Tian et al. (2025) An optimization framework for intelligent irrigation system installation in fragmented paddy fields

Identification

• Journal: Smart Agricultural Technology
• Year: 2025
• Date: 2025-12-19
• Authors: Runze Tian, Kyoji Takaki, Toshiaki Iida, Keigo Noda, Yohei Asada
• DOI: 10.1016/j.atech.2025.101740

Research Groups

• Department of Biological and Environmental Engineering, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Japan
• Department of Rural Environmental Sciences, Faculty of Agriculture, Iwate University, Japan
• Institute of Agriculture, Tokyo University of Agriculture and Technology, Japan

Short Summary

This study proposes and validates a two-stage optimization framework for Intelligent Irrigation System (IIS) unit installation in fragmented paddy fields, aiming to maximize labor reduction by optimizing patrol routes. The framework effectively reduces computational complexity and maintains solution quality, particularly under non-uniform spatial distributions, providing guidance for smart agricultural technology deployment.

Objective

• To develop and validate a field-selection framework for the installation of IIS units in fragmented paddy fields, identifying optimal field combinations that maximize labor reduction benefits under different installation numbers and farmer datasets, using patrol-route distance reduction as the evaluation metric.

Study Configuration

• Spatial Scale: Paddy field data from three rice farmers in Chiba Prefecture, Japan (Sakae Town and Sakura City). Farmer 1: 56 fields; Farmer 2: 82 fields; Farmer 3: 36 fields. Typical paddy field plots are approximately 30 m x 100 m, with inter-field distances in the order of hundreds of meters.
• Temporal Scale: Patrols are simulated twice daily (morning and evening) during irrigation. The framework was evaluated for IIS unit installation numbers ranging from n = 1 to 10.

Methodology and Data

• Models used:
  • Two-stage optimization framework (Cluster-First, Route-Second strategy)
  • Density-based Spatial Clustering of Applications with Noise (DBSCAN)
  • Normalized Nearest-distance (NN-distance) method for adaptive DBSCAN parameter selection (MinPts fixed at 2)
  • 2-opt algorithm for Traveling Salesman Problem (TSP) approximation and patrol route optimization
  • Nearest Neighbor Index (NNI) for evaluating spatial aggregation characteristics
• Data sources:
  • Paddy field data collected from three rice farmers in Chiba Prefecture, Japan.
  • Representative coordinates for each paddy field derived from the geometric centroid of boundary points on satellite imagery.
  • Inter-field distance matrix constructed using actual walking distances obtained via the Google Maps route search function.

Main Results

• The proposed framework reduced the candidate solution space by six to nine orders of magnitude compared to exhaustive search, effectively mitigating the computational complexity of the NP-hard problem.
• The NN-distance method for DBSCAN parameterization maintained solution quality, matching globally optimal labor reduction where exhaustive search was feasible, and outperformed the conventional k-distance approach by preventing over-clustering and over-segmentation in non-uniform spatial distributions.
• Sensitivity analysis confirmed the local stability of the NN-distance method's Eps parameter to ±5% perturbations and validated MinPts=2 for optimal computational load and solution quality.
• Case analyses demonstrated that labor reduction benefits from IIS unit installation are strongly influenced by the spatial distribution of fields (e.g., clustered, random, uniform), not solely by the number of installed units.
• The derived Eps values (approximately 200 m) were found to reflect the intrinsic spatial scale of Japanese paddy field layouts.
• Labor Reduction Rate (LRR) increased with the number of IIS units, but the growth patterns varied significantly among farmers based on their field distributions. Average Labor Reduction Rate per unit (ALRR) generally declined with increasing installations but showed rebounds for clustered distributions, indicating non-linear benefits.

Contributions

• Proposed an NN-distance method that stabilizes DBSCAN performance under highly non-uniform agricultural spatial distributions, addressing the sensitivity issues of density-based clustering.
• Developed a cluster-informed combinatorial reduction strategy that mitigates the exponential growth of installation combinations while preserving feasible solutions.
• Established an optimization framework that integrates adaptive clustering with deterministic route evaluation, enabling exhaustive and reproducible assessment of IIS installation benefits in fragmented paddy fields.

Funding

• JST SPRING, Grant Number JPMJSP2108.

Citation

@article{Tian2025optimization,
  author = {Tian, Runze and Takaki, Kyoji and Iida, Toshiaki and Noda, Keigo and Asada, Yohei},
  title = {An optimization framework for intelligent irrigation system installation in fragmented paddy fields},
  journal = {Smart Agricultural Technology},
  year = {2025},
  doi = {10.1016/j.atech.2025.101740},
  url = {https://doi.org/10.1016/j.atech.2025.101740}
}