Wu et al. (2025) Automated drone-borne GPR mapping of root-zone soil moisture for precision irrigation

Identification

• Journal: Remote Sensing of Environment
• Year: 2025
• Date: 2025-11-05
• Authors: Kaijun Wu, Jean Artois, Denis Tourneur, Merlin Mareschal, Maud Henrion, Sashini Pathirana, Lakshman Galagedara, Quentin Limbourg, Sébastien Lambot
• DOI: 10.1016/j.rse.2025.115110

Research Groups

• Universit´e catholique de Louvain, Earth and Life Institute, Environmental Sciences, Louvain-la-Neuve, Belgium
• Centre Wallon de Recherches Agronomiques, D´epartement Productions Agricoles, Gembloux, Belgium
• School of Science and the Environment, Memorial University of Newfoundland, Corner Brook, Newfoundland and Labrador, Canada

Short Summary

This study demonstrates the potential of drone-borne Ground-Penetrating Radar (GPR) to map spatial and temporal root-zone soil moisture dynamics across an agricultural field over an entire growing season. Using the gprSense® system with full-wave inversion, the research achieved precise and automated time-lapse soil moisture mapping, showing strong agreement with conventional methods and providing actionable insights for precision irrigation.

Objective

• To evaluate the potential of drone-borne GPR for high-resolution, time-lapse monitoring of root-zone soil moisture in an agricultural field during a spinach growing season.

Study Configuration

• Spatial Scale: A 5.72-hectare agricultural field in central Belgium; GPR measurements characterized soil moisture down to approximately 35–40 centimeters (cm) depth; final soil moisture maps had a 1-meter (m) grid resolution.
• Temporal Scale: Monitoring conducted over an entire spinach growing season (August to September 2023), with eight drone-borne GPR acquisition dates.

Methodology and Data

• Models used:
  • gprSense® system (frequency-domain radar with full-wave inversion)
  • Full-wave radar equation for inversion
  • Least-squares optimization in the time domain
  • Precomputed look-up table (LUT) for Green’s functions
  • Topp’s empirical equation for converting relative permittivity to volumetric water content
  • Ordinary Kriging for spatial interpolation
  • Boosted Regression Tree (BRT) model for soil moisture prediction
  • Mass balance approach for empirical characterization depth estimation
• Data sources:
  • Drone-borne GPR (gprSense® system, 110–120 MHz dipole antenna)
  • Time Domain Reflectometry (TDR) sensors (at 10, 20, 30, 40, 50, 60 cm depths)
  • Electrical Resistivity Tomography (ERT) (Veris Q2800 system, 30 cm depth used)
  • In-situ rain gauges
  • Sencrop weather station (daily precipitation, temperature, wind speed)
  • Raindancer application (irrigation events, GPS coordinates, water amount)
  • Digital Terrain Model (DTM) from G´eoportail de la Wallonie (altitude, flow accumulation index, slope)

Main Results

• The drone-borne GPR system successfully generated eight high-resolution (1 m grid) soil moisture maps over the growing season, capturing dynamic variations driven by precipitation and irrigation.
• The GPR system, operating in the 110–120 MHz frequency range, measured soil moisture down to an effective characterization depth of approximately 35–40 cm.
• Time-series analysis revealed that mean GPR-derived soil water content closely corresponded to rainfall and irrigation events.
• A strong correlation (Pearson correlation coefficient r = 0.7688) was observed between GPR-derived soil moisture and TDR measurements at 30 cm depth, consistent with the estimated radar characterization depth.
• RMSE values for GPR-derived versus TDR-measured soil moisture ranged from 0.0582 m³/m³ (at 60 cm) to 0.1319 m³/m³ (at 10 cm), with a value of 0.1188 m³/m³ at 30 cm.
• Spatial patterns of GPR-derived soil moisture aligned with predictions from Boosted Regression Tree (BRT) models (self-correlation r = 0.602, self-RMSE = 0.0258) and with rainfall/irrigation data collected by rain gauges (correlation r = 0.78).
• Consistently wetter areas were observed in the northwest corner of the field, correlating with higher soil electrical conductivity and topographical features (talweg).
• Slight spatial shifts in wetter areas relative to expected irrigation zones were attributed to wind drift from irrigation cannons.

Contributions

• This is the first study to implement time-lapse, root-zone soil moisture mapping using drone-borne GPR combined with real-time full-wave inversion.
• It demonstrates a complete, physically-based radar model integrated into a lightweight, automated platform suitable for operational deployment in agricultural fields.
• The approach provides non-invasive, spatially continuous monitoring of root-zone dynamics with minimal human intervention, addressing key limitations of existing in-situ, remote sensing, and modeling methods.
• The drone-borne GPR serves as a bridging tool, offering high-resolution ground-truthing for satellite-based soil moisture products and facilitating cross-scale integration of soil moisture information.
• The low-frequency operation (110–120 MHz) ensures the system is largely insensitive to most crop canopies and surface roughness, making the approach broadly transferable across various crop types and agricultural soils.

Funding

• DuraTechFarm project (Contract No. D65-7390) funded by the R´egion Wallonne (Belgium) and the Fonds de la Recherche Scientifique (FNRS, Belgium).
• gprSense® development supported by the agROBOfood Project (MIRAGE, Grant Agreement No. 825395) and the ICAERUS Project (gprSense®, Grant Agreement No. 101060643) through Open Calls, funded by the European Union’s Horizon Europe Research and Innovation Program.

Citation

@article{Wu2025Automated,
  author = {Wu, Kaijun and Artois, Jean and Tourneur, Denis and Mareschal, Merlin and Henrion, Maud and Pathirana, Sashini and Galagedara, Lakshman and Limbourg, Quentin and Lambot, Sébastien},
  title = {Automated drone-borne GPR mapping of root-zone soil moisture for precision irrigation},
  journal = {Remote Sensing of Environment},
  year = {2025},
  doi = {10.1016/j.rse.2025.115110},
  url = {https://doi.org/10.1016/j.rse.2025.115110}
}