Rosa et al. (2026) Remote Sensing–driven ensemble smoother assimilation of LAI for regional sugarcane yield estimation

Identification

• Journal: Remote Sensing Applications Society and Environment
• Year: 2026
• Date: 2026-01-01
• Authors: Juliano Mantellatto Rosa, Izael Martins Fattori Junior, Marina Luciana Abreu de Melo, Fábio Ricardo Marin
• DOI: 10.1016/j.rsase.2026.101952

Research Groups

• University of S˜ao Paulo, “Luiz de Queiroz” College of Agriculture (ESALQ-USP)
• Center for Carbon Research in Tropical Agriculture (CCARBON) - University of S˜ao Paulo

Short Summary

This study improved regional sugarcane yield estimation in S˜ao Paulo State, Brazil, by assimilating over 167,000 remotely sensed Leaf Area Index (LAI) observations into the DSSAT/SAMUCA model using an Ensemble Smoother. The approach significantly reduced Root Mean Square Error (RMSE) by 57% and Mean Absolute Error (MAE) by 72% across Technology Extrapolation Domains (TEDs).

Objective

• To improve regional sugarcane yield estimation by assimilating remotely sensed Leaf Area Index (LAI) data into a stochastic process-based crop model, addressing the underexplored application of comprehensive data assimilation frameworks at regional scales.

Study Configuration

• Spatial Scale: Regional scale, specifically sugarcane-growing areas in S˜ao Paulo State, Brazil, categorized by Technology Extrapolation Domains (TEDs).
• Temporal Scale: 2003–2013 period.

Methodology and Data

• Models used: DSSAT/SAMUCA (process-based crop model), Ensemble Smoother (ES) method (data assimilation).
• Data sources: Over 167,000 Leaf Area Index (LAI) observations derived from Landsat 7 ETM+ imagery.

Main Results

• Substantial improvements in yield estimation accuracy were achieved across all Technology Extrapolation Domains (TEDs).
• Root Mean Square Error (RMSE) was reduced from 43.98 megagrams per hectare (Mg ha⁻¹) to 17.83 Mg ha⁻¹, representing an average reduction of 57%.
• Mean Absolute Error (MAE) was reduced from 41.85 Mg ha⁻¹ to 11.65 Mg ha⁻¹, representing an average reduction of 72%.
• Some individual TEDs showed even larger improvements, with MAE reductions reaching up to 96%.

Contributions

• Developed and applied a comprehensive data assimilation framework at a regional scale for sugarcane yield estimation, integrating remotely sensed LAI into a process-based crop model.
• Demonstrated significant improvements in the accuracy of regional-scale sugarcane yield estimation, providing a robust method for retrospective assessment of crop performance.

Funding

• Not specified in the provided text.

Citation

@article{Rosa2026Remote,
  author = {Rosa, Juliano Mantellatto and Junior, Izael Martins Fattori and Melo, Marina Luciana Abreu de and Marin, Fábio Ricardo},
  title = {Remote Sensing–driven ensemble smoother assimilation of LAI for regional sugarcane yield estimation},
  journal = {Remote Sensing Applications Society and Environment},
  year = {2026},
  doi = {10.1016/j.rsase.2026.101952},
  url = {https://doi.org/10.1016/j.rsase.2026.101952}
}