Comparison of empirical and partly deterministic methods of time domain reflectometry calibration, based on a study of two tropical soils

The use of time domain reflectometry for the determination of volcanic soils' moisture content in tropical areas has shown the necessity to set up an accurate calibration. We chose to test the most representative calibration models on two tropical volcanic soils coming from two different climatic areas of the Réunion Island. The empirical model of Topp et al. [Topp, G.C., Davis, J.L., Annan, A.P., 1980. Electromagnetic determination of soil water content: measurements in coaxial transmission lines. Water Resources Research, 16 (3) 574–582.] was found to offer the best accuracy/simplicity trade-off for simulations with a cambisol, with an accuracy of ±2.46 dielectric constant points (or about ±2.7% for a water content close to 45%). In contrast, it gave poorer results for analysis of the andosol containing the greatest quantity of amorphous minerals (±5.74 dielectric constant points). The partly deterministic models were found to be better for this latter application, including that of Ansoult et al. [Ansoult, M., De Backer, L.W., Declercq, M., 1985. Statistical relationship between apparent dielectric constant and water content in porous media. Soil Sci. Soc. Am. J., 49, 47–50.] which gave the best accuracy (±3.84 dielectric constant points). For both soil materials studied, there was only a slight gain of accuracy by introducing a specific polynomial calibration determined in the laboratory: ±1.72 and ±3.46 dielectric points for the cambisol and the andosol, respectively. Choices of models to be used should thus be based on the type of soil to be analysed and accuracy requirements.