Mapping irrigation types in the northwestern US using deep learning classification

Authors: Nouwakpo, Sayjro; Bjorneberg, David - Dave; MCGWIRE, KENNETH - Desert Research Institute

Submitted to: IEEE Transactions on Geoscience and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/23/2024
Publication Date: 8/1/2024
Citation: Nouwakpo, S.K., Bjorneberg, D.L., Mcgwire, K. 2024. Mapping irrigation types in the northwestern US using deep learning classification. IEEE Transactions on Geoscience and Remote Sensing. 60(8):1-16. https://doi.org/10.1029/2023WR036155.
DOI: https://doi.org/10.1029/2023WR036155

Interpretive Summary: Many agricultural areas of the western United States practice irrigation using a variety of irrigation methods. Maps of irrigation methods are needed but no technology exists to produce these maps at broad spatial scales. In this study, we develop an irrigation methods mapping tool by training a deep learning model on satellite-derived input images. Training data consisted in irrigation methods maps for the state of Utah obtained from the Utah Water Related Land Use (WRLU) dataset and additional labeling in selected areas of southern Idaho. The trained model was able to predict the correct irrigation class 78% of the time. The model performance was stable over time but varied depending on the regions where it was evaluated. Model prediction of sprinkler irrigation in an irrigated watershed of southern Idaho for 2006, 2011, 2013, and 2016 were consistent with previously published survey data. Performance improvements are expected with the utilization of higher resolution satellite products. This methodology provides a tool for water resource managers to estimate irrigation methods in large agricultural areas and identify priority areas in need of irrigation methods conversion.

Technical Abstract: Many agricultural areas of the western United States practice irrigation using a variety of irrigation methods. Maps of irrigation methods are needed but no technology exist to produce these maps at broad spatial scales. In this study, we develop an irrigation methods mapping tool by training a U-Net model on Landsat 5- and 8-derived input images. Training data consisted in irrigation methods classified as Flood (F), Sprinkler (S) or Other (O) on agricultural fields from the Utah Water Related Land Use (WRLU) dataset and additional labeling in selected areas of southern Idaho. An ensemble of 10 trained models had an overall accuracy of 0.78. Precision for F, S and O were 0.73, 0.82 and 0.80 while recall values were 0.75, 0.74 and 0.84 respectively. Model performance was generally stable throughout the training years but varied by areas. The best performance was obtained in regions with uniform irrigation method across large patches while small fields of contrasting irrigation method with their surroundings were inadequately predicted. Model prediction of sprinkler irrigation in an irrigated watershed of southern Idaho for 2006, 2011, 2013, and 2016 were consistent with previously published survey data. Performance improvements are expected with the utilization of higher resolution satellite products. This methodology provides a tool for water resource managers to estimate irrigation methods in large agricultural areas and identify priority areas in need of irrigation methods conversion.