Ensemble frequency ratio-maximum entropy landslide susceptibility modelling for climate‑resilient road corridors in Bocas del Toro and Chiriquí, Panama
5 affiliations
- 1Smithsonian Tropical Research Institute, Bocas Research Station, Balboa Ancon, Panama
- 2Disaster Management Training and Education Center for Africa (DiMTEC), University of the Free State, Bloemfontein, South Africa
- 3School of Environmental Studies, University of British Columbia, Vancouver, British Columbia, Canada
- 4Centro Regional de Bocas del Toro, Universidad de Panama, Panamá
- 5Centro Regional de Bocas del Toro, Universidad Tecnologica de Panama, Panamá
Abstract
Landslides are a major rainfall-triggered hazard affecting road infrastructure across the mountainous terrain of western Panama. This study develops an ensemble Frequency Ratio (FR)-Maximum Entropy (MaxEnt) modelling framework to assess landslide susceptibility, identify critical management zones, and determine key drivers of slope failure along road corridors in the mainland districts of Bocas del Toro and along the Rambala-Hornito-David route. Within a 1,000 m corridor buffer, 87 mapped landslides (63 for training and 24 for validation) and 15 environmental conditioning factors were used to develop and validate the FR and MaxEnt models. An AUC-weighted linear combination of the independently calibrated FR and MaxEnt models was used to generate the ensemble susceptibility map. The ensemble delineated 9.67% and 3.99% of the corridor as high and very high susceptibility, respectively, achieving the highest predictive performance (AUC = 0.945) and outperforming the FR (0.873) and MaxEnt (0.943) models. Sensitivity analyses based on alternative training-validation partitions (57:30 and 70:17) produced consistently good performance (prediction-rate AUC > 0.88), indicating that the principal findings were not strongly influenced by the inventory allocation. Hotspot analysis identified three major clusters of elevated susceptibility and confirmed statistically significant spatial clustering at the 95% and 99% confidence levels. More than 71 km of primary roads intersect high-susceptibility zones, with Hornito, Bajo Cedro, and Valle de Agua Arriba exhibiting the greatest exposure. Slope gradient and NDVI were identified as the most influential predictors. These findings provide a practical basis for prioritising monitoring, slope stabilisation, drainage improvement, and ecosystem-based interventions to strengthen the resilience of road infrastructure in tropical mountain environments. Future work should incorporate larger landslide inventories, event-based rainfall information, and climate-change scenarios to improve the representation of landslide dynamics and support risk-informed infrastructure planning.
Related
- Landslide Susceptibility Modelling and Infrastructure Exposure in the Zillertal, Austrian Alps
- Machine-learning modelling of regional landslide susceptibility in eThekwini, South Africa
- Landslide Susceptibility Modeling and Mapping: A Comparison of Frequency Ratio and Decision Tree Models
- Role of Topo-Climatic Factors in Landslide Susceptibility Mapping: A Machine Learning Perspective
- Empirically informed AHP-FR modeling for multi-hazard susceptibility mapping in a mountainous transport corridors
BibTeX
@article{petiangma2026ensemble,
title = {{Ensemble frequency ratio-maximum entropy landslide susceptibility modelling for climate‑resilient road corridors in Bocas del Toro and Chiriquí, Panama}},
author = {Dereck Mbeh Petiangma and Gerald G. Singh and Enrique Williams and Jeffrey Gerardo Ramos Perez and Rachel Collin},
journal = {PLOS One},
year = {2026},
month = nov,
doi = {10.1371/journal.pone.0362084},
url = {https://doi.org/10.1371/journal.pone.0362084},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Ensemble frequency ratio-maximum entropy landslide susceptibility modelling for climate‑resilient road corridors in Bocas del Toro and Chiriquí, Panama
AU - Dereck Mbeh Petiangma
AU - Gerald G. Singh
AU - Enrique Williams
AU - Jeffrey Gerardo Ramos Perez
AU - Rachel Collin
PY - 2026
DA - 2026/11/05
JO - PLOS One
DO - 10.1371/journal.pone.0362084
UR - https://doi.org/10.1371/journal.pone.0362084
AB - Landslides are a major rainfall-triggered hazard affecting road infrastructure across the mountainous terrain of western Panama. This study develops an ensemble Frequency Ratio (FR)-Maximum Entropy (MaxEnt) modelling framework to assess landslide susceptibility, identify critical management zones, and determine key drivers of slope failure along road corridors in the mainland districts of Bocas del Toro and along the Rambala-Hornito-David route. Within a 1,000 m corridor buffer, 87 mapped landslides (63 for training and 24 for validation) and 15 environmental conditioning factors were used to develop and validate the FR and MaxEnt models. An AUC-weighted linear combination of the independently calibrated FR and MaxEnt models was used to generate the ensemble susceptibility map. The ensemble delineated 9.67% and 3.99% of the corridor as high and very high susceptibility, respectively, achieving the highest predictive performance (AUC = 0.945) and outperforming the FR (0.873) and MaxEnt (0.943) models. Sensitivity analyses based on alternative training-validation partitions (57:30 and 70:17) produced consistently good performance (prediction-rate AUC > 0.88), indicating that the principal findings were not strongly influenced by the inventory allocation. Hotspot analysis identified three major clusters of elevated susceptibility and confirmed statistically significant spatial clustering at the 95% and 99% confidence levels. More than 71 km of primary roads intersect high-susceptibility zones, with Hornito, Bajo Cedro, and Valle de Agua Arriba exhibiting the greatest exposure. Slope gradient and NDVI were identified as the most influential predictors. These findings provide a practical basis for prioritising monitoring, slope stabilisation, drainage improvement, and ecosystem-based interventions to strengthen the resilience of road infrastructure in tropical mountain environments. Future work should incorporate larger landslide inventories, event-based rainfall information, and climate-change scenarios to improve the representation of landslide dynamics and support risk-informed infrastructure planning.
SN - 1932-6203
ER -