Forest Ecology and Conservation Group

Changing the Landscape of Conservation

Species Distribution & Modeling

This cluster explores advances in species distribution modeling (SDMs), focusing particularly on MaxEnt methods, spatial bias correction, and applications in biodiversity assessment and conservation practice. The works demonstrate how modeling can reveal patterns in species richness across elevational gradients, inform IUCN Red List extinction risk evaluations, and highlight the importance of sampling design and feature complexity in predictive performance. Collectively, these studies help refine SDM methodologies and enhance their utility in ecological research and global biodiversity policies.

Featured Publications

Inferring diversity patterns along an elevation gradient from stacked SDMs: A case study on Mesoamerican ferns
Syfert, Mindy M.; Brummitt, Neil A.; Coomes, David A.; Bystriakova, Nadia; Smith, Matthew J. – Global Ecology and Conservation (2018)
DOI: 10.1016/j.gecco.2018.e00433
Using stacked SDMs for 86 Mesoamerican ferns, this study reveals a hump-backed relationship between species richness and elevation (peaking around 1800–2000 m), validated against field survey data.

Using species distribution models to inform IUCN Red List assessments
Syfert, Mindy M.; Joppa, Lucas; Smith, Matthew J.; Coomes, David A.; Bachman, Steven P.; Brummitt, Neil A. – Biological Conservation (2014)
DOI: 10.1016/j.biocon.2014.06.012
Compares extent-of-occurrence (EOO) estimates based on specimen records versus SDM-derived range maps for Central American plant species, showing SDMs can provide more representative range coverage when sample sizes are small.

The Effects of Sampling Bias and Model Complexity on the Predictive Performance of MaxEnt Species Distribution Models
Syfert, Mindy M.; Smith, Matthew J.; Coomes, David A. – PLOS ONE (2013)
DOI: 10.1371/journal.pone.0055158
Using New Zealand tree fern data, this work assesses both spatial sampling bias correction and MaxEnt parameter complexity, demonstrating that bias correction significantly improves fit while model complexity has limited additional benefit.