Peña-Arancibia et al. (2026) Integrating Landsat-based irrigation mapping and actual evapotranspiration estimates with hydrological modelling to assess regional irrigation dynamics in a large basin experiencing extreme climate variability: Towards a national operational irrigation water accounting system

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-08-20
• Authors: Jorge L. Peña-Arancibia, Yingying Yu, Tim R. McVicar, Thomas G. Van Niel, Francis H. S. Chiew, Darin Hodgson, Jamie Vleeshouwer, Aarond Dino, Zachary Brown, Anthony Nadelko
• DOI: 10.1016/j.rse.2026.115611

Research Groups

• CSIRO Environment
• CSIRO Agriculture and Food
• Terrestrial Ecosystem Research Network (TERN)

Short Summary

This research developed a multi-decadal (1987–2024), high-resolution (approximately 30 m) remote sensing framework to quantify irrigation water-use (IWU) dynamics across Australia's Murray-Darling Basin. The framework successfully produced maps of irrigated extent, phenology, and net IWU, revealing distinct regional responses to climate variability and policy changes.

Objective

• To develop a remote sensing framework for quantifying long-term irrigation areal extent, vegetation types, and associated net irrigation water-use (IWUnet) in the Murray-Darling Basin (MDB).
• To enable the identification of irrigation extent and IWUnet by developing gap-free monthly approximately 30 m remote sensing time series.
• To evaluate gap-filling and actual evapotranspiration (ETa) estimates.
• To identify broad irrigated agriculture types (crops, pastures, orchards) using an Artificial Neural Network (ANN) classification framework.
• To estimate IWUnet from ETa and rainfall using a water balance model incorporating soil moisture accounting.
• To quantify patterns and long-term changes in irrigation dynamics across the MDB over the period 1987–2024.

Study Configuration

• Spatial Scale: Murray-Darling Basin (approximately 1.06 × 10^6 km²), with field-scale resolution (approximately 30 m).
• Temporal Scale: Multi-decadal (July 1987 to December 2024), monthly.

Methodology and Data

• Models used:
  • Spatiotemporal gap-filling (Wang et al., 2012)
  • Sub-pixel class fraction change information Flexible Spatiotemporal DAta Fusion (SFSDAF) method (Li et al., 2020)
  • CMRSET (CSIRO MODIS Reflectance-based Scaling EvapoTranspiration) model (version 2)
  • Artificial Neural Network (ANN) Self-Organising Maps (SOM)
  • Supervised Pattern Recognition Neural Network (PRNN)
  • Thornthwaite-Mathern soil water balance framework
• Data sources:
  • Satellite: Landsat-5 TM, Landsat-7 ETM+, Landsat-8 OLI, Landsat-9 OLI-2 (L5789) surface reflectance; MODIS Nadir Bidirectional Reflectance Distribution Function Adjusted Reflectance (NBAR) product MCD43A4.
  • Climate: Australian Gridded Climate Data (AGCD) daily precipitation (approximately 5 km resolution).
  • Evapotranspiration: Landsat-based CMRSET V2 ETa estimates.
  • Soils: Soil and Landscape Grid of Australia, Australian Water Outlook (for total available water capacity and saturated hydraulic conductivity).
  • Land Use: Catchment Scale Land Use Mapping for Australia (CLUM), ePaddocks data.
  • Evaluation Data: In situ ETa from eddy covariance flux towers (OzFlux network), multi-scale agricultural survey statistics, irrigation district annual compliance reports, and Murray-Darling Basin Authority (MDBA) water take records.

Main Results

• Gap-filling and blending achieved moderate to high accuracy for NDVI prediction (median R² = 0.77, RMSE = 0.08 for spatiotemporal gap-filling; median R² = 0.65, RMSE = 0.07 for SFSDAF blending).
• CMRSET V2 ETa estimates captured temporal patterns well at an irrigated site (R² = 0.86, nRMSE = 0.26, Bias = 0.04), but showed systematic underestimation of magnitudes when flux tower ETa was energy balance corrected.
• PRNN classification of irrigated land cover types achieved high accuracies (>80% per class) and a weighted Kappa coefficient of 0.91, with very high accuracies (>90%) for cotton-like, rice-like, pasture-like, and fallow/bare classes.
• The developed irrigable area mask identified approximately 4.16 × 10^10 m² of irrigable land, an 18% increase compared to the CLUM mask (approximately 3.45 × 10^10 m²), indicating expansion along headwater rivers.
• IWUnet contributed an average of approximately 39% of total ETa at the Myall Vale B irrigated cotton site over six summer cropping seasons, peaking at approximately 80% during the 2019-2020 Tinderbox drought.
• District to catchment scale comparisons of mapped IWUnet against reported data generally showed moderate to high accuracy (R² > 0.8, nRMSE < 0.3–0.5), though some regional discrepancies were noted due to factors like dryland cotton irrigation or new orchard establishment.
• Temporal dynamics revealed strong interannual fluctuations in irrigated area and IWUnet in the northern MDB (opportunistic, cotton-dominated), contrasting with smoother trajectories and gradual crop-mix shifts in the southern MDB (e.g., orchard expansion in Lower Murray, shift from rice to cotton-rice in Murrumbidgee).

Contributions

• Delivered the first multi-decadal (1987–2024), high-resolution (approximately 30 m), basin-wide mapping of irrigated extent, phenology, and net irrigation water-use (IWUnet) for the Murray-Darling Basin.
• Developed and operationalized a robust framework for generating gap-free, monthly, approximately 30 m NDVI and ETa time series across a large, heterogeneous basin by integrating heterogeneous satellite archives (Landsat and MODIS).
• Introduced an innovative semi-supervised workflow combining Self-Organising Maps (SOM) and a Pattern Recognition Neural Network (PRNN) with NDVI and ETa time series to accurately identify and classify broad irrigated agriculture types, including distinguishing rice-like from cotton-like phenologies.
• Quantified the green/blue water components of IWUnet using a mass-conserving soil water balance model, providing critical insights into irrigation water sources and consumption under varying hydroclimatic conditions.
• Provided unprecedented insights into how long-term irrigation dynamics in the MDB respond to extreme climate variability and policy reforms, highlighting distinct regional adaptation strategies.
• Established a scalable and transferable pathway towards a sustained, operational irrigation information product, addressing a critical data gap for agricultural water-use monitoring and supporting evidence-based decision-making.

Funding

• Murray-Darling Basin Water and Environment Research Program (MD-WERP)
• Australian Government (through the National Collaborative Research Infrastructure Strategy (NCRIS) for the Terrestrial Ecosystem Research Network (TERN))

Citation

@article{PeñaArancibia2026Integrating,
  author = {Peña-Arancibia, Jorge L. and Yu, Yingying and McVicar, Tim R. and Niel, Thomas G. Van and Chiew, Francis H. S. and Hodgson, Darin and Vleeshouwer, Jamie and Dino, Aarond and Brown, Zachary and Nadelko, Anthony},
  title = {Integrating Landsat-based irrigation mapping and actual evapotranspiration estimates with hydrological modelling to assess regional irrigation dynamics in a large basin experiencing extreme climate variability: Towards a national operational irrigation water accounting system},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115611},
  url = {https://doi.org/10.1016/j.rse.2026.115611}
}