Islam et al. (2026) Accuracy is not certainty: using model agreement and human judgment to assess spatial uncertainty in high-resolution mangrove mapping

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

Identification

• Journal: Environmental Research Letters
• Year: 2026
• Date: 2026-07-06
• Authors: K. M. Ashraful Islam, Paulo J. Murillo‐Sandoval, Robert E. Kennedy
• DOI: 10.1088/1748-9326/ae867e

Research Groups

• Department of Earth and Environmental Sciences, University of California, Los Angeles (UCLA)
• Bangladesh Forest Department
• International Centre for Tropical Agriculture (CIAT)

Short Summary

This study examines the representation of spatial uncertainty in probabilistic ensemble models for mangrove classification using high-resolution imagery of the Sundarbans in Bangladesh. The results show that stacked probabilities can serve as a proxy for epistemic uncertainty, highlighting areas where predictions are stable or contested.

Objective

• Investigate how probabilistic ensemble models represent spatial uncertainty in mangrove classification

Study Configuration

• Spatial Scale: Regional scale (Sundarbans in Bangladesh)
• Temporal Scale: Not specified

Methodology and Data

• Models used: Stacked generalization with multiple base learners
• Data sources: High-resolution imagery, human interpretation

Main Results

• Individual and stacking models achieved similarly high accuracy.
• Stacked probabilities served as a proxy for epistemic uncertainty in the map.
• Disagreement among heterogeneous base learners reflected where predictions are stable or contested.
• Extreme values corresponded to strong model consensus and higher interpreter confidence.
• Intermediate values coincided with greater variability in both model predictions and human judgment.

Contributions

• This study provides new insights into the representation of spatial uncertainty in probabilistic ensemble models for mangrove classification.
• The results highlight the importance of quantifying epistemic uncertainty in environmental monitoring applications.

Funding

• Not specified

Citation

@article{Islam2026Accuracy,
  author = {Islam, K. M. Ashraful and Kilbride, John B and Murillo‐Sandoval, Paulo J. and Kennedy, Robert E.},
  title = {Accuracy is not certainty: using model agreement and human judgment to assess spatial uncertainty in high-resolution mangrove mapping},
  journal = {Environmental Research Letters},
  year = {2026},
  doi = {10.1088/1748-9326/ae867e},
  url = {https://doi.org/10.1088/1748-9326/ae867e}
}