Paciolla et al. (2026) Irrigation quantification through ensemble physical modelling

Identification

• Journal: Agricultural Water Management
• Year: 2026
• Date: 2026-10-17
• Authors: Nicola Paciolla, michele Polletta, Francesco Morari, Chiara Corbari
• DOI: 10.1016/j.agwat.2026.110866

Research Groups

• Politecnico di Milano, Milan, Italy
• Università di Padova, Padua, Italy

Short Summary

This study developed a Monte Carlo ensemble physical modeling methodology to quantify irrigation incidence and volume by integrating ground and satellite data with an energy-water balance model. The algorithm proved most effective for fields with fewer irrigation events and during high-evapotranspiration periods, though its accuracy is sensitive to satellite data uncertainty, particularly in heterogeneous agricultural landscapes.

Objective

• To develop and test a novel methodology for detecting and quantifying irrigation incidence and volume at the field scale, integrating a Monte Carlo ensemble approach with multi-source ground and satellite data and an energy-water balance model (FEST-EWB), aiming for applicability in data-scarce environments with relatively low uncertainties.

Study Configuration

• Spatial Scale: Field scale (e.g., 1.08 ha to 33 ha fields) across various Italian case studies. Satellite data used had resolutions ranging from 10 m (Sentinel-2) to 100 m (Landsat-8/9 thermal) and 1 km (Copernicus SSM).
• Temporal Scale: Time-continuous model simulations, with daily averages for hydrological variables. Satellite data revisit times ranged from 4 days (Sentinel-1) to 5 days (Sentinel-2) and 16 days (Landsat-8/9). Irrigation event comparisons were made at 1-day, bi-weekly, and monthly aggregation intervals.

Methodology and Data

• Models used:
  • FEST-EWB hydrological model (Corbari et al., 2011)
  • FEST-EWB-2layers (for superficial and deep soil moisture)
  • FEST-2x2-EWB (two-source configuration for heterogeneous crops)
  • Monte Carlo ensemble approach for simulating multiple irrigation schedules.
• Data sources:
  • Ground-based observations: In-situ soil moisture sensors (TDR instruments TEROS 10 at 10 cm, 30 cm, 50 cm depths), meteorological stations (METER ATMOS 41), water table depth sensors (HYDROS 21), eddy-covariance stations (for evapotranspiration), in-situ radiometers (for land surface temperature).
  • Satellite data:
    • Copernicus Surface Soil Moisture 1 km Version 1 product (from Sentinel-1 SAR, C-band).
    • NASA Landsat-8/9 (Land Surface Temperature at 100 m, Albedo, Leaf Area Index, Vegetation Fraction at 30 m).
    • ESA Sentinel-2 (Albedo, Leaf Area Index, Vegetation Fraction at 10 m to 60 m).
  • Reference data: Observed irrigation records (volume and timing) for benchmark simulations.

Main Results

• The algorithm performs best with fewer irrigation events (less than 10 events), as very frequent events (more than 2–3 per week) obscure the signal.
• High-evapotranspiration periods enhance the ensemble spread, improving the algorithm's ability to distinguish between different irrigation schedules.
• Uncertainty in satellite surface soil moisture (SSM) retrievals, particularly over heterogeneous agricultural landscapes, negatively impacts the algorithm's accuracy.
• Surface soil moisture (SSM) was identified as the most suitable hydrological variable for the algorithm, outperforming deep soil moisture (SM2), evapotranspiration (ET), and land surface temperature (LST) due to its larger ensemble spread and direct response to irrigation.
• Deep soil moisture (SM2) and ET showed limited ensemble spread (maximum 0.03 m³ m⁻³ for SM2, 3.7 mm d⁻¹ for ET), making them generally unsuitable for the procedure. LST also had a narrow spread (often less than 10 °C) except under specific high-energy and dense canopy conditions.
• Using a benchmark simulation (fed with observed irrigation) as a reference yielded the smallest errors. Comparisons with in-situ and satellite observations introduced model error but maintained comparable accuracy.
• The algorithm can estimate irrigation volumes and event numbers, with SSM-based extraction against benchmark simulation showing the lowest errors.
• The coefficient of determination (R²) between observed and synthetic irrigation events was low for 1-day predictions (less than 0) but significantly increased with temporal aggregation, reaching 0.5–0.9 for monthly intervals.

Contributions

• Presents a novel, physically-based Monte Carlo ensemble modeling approach for irrigation quantification, integrating diverse ground and satellite data.
• Offers a method specifically tailored for detailed field-level studies, contrasting with existing literature that often focuses on larger spatial domains (e.g., irrigation districts or hydrological basins) and coarser temporal aggregations.
• Demonstrates the potential for irrigation detection and quantification in data-scarce environments by leveraging satellite observations, while explicitly addressing the limitations and uncertainties of such data.
• Provides insights into the optimal conditions (e.g., irrigation frequency, hydrological variable choice, climate periods) for effective application of the proposed methodology.

Funding

• European Space Agency (ESA) within the HYPERCROP project (HYPER-resolution CROP yield estimates and extreme events crops shocks monitoring by integrating multiple satellite data and monitoring).

Citation

@article{Paciolla2026Irrigation,
  author = {Paciolla, Nicola and Polletta, michele and Morari, Francesco and Corbari, Chiara},
  title = {Irrigation quantification through ensemble physical modelling},
  journal = {Agricultural Water Management},
  year = {2026},
  doi = {10.1016/j.agwat.2026.110866},
  url = {https://doi.org/10.1016/j.agwat.2026.110866}
}