Wen et al. (2026) Rule-driven functional zoning index prediction for sustainable agricultural landscapes design based on multimodal fusion deep learning

Identification

• Journal: Scientific Reports
• Year: 2026
• Date: 2026-09-18
• Authors: Ruifen Wen, Juan Du
• DOI: 10.1038/s41598-026-72129-2

Research Groups

• Department of Computer Science, University of California, Fresno
• Department of Agricultural and Biological Engineering, University of California, Davis
• NASA Ames Research Center, Mountain View, CA

Short Summary

This study proposes a multimodal fusion deep learning model (MMLF-Net) to predict functional zoning indices for sustainable agricultural landscapes. The model integrates multi-source variables, including remote sensing spectral information, landscape structure, topographic conditions, crop configuration, and land cover.

Objective

• Develop an end-to-end spatial modeling framework capable of effectively learning the mapping relationship between multimodal spatial data and a predefined functional zoning index system.

Study Configuration

• Spatial Scale: Local to regional scale (Fresno region, California, USA)
• Temporal Scale: Continuous growing season from April to September 2021

Methodology and Data

• Models used: MMLF-Net, a multimodal fusion deep learning model combining Vision Transformer (ViT) and Convolutional Neural Network (CNN)
• Data sources:
  • Sentinel-2 Level-2 A remote sensing images
  • European Space Agency (ESA) WorldCover 10 m land cover product
  • Shuttle Radar Topography Mission (SRTM) DEM
  • USDA Cropland Data Layer (CDL)

Main Results

• MMLF-Net achieves an overall accuracy of 87.34% and a Kappa coefficient of 0.84 for functional identification and zoning.
• The model outperforms existing methods in terms of accuracy and stability.

Contributions

• This study provides a practical spatial decision-making tool for the sustainable development of regional agriculture.
• MMLF-Net offers a more stable and scalable technical pathway for SAL functional zoning compared to traditional zoning methods based on rules or multi-index weighting.

Funding

• This research was funded by the NASA Ames Research Center (Grant Number: NNX17AC54A) and the University of California, Fresno (Internal Grant Number: UCFFR-2020-001).

Citation

@article{Wen2026Ruledriven,
  author = {Wen, Ruifen and Du, Juan},
  title = {Rule-driven functional zoning index prediction for sustainable agricultural landscapes design based on multimodal fusion deep learning},
  journal = {Scientific Reports},
  year = {2026},
  doi = {10.1038/s41598-026-72129-2},
  url = {https://doi.org/10.1038/s41598-026-72129-2}
}