Hendrickx et al. (2026) Field‐Scale Soil Moisture Predictions in Real Time Using In Situ Sensor Measurements in an Inverse Modeling Framework: SWIM ²

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available due to publisher restrictions.

Identification

• Journal: Water Resources Research
• Year: 2026
• Date: 2026-01-30
• Authors: Marit G. A. Hendrickx, Jan Vanderborght, Pieter Janssens, E. Laloy, Sander Bombeke, Evi Matthyssen, Anne Waverijn, Jan Diels
• DOI: 10.1029/2025wr041324

Research Groups

Not explicitly mentioned in the abstract, but the study was conducted in Flanders, Belgium (likely KU Leuven and/or related research institutes).

Short Summary

This study introduces SWIM 2 (Sensor Wielded Inverse Modeling of a Soil Water Irrigation Model), an irrigation decision support system designed as a digital twin. It integrates real-time soil sensor data and periodic soil samples into an FAO-based soil water balance model using the Bayesian inverse modeling algorithm DREAM (ZS). Validated across 18 vegetable cropping cycles in Flanders, Belgium, the system provides robust 7-day soil moisture predictions with uncertainty estimates, even with minimal prior knowledge and biased sensors.

Objective

To develop and validate a real-time framework (SWIM 2) that integrates in situ sensor data into a soil water balance model via inverse modeling to improve field-scale soil moisture predictions and irrigation decision-making.

Study Configuration

• Spatial Scale: Field-scale (agricultural fields in Flanders, Belgium).
• Temporal Scale: Real-time setup across 18 vegetable cropping cycles; 7-day prediction horizon.

Methodology and Data

• Models used: FAO-based soil water balance model; DREAM (ZS) (DiffeRential Evolution Adaptive Metropolis) for Bayesian inverse modeling.
• Data sources: In situ autonomous soil moisture sensors (IoT); unbiased periodic soil samples; in situ precipitation data.

Main Results

• Robust Predictions: Achieved robust soil moisture predictions for a 7-day horizon with accuracies comparable to direct sensor measurements.
• Rapid Calibration: Prediction precision improved substantially within the first 20 days of calibration and remained high throughout the season.
• Bias Correction: The study highlights that independent soil samples and knowledge of error covariance are essential to correct for sensor bias and ensure accurate calibration.
• Data Synergy: Continuous sensor data are vital for capturing dynamics, while periodic unbiased samples ensure accuracy.

Contributions

• Implementation of a Bayesian inverse modeling framework for real-time field-scale soil moisture forecasting.
• Proof-of-concept for a "digital twin" approach to agricultural water management using affordable IoT sensors and minimal prior knowledge.

Funding

Not extractable from the abstract.

Citation

@article{Hendrickx2026FieldScale,
  author = {Hendrickx, Marit G. A. and Vanderborght, Jan and Janssens, Pieter and Laloy, E. and Bombeke, Sander and Matthyssen, Evi and Waverijn, Anne and Diels, Jan},
  title = {Field‐Scale Soil Moisture Predictions in Real Time Using In Situ Sensor Measurements in an Inverse Modeling Framework: SWIM <sup>2</sup>},
  journal = {Water Resources Research},
  year = {2026},
  doi = {10.1029/2025wr041324},
  url = {https://doi.org/10.1029/2025wr041324}
}