Pedophysics: a python package for soil geophysics.

Near surface geophysical electromagnetic techniques are proven tools to support ecosystem services such as agriculture, soil remediation, nutrient management, and heritage conservation. Key influencing geophysical properties are electrical conductivity (σ, or resistivity (ρ)), dielectric permittivity (𝜀) or magnetic susceptibility (μ), that are targeted to model soil properties and state variables such as texture, bulk density, cation exchange capacity (CEC) or water content. To translate geophysical properties into quantitative information on the targeted soil properties, relationships between these have to be considered in appropriate models. These so-called pedophysical models can then be integrated into interpretation schemes (e.g., after inversion, or through incorporating this into forward modelling procedures). This modelling step, translating geophysical properties into soil properties (and vice versa), thus constitutes a key aspect of near surface exploration.

While hundreds of pedophysical models exist to perform this task, these often depend on many properties and parameters defined within a specific range e.g., electromagnetic frequency, texture, and salinity; impeding applications to cases where information about the studied soil is scarce. Therefore, selecting an appropriate pedophysical model for a given scenario is often a very complex task.

To facilitate solutions for pedophysical modelling we present pedophysics, an open source python package for soil geophysical characterization. The package implements up-to-date models from the literature and, based on the user’s needs, automatically provides an optimal solution given a set of input parameters and the targeted output.

First, a virtual soil is defined by inputting any of its available properties. This soil can be defined in discrete states to simulate the evolution of its properties over time. Secondly, a module (predict) is called to predict the target property of interest. Following this workflow, for example, a soil with a given texture and changing water content could be defined to obtain its 𝜀 or σ at a predefined frequency, or, inversely, its water content could be predicted based on changing σ.

However, as soil properties required as input parameters for pedophysical models are often unknown, it can, in such cases, be impossible to obtain a viable prediction outcome. The pedophysics package accounts for such limitations by implementing pedotransfer functions, that allow obtaining the missing properties from the available ones. For example, if CEC is unknown, it is determined based on soil texture and a location.

In summary, the package synthesizes specific pedophysical modeling knowledge. Time-varying properties can be calculated in a straightforward way, and, through the integration of pedotransfer functions, target properties can be predicted with a minimum of information about the studied soil. Thus, by translating known properties to targeted ones, pedophysics is contributing to improve interpretability of near surface modeling schemes; enhancing soil electromagnetic geophysical exploration techniques in ecosystem services applications.