Advancing Field-Scale Soil Mapping Using An Autonomous Rover With Multi-Layer Geophysical Sensors

Geophysical methods are non-invasive techniques employed to measure the physical properties of the investigated media—primarily electrical and mechanical—while preserving the dynamics of soil structure without altering its state. These methods can be used qualitatively to detect soil anomalies and spatial heterogeneities, as well as quantitatively to correlate primary soil properties with physical measurements. Soil compaction resulting from traffic with modern agricultural machinery has significantly increased, leading to substantial impacts on soil ecosystem services and crop yields.
The quantification of soil structure and compaction has traditionally been performed through destructive soil sampling followed by, dry bulk density and porosity measurements, or through inferential methods (e.g., pedotransfer functions).
This study investigated the potential of an integrated geophysical approach using autonomous driving rover (Robotti 150D, Agrointelli-DK) for mapping soil variability and compaction status on arable land.
The experiment was conducted at the L. Toniolo experimental farm of the University of Padua on a 1-ha field comprising a complete randomize design testing two traffic treatments (conventional and controlled traffic with autonomous guidance vehicle) and four replicates covering 8 plots (130 m x 10 m). The geophysical instruments mounted on the rover included: a γ-ray detector (Agri Detector MS-2000, Medusa - NL) positioned at the front; a GPR (Stream DP, IDS - IT) and a cosmic ray neutron sensing probe (Finapp - IT) in the central section; and an electromagnetic conductivity meter (CMD-MiniExplorer, GF Instruments - CZ) mounted on a wooden sled at the rear. Measurements were conducted at a speed of 3.6 km h⁻¹, with swaths spaced every two meters. All instruments operated simultaneously and were connected to a GPS equipped with an RTK positioning system, ensuring a precision of 2 cm.
Moreover, eight (1 per plot) 3D electrical conductivity tomographies (ERTs) (Syscal Terra, IRIS - FR) were performed for each replicate on a ca. 3 m³ investigated volume (4.6 × 0.8 × 0.8 m) using a dipole-dipole array. Geophysical techniques were then complemented by traditional destructive measurements of bulk density (core method), soil penetration resistance (Eijkelkamp - NL) and soil texture on the top 1 m and by drone surveys to create a digital elevation model (DEM).
Preliminary results demonstrated that the combination of an autonomous robot with several multi-layer geophysical sensors can act as a proxy for expeditive digital soil mapping on large surfaces. Nevertheless, the ERT capability to capture the presence of resistivity anomalies and its combination with traditional method seemed fundamental to precisely adjust the multi-mapping survey.
In conclusion, the tested approach might provide a consistent set of real time data valuable also for training machine learning algorithms and give new insight to precision agriculture technique.