Xue et al. (2026) An Interpretable BO-TCBDA Deep Learning Framework for Winter Wheat Yield Estimation Using Multi-Source Remote Sensing Data

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-09-08
• Authors: Anqi Xue, Shufang Tian, Tingyan Fu
• DOI: 10.3390/rs18183061

Research Groups

• Key Laboratory of Arable Land Quality and Fertilizer Efficiency in China (Ministry of Agriculture and Rural Affairs)
• Institute of Soil Science, Chinese Academy of Sciences
• College of Resources and Environmental Sciences, Nanjing Agricultural University

Short Summary

This study introduces a Bayesian Optimization–Temporal Convolutional Network–Bidirectional Long Short-Term Memory–Dual Attention (BO-TCBDA) deep learning framework for accurate and interpretable winter wheat yield estimation. The proposed model achieved the best performance with an R2 of 0.823 and an RMSE of 561.26 kg/ha.

Objective

• Investigate the effectiveness of a novel BO-TCBDA deep learning framework in predicting county-level winter wheat yields using multi-source features.

Study Configuration

• Spatial Scale: County-level (Henan Province, China)
• Temporal Scale: Annual data from 2013 to 2022

Methodology and Data

• Models used: BO-TCBDA, Temporal Convolutional Network (TCN), Bidirectional Long Short-Term Memory (BLSTM), Dual Attention Mechanism (DAM), Bayesian Optimization (BO)
• Data sources: Enhanced Vegetation Index (EVI), Leaf Area Index (LAI), Solar-Induced Chlorophyll Fluorescence (SIF), climate data

Main Results

• BO-TCBDA achieved the best performance with an R2 of 0.823 and an RMSE of 561.26 kg/ha.
• SIF improved predictive performance in all models, with statistically significant gains observed in deep learning models.
• Dual-attention mechanism provided interpretable insights by highlighting relatively balanced contributions among input features and grain-filling stage through temporal attention.

Contributions

• Original value lies in the development of a novel BO-TCBDA framework that combines Bayesian Optimization with Temporal Convolutional Network, Bidirectional Long Short-Term Memory, and Dual Attention Mechanism for accurate and interpretable winter wheat yield estimation.
• The study demonstrates the effectiveness of SIF in improving predictive performance of deep learning models.

Funding

• This research was funded by the National Key Research and Development Program of China (Grant No. 2021YFA1402300) and the Natural Science Foundation of Jiangsu Province (Grant No. BK20190312).

Citation

@article{Xue2026Interpretable,
  author = {Xue, Anqi and Tian, Shufang and Fu, Tingyan},
  title = {An Interpretable BO-TCBDA Deep Learning Framework for Winter Wheat Yield Estimation Using Multi-Source Remote Sensing Data},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18183061},
  url = {https://doi.org/10.3390/rs18183061}
}