Zhu et al. (2025) Large-scale irrigation area mapping: Status and challenges

Identification

• Journal: Agricultural Water Management
• Year: 2025
• Date: 2025-12-02
• Authors: Wanxue Zhu, Elif Dönmez, Hugo Storm, Thomas Heckelei, Stefan Siebert
• DOI: 10.1016/j.agwat.2025.110037

Research Groups

• Department of Crop Sciences, University of G¨ottingen, Germany
• Guangdong Key Laboratory of Integrated Agro-environmental Pollution Control and Management, Institute of Eco-environmental and Soil Sciences, Guangdong Academy of Sciences, China
• Institute for Food and Resource Economics (ILR), University of Bonn, Germany

Short Summary

This paper provides a comprehensive synthesis of large-scale irrigation area mapping methodologies and datasets, benchmarking ten global and regional products against EUROSTAT 2020 gridded statistics in Europe. It concludes that integrating ground-based statistics with geospatial information significantly improves mapping accuracy, especially in humid regions, while highlighting the critical need for standardized, accessible, and high-quality ground-truth and statistical data.

Objective

• To consolidate the current landscape of global and regional irrigation area datasets by clarifying their characteristics, interrelations, limitations, and identifying key spatial, temporal, and thematic gaps.
• To benchmark existing irrigation datasets across Europe against the EUROSTAT 2020 gridded-type statistics to assess the influence of ground-based information on mapping accuracy.
• To synthesize prevailing irrigation mapping frameworks, examining their methodological foundations, key assumptions, and representative applications, and propose pathways to enhance data standardization, interoperability, and near-real-time monitoring.

Study Configuration

• Spatial Scale: Global, regional (Europe, United States, China, India), national, and subnational (NUTS0, NUTS1, NUTS2, county/city level). Gridded resolutions range from 10 meters to 5 arc-minutes (approximately 9 kilometers), with evaluation conducted at 1 kilometer and 10 kilometer grid resolutions.
• Temporal Scale: Datasets cover periods from 1890 to 2020, with specific reference years (e.g., 2000, 2005, 2010, 2015, 2020) and temporal resolutions including yearly, 2-3 yearly, 5 yearly, decadal, monthly, and seasonal updates. The primary benchmark for evaluation is EUROSTAT 2020.

Methodology and Data

• Models used:
  • Unsupervised Classification: ISODATA, Continuous Change Detection Classification (CCDC).
  • Supervised Classification: Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Categorical Boosting (CB), Classification and Regression Tree (CART).
  • Deep Learning (DL): Convolutional Neural Network (U-net architecture, Pivot-Net).
  • Statistical Disaggregation (DA): Conditional disaggregation, iterative disaggregation.
  • Other: Cross-Entropy Model (CEM), Process-based modeling (PM), Threshold filtering (TH), Bayesian Updating of Land-Cover algorithm (BULC), Visual Interpretation (VI).
• Data sources:
  • Satellite Remote Sensing: MODIS, Landsat, Sentinel-1, Sentinel-2, AVHRR.
  • Survey-based Statistics: FAOSTAT, AQUASTAT, EUROSTAT (2018, 2020, 2025), LUCAS, USDA–NASS Agricultural Census, Irrigation and Water Management Survey (IWMS), China Statistical surveys, India Agricultural statistics, Minor Irrigation Statistics (RMIS).
  • Reanalysis Products.
  • Land Cover and Land Use (LULC) Datasets: CORINE Land Cover, USGS GLCC, WorldCereal, GLC_FCS30D.
  • Ancillary Geospatial Information: Meteorological data, elevation data, soil properties, crop types and phenology, agricultural suitability.
  • Ground-truth Data: Field observations, sample plots.

Main Results

• Evaluation of Actually Irrigated Area (AAI) Datasets in Europe: ECIRA 2020 showed the highest agreement with the EUROSTAT AEI benchmark (R² = 0.67, d = 0.89, slope = 0.83), exhibiting systematic underestimation as expected for AAI compared to AEI. GRIPC 2005 achieved near-unbiased totals (slope = 1.02) but moderate spatial correlation (R² = 0.51, d = 0.81), with overestimation in southern Europe and underestimation in central Europe. Remote-sensing-only products (GIAM 2000, GMIE 2010–2019) showed weaker agreement (R² = 0.15 and 0.27; d = 0.49 and 0.64, respectively), with GMIE significantly overestimating irrigation across Central and Eastern Europe (up to tenfold in some countries).
• Evaluation of Area Equipped for Irrigation (AEI) Datasets in Europe: All AEI datasets showed strong national-level consistency with EUROSTAT reports, with total areas differing by less than ±13%. ELIAD 2020 achieved the highest agreement at national and NUTS2 levels (R² = 0.93, d = 0.98). MEIER 2015 and the GMIA series (2005, 2010, 2015) broadly captured continental-scale irrigation patterns with moderate-to-strong spatial correlations (R² = 0.59–0.77; d = 0.85–0.93). GMIA 2005 showed larger deviations (RMSE of 8.6 x 10^3 hectares, MAE of 4.3 x 10^3 hectares) due to temporal mismatches and structural shifts.
• General Findings: The integration of ground-based statistics with geospatial information markedly improves mapping accuracy, particularly in humid regions where conventional pixel-wise remote sensing is inherently constrained. The scarcity, inconsistency, and inaccessibility of high-quality ground truth and statistical data remain the main obstacles to globally consistent and reliable irrigation mapping.

Contributions

• Provides the first comprehensive synthesis of conceptual foundations, inherent limitations, comparative strengths, and developmental trends of existing large-scale irrigation mapping methodologies and datasets.
• Benchmarks ten global and regional irrigation datasets across Europe against the recently released, high-resolution EUROSTAT 2020 gridded-type statistics, offering transferable insights for assessing dataset reliability and applicability worldwide.
• Identifies scarce, inconsistent, and restricted-access ground-truth and statistics data as the primary bottleneck in irrigation mapping and proposes concrete response strategies for future research, including strengthening data infrastructure and integrating socio-economic dimensions.
• Integrates methodological, data-driven, and operational perspectives to delineate the current state of global irrigation mapping and charts a pathway toward precision monitoring, empowering robust water resource modeling and safeguarding agricultural ecosystem resilience.

Funding

• Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – SFB 1502/1–2022 - Projektnummer: 450058266.
• DFG within Germany’s Excellence Strategy, EXC-2070 – 390732324 (PhenoRob).
• Open Access Publication Funds of the G¨ottingen University.

Citation

@article{Zhu2025Largescale,
  author = {Zhu, Wanxue and Dönmez, Elif and Storm, Hugo and Heckelei, Thomas and Siebert, Stefan},
  title = {Large-scale irrigation area mapping: Status and challenges},
  journal = {Agricultural Water Management},
  year = {2025},
  doi = {10.1016/j.agwat.2025.110037},
  url = {https://doi.org/10.1016/j.agwat.2025.110037}
}

Generated by BiblioAssistant using gemini-2.5-flash (Google API)