Additionally, inversion of EM data could be used to improve EC e mapping considering the relatively large model error associated with EM data in the subsoil of inverted salinity profiles.