This paper presents a method to accurately determine the concentration of nitrates and chlorides in soil by way of multi-frequency impedance measurements of electrode surrounded by the soil under observation. We use the Bruggeman's symmetric mixing model that relates the permittivity of soil to permittivity and concentrations of constituents such as rock particles, air, water and ions. The presence of water as bound and free water has been accounted for. It is empirically shown that a parametric estimate of the permittivity in the vicinity of particles can provide a better estimate of saline water concentration. Since effective as well as individual constituent permittivity varies with frequency, permittivity measurements over multiple frequencies provide multiple equations with only the concentrations of the soil constituents as unknowns. We observe that the dielectric mixing model works well for low RF (radio-frequency) range. We validate our approach by making multi-frequency impedance measurements of a soil mixture at different concentrations of various components. To perform the experiment, a cylindrical coaxial capacitive cell was constructed and was filled with soil. As the concentration of water, nitrate and chloride salts is changed in the cell, the real and imaginary parts of permittivity change. This change is reflected in the input impedance measurement of the cylindrical cell. The impedance measurements are used to determine the permittivity using R,L,C,G parameter determination for ideal coaxial cable. The capacitance parameter, C, determines the real part of permittivity while the conductance part, G, determines the imaginary part. The method provides accurate estimates of  air, water and ions like nitrates and chlorides. This work has the potential to be combined with out ongoing work on on-board, real-time and in-situ soil impedance sensors and provide a real time estimate of local soil nitrate and chloride concentrations.