Muturi et al. (2026) An integrated satellite remote sensing approach for mapping field-scale irrigation dynamics to support water management in highly regulated semi-arid catchments

Identification

• Journal: Agricultural Water Management
• Year: 2026
• Date: 2026-07-23
• Authors: Jackline W. Muturi, Sayantan Majumdar, Christopher E. Ndehedehe, Nathan O. Agutu, Mark J. Kennard
• DOI: 10.1016/j.agwat.2026.110653

Research Groups

• School of Environment & Science, Griffith University, Nathan, QLD, Australia
• Australian Rivers Institute, Griffith University, Nathan, QLD, Australia
• Division of Hydrologic Sciences, Desert Research Institute, Reno, NV, USA
• Jomo Kenyatta University of Science and Technology, Kenya

Short Summary

This study develops and evaluates a simple thresholding approach using the fraction of reference evapotranspiration (ET0F) and enhanced vegetation index (EVI) to map field-scale irrigation dynamics in the Namoi catchment, Australia, from 2019 to 2025. The method serves as a conservative classifier with moderate accuracy (61.4%) and specificity (79.9%), proving useful as a low-cost screening tool for identifying potential irrigated areas.

Objective

• Quantify the spatio-temporal variability of irrigated cropland in the Namoi catchment from 2019 to 2025.
• Compare the satellite-derived irrigated areas with existing historical irrigated area datasets to identify and interpret discrepancies.
• Validate the performance and assess the suitability and operational value of the ET0F-EVI thresholding approach for discriminating irrigated areas from rainfed cropland in semi-arid regions.

Study Configuration

• Spatial Scale: Namoi catchment (approximately 42,000 km²). Input data resolutions range from 10 m (Sentinel-2) to 5 km (rainfall), with final irrigation classification outputs analyzed at an approximate spatial resolution of 1 km.
• Temporal Scale: A 7-year period from 2019 to 2025, with monthly EVI composites, monthly actual evapotranspiration (ETa), dekadal/monthly reference evapotranspiration (ET0), and daily rainfall data.

Methodology and Data

• Models used:
  • Thresholding approach based on the fraction of reference evapotranspiration (ET0F) and Enhanced Vegetation Index (EVI).
  • Receiver Operating Characteristic (ROC) analysis with Youden Index for optimal threshold selection.
  • Bootstrap resampling (n = 500 iterations) for uncertainty quantification of irrigated area estimates.
• Data sources:
  • Satellite: Sentinel-2 Level-2A surface reflectance data (for EVI).
  • Derived/Reanalysis:
    • Actual crop evapotranspiration (ETa) from the CSIRO MODIS-Constrained Regional Scale Estimate of Evapotranspiration (CMRSET) product (Guerschman et al., 2022).
    • Gridded grass reference evapotranspiration (ET0) from the FAO WaPOR dataset (FAO, 2018).
    • Gridded monthly rainfall data from the Scientific Information for Land Owners (SILO) 5 km Australian grid dataset (Jeffrey et al., 2001).
  • Ancillary/Reference:
    • Manually digitized training and independent validation polygons of irrigated and non-irrigated areas from Google Earth Pro imagery.
    • Cropland mask of the Namoi catchment derived from the Catchment Scale Land Use of Australia (CLUM) version 2 dataset (ABARES, 2024).
    • Historical irrigated area datasets for comparison: Hung et al. (2025) (60 m resolution) and Nagaraj et al. (2021) (9 km resolution).

Main Results

• Annual irrigated area in the Namoi catchment exhibited considerable interannual variability, ranging from approximately 1882 km² in 2023 to 4631 km² in 2024.
• Irrigated areas were consistently concentrated in the central and south-eastern portions of the catchment, forming a noticeable band, with limited areas showing persistent irrigation across all years.
• A weak positive correlation (r = 0.27, R² = 0.07) was observed between bootstrap-derived mean irrigated area estimates and mean seasonal rainfall, suggesting other factors significantly influence irrigation decisions.
• The ET0F-EVI thresholding approach achieved a moderate overall accuracy (61.4%) and specificity (79.9%), but low sensitivity (38.9%) and classification agreement (Kappa = 0.19), indicating its behavior as a conservative classifier.
• Bootstrap analysis revealed substantial uncertainty in irrigated area estimates associated with threshold selection, with wide 95% confidence intervals across years.
• Comparison with historical datasets showed spatial heterogeneity and dynamic changes in irrigation activity, with only 162 km² consistently identified as persistently irrigated by all three datasets (this study, Hung, Nagaraj). The recent dataset uniquely identified a substantial 7375 km² as irrigated.

Contributions

• Develops and evaluates a simple, transferrable, and data-efficient thresholding approach for mapping irrigated areas using freely available satellite-derived ET0F and EVI, avoiding complex machine learning models.
• Integrates both vegetation greenness and normalized water use signals, offering a more robust classification of irrigated areas, particularly in semi-arid regions.
• Provides a low-cost screening tool for identifying potential irrigated areas, supporting applications such as irrigation hotspot detection, baseline mapping, and prioritization of field-scale irrigation surveillance.
• Offers insights into the spatio-temporal dynamics of irrigation in a highly regulated semi-arid catchment, highlighting the influence of short-term climatic conditions, water availability, and infrastructure on irrigation patterns.
• Demonstrates the importance of careful selection and evaluation of vegetation metrics in threshold-based irrigation classification and suggests avenues for methodological improvement.

Funding

• Faculty for the Future Fellowship program by the Schlumberger Foundation (Jackline W. Muturi).
• Australian Research Council (DE230101327) (Christopher E. Ndehedehe).

Citation

@article{Muturi2026integrated,
  author = {Muturi, Jackline W. and Majumdar, Sayantan and Ndehedehe, Christopher E. and Agutu, Nathan O. and Kennard, Mark J.},
  title = {An integrated satellite remote sensing approach for mapping field-scale irrigation dynamics to support water management in highly regulated semi-arid catchments},
  journal = {Agricultural Water Management},
  year = {2026},
  doi = {10.1016/j.agwat.2026.110653},
  url = {https://doi.org/10.1016/j.agwat.2026.110653}
}