Soil dielectric permittivity modelling for 50 MHz instrumentation

Abstract

Near surface electromagnetic geophysical techniques are proven tools to support soil ecosystem services and soil exploration. Such geophysical techniques provide electromagnetic properties that are useful to characterize the studied soil. The link between relevant soil characteristics and geophysical properties, such as dielectric permittivity (ε), is commonly expressed by pedophysical models. However, some weaknesses remain in their application, such as the requirement of parameters that are difficult to measure or calculate. Therefore, these parameters are frequently fixed, but this oversimplifies the complexity of the investigated soils. Moreover, the validity of ε pedophysical models in the frequency range of operating soil moisture sensors (normally < 100 MHz) remains poorly investigated.In this study, the accuracy and adaptability of ε pedophysical models at different electromagnetic frequency ranges was tested and improved using newly collected laboratory and field data. Such data was collected on soils over a wide range of textures, physical and chemical properties.To achieve this, we review the measurement methods and characteristics of ε pedophysical models, soil phases and geometric parameters. Subsequently, we show how geometric parameters can explain the dependance of soil texture on ε by implementing pedotransfer functions. Then, drawing on a broad experimental basis of common soil types in Europe, we develop novel ε pedophysical models at 50 MHz. These models are not only easy to evaluate but also capture most of the soil’s complexity. Additionally, these new ε pedophysical models eliminate the need for calibration data due to the introduction of novel pedotransfer functions based on soil cation exchange capacity. An extensive model test shows an unprecedented decrease in the RMSE of the newly proposed models of up to 412%.In conclusion, despite it is unlikely to characterize soil structure, bulk density, or temperature at 50 MHz, these updated PPMs are useful for highly accurate water content and ε predictions, in both laboratory and field conditions, without the need for calibration data. As the developed modelling procedures are valid for a wide range of electromagnetic frequencies, these can be applied to soil exploration with TDR and GPR instrumentation.For reproducibility, all collected soil data are provided, alongside open-source Python code that contains the presented modelling procedures.