Evaluation of modified Hilhorst models for pore electrical conductivity estimation using a low-cost dielectric sensor

Real-time measurement of soil water content (θ) and pore electrical conductivity (ECp) is essential to improve water irrigation efficiency and agricultural productivity. Low-cost frequency domain reflectometry (FDR) sensors are now representing a powerful tool for irrigation management purposes. However, compared to the time domain reflectometry (TDR), FDR sensors’ accuracy to predict θ and ECp is negatively affected by saline conditions. Thus, it is necessary to determine the soil salinity range where FDR probes are not recommended in precise irrigated agriculture and to select the appropriate models for ECp estimation especially under saline conditions. Low-cost sensors, however, often use the default Hilhorst model for ECp determination, and in salty soils, this use is not correct. Thus, we present a new and improved Hilhorst model of ECp estimation. We also assess the performance of the low-cost Water, Electrical conductivity, and Temperature (WET) sensor and to test the new ECp model under saline conditions. Consequently, the ECp was predicted using, first, a polynomial model in which ECa effect on the soil parameter K0 is considered and second, a linear model in which the ECa effect on soil apparent dielectric permittivity Ka is considered. The performance of the proposed models is evaluated by measurements of the WET sensor in sandy porous media collected in the Tunisian Jemna oasis using seven different levels of NaCl solutions (0.02 to 8.2 dSm−1) and compared to TDR measurements. Results show that using the default Hilhorst model, the root mean square error (RMSE) of ECp predictions was higher than 0.5 dSm−1 using WET sensor. However, if considering the bulk electrical conductivity (ECa) effect on the soil parameter K0 instead of using the standard values in the Hilhorst model, the performance of the WET sensor to predict ECp increased with a mean RMSE equal to 0.1 dSm−1.