Sensibilité des modèles de distribution d'espèces à l'algorithme choisi : cas de Lophira lanceolata et implications pour sa conservation au Burkina Faso
Abstract
Lophira lanceolata is a native, multipurpose tree species that supports ecosystem functioning and remains a key resource for rural communities across its range. Given the adverse effects of climate change on natural habitats, this study primarily aims to analyse divergences among five ecological niche modelling algorithms and their implications for assessing suitable habitats for Lophira lanceolata in Burkina Faso. Current and future potential distribution projections for the 2050 and 2070 horizons are used as a basis for comparing the sensitivity of the models under different climate scenarios. Thus, five modelling methods were employed: Generalized Linear Model, Generalized Additive Model, Random Forest, Maximum Entropy, and Boosted Regression Trees. Occurrence records came from the Global Biodiversity Information Facility and Rainbio, filtered for quality and paired with environmental predictors chosen to minimize collinearity. Overall model performance was satisfactory (Area Under the Curve, True Skill Statistics, and kappa), but the algorithms differed substantially in their predictions of where the species could persist. Random Forest was the most conservative, restricting suitable habitat to a narrow range, while Boosted Regression Trees mapped the broadest extent of favorable conditions. Annual precipitation stood out as the dominant driver. Inside protected areas, GLM and GAM predicted more extensive suitable habitat than either Maxent or Random Forest. These differences became even more pronounced under future climate scenarios. Random Forest projected a steep loss of highly suitable habitat; Maxent proved highly sensitive to the choice of climate scenario; Boosted Regression Trees, by contrast, held relatively steady. GLM and GAM went the other direction entirely, pointing to an expansion of suitable area, most notably under MIROC6. These divergences, sometimes substantial depending on the algorithms compared, highlight the limitations of an uncritical use of ecological niche modelling tools and call for a cautious interpretation of their outputs when defining conservation priorities for Lophira lanceolata under climate change.
Related
- Molecular Screening for Potential Resistance to Toxic Invasive Prey: Implications for Island Conservation
- Impact de la mise à jour de l’algorithme du logo nutritionnel Nutri-Score sur la classification des produits de la catégorie des pains de mie
- Prediction of Potentially Suitable Distribution Areas and Conservation Status of Sponges
- Effects of Shading-Induced Senescence and Retention of Lower-Canopy Needles on Photosynthetic Dynamics in Upper-Canopy Needles of Cunninghamia lanceolata
- Mongolian medicinal plants in China: Distribution, conservation priorities, and phylogenetic insights
BibTeX
@article{tietiambou2026sensibilité,
title = {{Sensibilité des modèles de distribution d'espèces à l'algorithme choisi : cas de Lophira lanceolata et implications pour sa conservation au Burkina Faso}},
author = {Fanta Reine Sheirita Tiétiambou and Rodrigue Idohou and Fabrice Jean Eudes Crédo Gahounga and Benjamin Lankoandé and Abdoulaye Tyano and Amadé Ouédraogo},
journal = {Physio-Géo},
year = {2026},
month = oct,
doi = {10.4000/16xa8},
url = {https://doi.org/10.4000/16xa8},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Sensibilité des modèles de distribution d'espèces à l'algorithme choisi : cas de Lophira lanceolata et implications pour sa conservation au Burkina Faso
AU - Fanta Reine Sheirita Tiétiambou
AU - Rodrigue Idohou
AU - Fabrice Jean Eudes Crédo Gahounga
AU - Benjamin Lankoandé
AU - Abdoulaye Tyano
AU - Amadé Ouédraogo
PY - 2026
DA - 2026/10/01
JO - Physio-Géo
DO - 10.4000/16xa8
UR - https://doi.org/10.4000/16xa8
AB - Lophira lanceolata is a native, multipurpose tree species that supports ecosystem functioning and remains a key resource for rural communities across its range. Given the adverse effects of climate change on natural habitats, this study primarily aims to analyse divergences among five ecological niche modelling algorithms and their implications for assessing suitable habitats for Lophira lanceolata in Burkina Faso. Current and future potential distribution projections for the 2050 and 2070 horizons are used as a basis for comparing the sensitivity of the models under different climate scenarios. Thus, five modelling methods were employed: Generalized Linear Model, Generalized Additive Model, Random Forest, Maximum Entropy, and Boosted Regression Trees. Occurrence records came from the Global Biodiversity Information Facility and Rainbio, filtered for quality and paired with environmental predictors chosen to minimize collinearity. Overall model performance was satisfactory (Area Under the Curve, True Skill Statistics, and kappa), but the algorithms differed substantially in their predictions of where the species could persist. Random Forest was the most conservative, restricting suitable habitat to a narrow range, while Boosted Regression Trees mapped the broadest extent of favorable conditions. Annual precipitation stood out as the dominant driver. Inside protected areas, GLM and GAM predicted more extensive suitable habitat than either Maxent or Random Forest. These differences became even more pronounced under future climate scenarios. Random Forest projected a steep loss of highly suitable habitat; Maxent proved highly sensitive to the choice of climate scenario; Boosted Regression Trees, by contrast, held relatively steady. GLM and GAM went the other direction entirely, pointing to an expansion of suitable area, most notably under MIROC6. These divergences, sometimes substantial depending on the algorithms compared, highlight the limitations of an uncritical use of ecological niche modelling tools and call for a cautious interpretation of their outputs when defining conservation priorities for Lophira lanceolata under climate change.
SN - 1958-573X
ER -