Proximal electrical conductivity (EC) and spectral sensing has been widely used as a cost-effective tool for soil mapping at field scale. The traditional method of calibrating proximal sensors for functional soil property prediction (e.g., soil organic matter, sand, silt, and clay contents) requires the local soil sample data, which results in a field-specific calibration. In this large-scale study consisting of 126 fields, we found that the traditional local calibration method had suffered weak correlation or uninterpretable models due to confounding factors (e.g., soil moisture) and small field variability. We proposed a new global calibration method that integrates satellite remote sensing (i.e., SMAP soil moisture, Landsat 8, and MODIS) and topographic information to explicitly account for the confounding effects in a large domain for a more robust calibration. Results show that both global and local calibration showed marked reduction in total root mean squared error (RMSE) over SSURGO. Global calibration without local soil samples had comparable accuracy as the local calibration in predicting organic matter (OM), sand, silt, and clay at the depths of 0-30, 30-60, and 60-90 cm. Adding five local sample measurements to the global models (spiking) reduced the overall errors for all four soil attributes by correcting the cross-field errors resulting in the most accurate predictions. Our findings suggest that global calibration is a robust, accurate, and cost-effective solution for operational functional soil property mapping.