Cheng et al. (2026) Improving regional winter wheat yield estimation by assimilating cross-scale heterogeneity information from UAV and satellite remote sensing

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-10-01
• Authors: Zhikai Cheng, Xiaobo Gu, Yang Xu, Xiaohai Fang, Zhengtao Zhang, Penglin Li, Yupeng Li, Shikun Sun, Yadan Du, Huanjie Cai
• DOI: 10.1016/j.rse.2026.115701

Research Groups

• Key Laboratory of Agricultural Soil and Water Engineering in Arid and Semiarid Areas, Ministry of Education, Northwest A&F University
• Center for Agricultural Water Research in China, China Agricultural University
• Shiyanghe Experimental Station for Improving Water Use Efficiency in Agriculture, Ministry of Agriculture and Rural Affairs, Ministry of Education and Gansu Government

Short Summary

This study developed a novel cross-scale data assimilation framework integrating UAV and satellite remote sensing with the WOFOST crop model to improve high-resolution winter wheat yield estimation by accounting for within-pixel heterogeneity. The framework achieved robust and transferable yield estimates, outperforming conventional assimilation methods.

Objective

• To develop a novel cross-scale data assimilation framework that integrates UAV and satellite remote sensing with a crop model to enhance high-resolution winter wheat yield estimation by incorporating within-pixel heterogeneity information.

Study Configuration

• Spatial Scale: Regional scale, with high-resolution yield estimates at 10 m × 10 m.
• Temporal Scale: Yield estimation performance evaluated for 2024–2025, with interannual stability tested from 2017–2023.

Methodology and Data

• Models used:
  • WOFOST crop model
  • Mixture Density Network (MDN) for upscaling and heterogeneity quantification
  • Random Forest for crop-soil parameter inversion
  • Improved Particle Filter for data assimilation
• Data sources:
  • Unmanned Aerial Vehicle (UAV) remote sensing (multispectral and thermal images)
  • Sentinel satellite data
  • In situ sampling data (used to link UAV data)
  • Crop-soil parameters: Leaf Area Index (LAI), Aboveground Biomass (AGB), Soil Moisture at 10 cm depth (SMC-10), Soil Moisture at 20 cm depth (SMC-20)

Main Results

• Robust crop-soil parameter inversion models were constructed using random forest with UAV images (R²: 0.73–0.88; NRMSE: 12.30%–22.83%).
• The Mixture Density Network (MDN) successfully quantified within-pixel heterogeneity of crop-soil parameters through a UAV-satellite upscaling approach.
• Joint assimilation of LAI, AGB, and SMC-10 achieved optimal winter wheat yield estimation performance for MDN-based cross-scale data assimilation (R² = 0.83, NRMSE = 15.70% in 2024–2025).
• The proposed framework outperformed satellite-based and mean-based cross-scale assimilation methods.
• The framework demonstrated stable interannual performance in Xianyang (RMSE ≤ 758 kg ha⁻¹, NRMSE ≤ 17.20% from 2017–2023) and reasonable transferability in two external regions (RMSE ≤ 943 kg ha⁻¹; NRMSE ≤ 18.29% in Baoding and Changji).

Contributions

• Developed a novel cross-scale data assimilation framework that effectively integrates UAV and satellite remote sensing to account for within-pixel heterogeneity in yield estimation.
• Introduced a Mixture Density Network (MDN) for upscaling UAV-derived crop-soil parameters to satellite pixel scale, providing probability distributions of heterogeneity.
• Demonstrated improved accuracy and transferability of high-resolution winter wheat yield estimates compared to conventional assimilation approaches that neglect within-pixel heterogeneity.

Funding

• [No specific funding information provided in the given paper text snippet.]

Citation

@article{Cheng2026Improving,
  author = {Cheng, Zhikai and Gu, Xiaobo and Xu, Yang and Fang, Xiaohai and Zhang, Zhengtao and Li, Penglin and Li, Yupeng and Sun, Shikun and Du, Yadan and Cai, Huanjie},
  title = {Improving regional winter wheat yield estimation by assimilating cross-scale heterogeneity information from UAV and satellite remote sensing},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115701},
  url = {https://doi.org/10.1016/j.rse.2026.115701}
}