Degree Type


Date of Award


Degree Name

Doctor of Philosophy


Agricultural and Biosystems Engineering


Environmental Science

First Advisor

Amy L. Kaleita

Second Advisor

Stuart J. Birrell


A fast and reliable method for in situ monitoring of soil nitrate-nitrogen (NO3-N) concentration is vital for evaluation of N management practices focused on reduction of NO3-N losses to ground and surface waters from agricultural systems. Using dielectric measurements at multiple frequencies can help to estimate several physical and chemical soil properties simultaneously. Hence the goal of this study is to examine the feasibility to estimate changes in pore water NO3-N concentration together with volumetric water content (VWC) from the dielectric measurements obtained at multiple frequencies below several MHz where conductive behavior of soil dominates.

An initial experiment with two off-the-shelf capacitance probes showed that at a relatively high frequency response of the probe was primarily correlated with VWC, while measurements at a lower frequency made by the second probe also incorporated the effect induced by changes in pore water ionic concentration. These results confirmed that using measurements at multiple frequencies can provide information about several soil properties, including NO3-N concentration. Consequently two follow-up laboratory experiments used impedance spectroscopy to estimate changes in NO3-N concentration in pure solutions and soil water, respectively, using a multivariate chemometric analysis, particularly partial least squares (PLS) regression. The results showed that change in NO3-N concentration could be estimated with sufficient accuracy when its concentration was greater than concentration of other anions (chloride in our case). In addition, estimation of NO3-N in soil water improved significantly with increasing VWC. A good agreement was found between actual and estimated NO3-N concentration when the PLS model was built using permittivity data obtained at VWC ≥ 0.20 m3 m-3. R2 and the root mean square error (RMSE) of NO3-N estimation for the best model (VWC ≥ 0.20 m3 m-3 and concentration of chloride < 500 mg L-1) were 0.84 and 28 mg L-1, respectively. In general, the study demonstrated that PLS regression method coupled with the dielectric measurements obtained at multiple frequencies below several MHz can be used to indirectly estimate VWC and NO3-N concentration, but after the proper calibration equally covering the expected variations in VWC and NO3-N. For in situ application other environmental variables such as temperature should also be incorporated into the calibration process.


Copyright Owner

Giorgi Chighladze



Date Available


File Format


File Size

143 pages