Abstract
Water-saving irrigation is critical for improving water use efficiency and ensuring sustainable food production. However, accurate prediction of farmland water demand remains difficult because soil, crop, and meteorological processes interact across space, depth, and time. To address this problem, we develop an AI-STGNN-based irrigation decision framework in which AI-STGNN serves as the core predictive model. The main methodological contribution is a dynamic three-dimensional (3D) graph that represents vertical soil heterogeneity and time-varying hydrological connectivity, while channel-wise meteorological attention and multi-scale temporal memory are incorporated as agro-hydrology-specific enhancements. On the held-out 2021–2022 test set, AI-STGNN achieved RMSE, MAE, and R² values of 3.63 vol%, 2.88 vol%, and 0.89, respectively, reducing RMSE by 7.2% relative to STGCN and by 16.7% relative to LSTM. Under heavy rainfall, its RMSE increased by 45%, compared with 59–70% for the baseline models. In the two-season field comparison across three representative farms, the AI-STGNN-guided strategy reduced irrigation water use by 12.3–15.7%, increased crop yield by 3.5–4.2%, and improved water use efficiency by 13.7–17.8% relative to conventional empirical irrigation. Because the field comparison covers only two growing seasons and three sites, these practical gains should be interpreted as preliminary rather than definitive. Overall, the proposed framework provides a physically informed, data-driven basis for dynamic water-demand prediction and irrigation scheduling in the studied Midwest cropping system.
Similar content being viewed by others
Alleviating water scarcity by optimizing crop mixes
Participatory evaluation of an irrigation decision support system for water-saving and productivity gains in Lake Urmia Basin
Selection of alfalfa water and nitrogen management regimes based on the DSSAT model
Author information
Authors and Affiliations
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Funding
This study was supported by the Basic Research Project of Liaoning Provincial Education Department (LJ222510157006; LJ212510157030; LJ212510157002).
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Reprints and permissions
About this article
Cite this article
Fang, Y., Zheng, J., Shen, X. et al. Optimization of water-saving irrigation decision using dynamic spatiotemporal graph neural networks.
Sci Rep (2026). https://doi.org/10.1038/s41598-026-54923-0
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-026-54923-0
Keywords
- Spatiotemporal graph neural network
- Dynamic water demand prediction
- Irrigation decision optimization
- Multimodal data fusion
Source: Resources - nature.com
