What is it about?
Farms use different irrigation methods—like flooding fields or using sprinklers or drip systems—to water crops. Choosing the right method affects how much water is used and how efficiently it's applied. However, there hasn’t been a large, detailed dataset that shows how these methods are used across different areas. Our work introduces IrrMap, the first large-scale dataset that uses satellite images to identify how farms are irrigated. It covers over 1.6 million farms and 14 million acres across the western United States from 2013 to 2023. We used satellite data from LandSat and Sentinel along with information about crop types, land use, and vegetation to build over 1 million ML-ready image patches. Researchers can use this dataset to develop AI models that better monitor irrigation practices, detect changes over time, and support water-saving strategies in agriculture. Everything—from data to code—is openly shared to help others use and build upon our work.
Featured Image
Photo by Pascal Meier on Unsplash
Why is it important?
Water is becoming increasingly scarce, especially in agricultural areas. Since irrigation accounts for 70% of global freshwater use, understanding how water is applied on farms is crucial for sustainable agriculture. By providing a high-quality, open-source dataset, our work enables machine learning and environmental research communities to track irrigation methods, detect inefficiencies, and support smarter water policies. This dataset also lays the foundation for developing agricultural AI models that can generalize across regions—something that wasn't previously possible at this scale.
Perspectives
As a researcher passionate about the intersection of AI, sustainability, and agriculture, this project was particularly meaningful. It took months of effort to harmonize diverse datasets, overcome geospatial alignment issues, and ensure the quality of over a million satellite image patches. What excites me most about IrrMap is its potential to democratize access to irrigation intelligence. By making high-resolution irrigation method labels publicly available for the first time, we’re giving researchers, policy-makers, and practitioners a tool they can build on—from tracking water use trends to training region-specific AI models. I also learned a great deal about the nuances of working with large-scale remote sensing data and the importance of balancing methodological rigor with usability. Seeing the community already engaging with the dataset is incredibly rewarding—and I hope it inspires similar work in agricultural AI globally.
Nibir Chandra Mandal
Read the Original
This page is a summary of: IrrMap: A Large-Scale Comprehensive Dataset for Irrigation Method Mapping, August 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3711896.3737380.
You can read the full text:
Contributors
The following have contributed to this page
