20th Annual Student Research Conference
Interdisciplinary
Habitat suitability modeling of a federally threatened plant species, Missouri bladderpod (Lesquerella filiformis): Evaluation of model selection criteria and model performance.
James D. Franklin* and Corey A. Elledge
Dr. Hyun-Joo Kim and Dr. Michael I. Kelrick, Faculty Mentors
Habitat suitability modeling (HSM) using logistic regression is a valuable conservation biology tool to predict or explain species’ responses to habitat attributes. We used HSM to investigate habitat preferences of Missouri bladderpod (Lesquerella filiformis), a federally threatened plant species. Missouri bladderpod abundance and microhabitat attribute cover have been monitored at Wilson’s Creek National Battlefield (Republic, MO) intermittently since 1997. This study examined various model selection procedures, initially using proportion of cells predicted correctly as a measure of model performance. Two methods – model averaging and single model selection – were evaluated, for four candidate model selection Information Criteria. No significant differences were observed in signed-rank tests of comparisons of combinations of these methodologies and criteria. So, despite their broad theoretical appeal, these approaches have not clarified which HSM might be best. Consequently, we are pursuing alternative approaches to model evaluation, involving “out-of-bag” simulations, as well as assessment of prediction error types.
Keywords: plant ecology, statistics, MBI, modeling
Topic(s):Interdisciplinary
Presentation Type: Oral Paper
Session: 54-4
Location: OP 2210
Time: 3:30 pm