Huang et al. (2025) A GCN-Attention Model for Precision Irrigation Evaluation

Identification

• Journal: Electronics ETF
• Year: 2025
• Date: 2025-11-18
• Authors: Ying Huang, Meng Liu
• DOI: 10.53314/els2529070h

Research Groups

Faculty of Electrical Engineering, University of Banja Luka

Short Summary

This paper proposes UFOGCN-SPANet, a novel and computationally efficient GCN-attention model for precision irrigation evaluation, which integrates a linear-complexity Vision Transformer, Graph Convolutional Networks, and a Salient Positions-based Attention Network to overcome limitations of traditional irrigation methods and improve performance in resource-constrained agricultural settings.

Objective

• To address the challenges of traditional quantitative irrigation methods, including their inability to adapt to dynamic soil moisture and meteorological changes, issues of hysteresis and adaptability in soil moisture threshold methods, lack of comprehensive consideration of meteorological factors and growth dynamics, and limitations in processing efficiency and sensitivity to subtle soil moisture changes, by developing a computationally efficient and effective context modeling architecture.

Study Configuration

• Spatial Scale: Not explicitly defined, but implied for precision agriculture applications and real-world irrigation systems.
• Temporal Scale: Dynamic, continuous, for time series forecasting of soil moisture and meteorological changes.

Methodology and Data

• Models used: UFOGCN-SPANet, a cascaded architecture comprising:
  1. Unit Force Operated Vision Transformer (UFOViT) for efficient global spatio-temporal feature extraction.
  2. Graph Convolutional Networks (GCNs) for modeling spatial dependencies.
  3. Salient Positions-based Attention Network (SPANet) with a Significant Position Selection (SPS) algorithm for dynamic computational focus. The model was compared against traditional GNN models (e.g., SAGEConv) and 12 standard time series forecasting methods.
• Data sources: Implied from actual soil moisture content and meteorological changes; specific data sources (e.g., satellite, observation, reanalysis) are not detailed in the provided text.

Main Results

• The proposed UFOGCN-SPANet method significantly outperforms traditional GNN models, such as SAGEConv, and 12 standard time series forecasting methods across key metrics including accuracy, precision, recall, and F1-Score.
• The model demonstrates enhanced discriminative power while drastically reducing computational complexity, making it suitable for resource-constrained precision agriculture.

Contributions

• Introduction of UFOGCN-SPANet, a novel and computationally efficient architecture specifically designed for resource-constrained precision agriculture.
• Innovation through the cascaded integration of a linear-complexity UFOViT, GCNs, and a SPANet employing a novel SPS algorithm.
• Direct addressing of critical challenges in real-world irrigation systems concerning computational efficiency and effective context modeling.

Funding

Not specified in the provided text.

Citation

@article{Huang2025GCNAttention,
  author = {Huang, Ying and Liu, Meng},
  title = {A GCN-Attention Model for Precision Irrigation Evaluation},
  journal = {Electronics ETF},
  year = {2025},
  doi = {10.53314/els2529070h},
  url = {https://doi.org/10.53314/els2529070h}
}