Soil sampling and laboratory analyses are onerous and expensive, especially for large fields. Therefore, alternative rapid and inexpensive methods to predict plant-available soil nutrients are needed – such as proximal sensors. An on-farm evaluation of soil chemical properties was used to evaluate the viability of using electrical conductivity (EC) or apparent electrical conductivity (ECa) to model California soils' chemical characteristics. This survey used a Veris electrical conductivity meter and an EM38 apparent electrical conductivity meter to build data sets for two test sites. Soils were manually sampled (15 cm depth) to determine whether the EC or EC_a values could predict the lab analyzed soil sample results. Two sites were investigated. The soils at both locations mapped primarily as the Yolo soil series, a very deep, well-drained soil formed in alluvium from mixed rocks.

EC data from the Veris ranged from 32 to 63 mS/m, and the EM38 returned EC_a values from 23 to 46 mS/m. Evaluation of the raw Veris and EM38 data showed significant outliers, limiting our ability to model the data between the two reliably. The Veris and EM38 readings showed significant correlation with soil-test nitrate-nitrogen in the upper 15 cm of the soil at Site 1 (r = 0.68, p = 0.001) (r = 0.49, p = 0.001), respectively. Unfortunately, Site 2 did not have nitrate-nitrogen data due to sampling error.

The EC_a values of the EM38 unit correlated positively(r = 0.48, p = 0.001 ) with soil-test potassium for Site 1. The relationship between the EM38 EC_a values and potassium at Site 2, however, had a negative relationship (r = -0.39, p = 0.01). These contrasting findings are most likely because of clay's water-holding ability, as it strongly affects EC readings. Also, clay-size minerals are the main soil components that retain exchangeable potassium. Therefore, we felt clay concentration was a pertinent parameter to explore further. These significant correlations deserved exploration with a primary focus on reliable EM38 data. Using seven different models, we found the ability to capture adequate information behind the relationship between EM38 ECa values and potassium. The further addition of soil test clay data improved evaluations. We focused our efforts on determining the viability of utilizing these improved readings to eliminate the need for future soil testing for potassium. Evaluation using the mean predicted error of our models would outline any significant relationships from the predicted values. In combination with soil sampling procedures, cost estimates will effectively determine our ability to replace soil sampled potassium with estimates derived from soil EC and EC_a values.