Abstract Detail

Conservation Biology

Schmidt, Thomas J [1], Greer, Gary K. [2].

A Maxent approach to model pteridophyte community reestablishment in secondary forests of Puerto Rico.

Approximately 94% of Puerto Rico’s forests were converted into agricultural systems by 1950. Since then, however, there has been extensive abandonment of agricultural land due to an economic shift towards industry and services, initiating an increase in forest regeneration and thus a substantial amount of secondary forest throughout the main island. Pteridophytes are a major herbaceous component of oceanic, tropical island forests; consequently, the composition and community structure of ferns are indicative of the relative richness of the revegetated landscapes of Puerto Rico. Predictive models of species distributions using geographic information systems (GIS) and statistical algorithms are becoming increasingly popular due to their accuracy and have great value in conservation prioritization. We used Maximum Entropy (Maxent), a widely-used mathematical tool for distinguishing suitable versus unsuitable niche space, to model thirty-six tropical fern distributions (including terrestrial and epiphytic species, some endemic or threatened, representing 13 pteridophyte families). Ten environmental variables in four categories including climate, topography, geologic substrate, and vegetation indices (land cover transformations indicating vegetation denseness, water retained by soil and vegetation, and exposed soil) were selected in conjunction with documented species occurrences collected from three herbaria for model construction. Model discrimination was assessed via area under the receiver operating characteristic curve (AUC) for each species. Sample size corrected Akaike information criteria were obtained for model variations ranging in complexity using ENMTools and used to identify the most parsimoniously constructed version. Parsimony significantly increased model discrimination relative to models ran with default settings. Predicted distributions were used to create threshold maps, a binary output indicating suitable versus unsuitable habitat for each species, and overlaid to depict areas of predicted high species richness. To compare model predictions with actual species richness, twenty-two, 20 m × 20 m secondary forest plots, ranging in age since agricultural abandonment from 13 – 76 years, were sampled throughout the main island of Puerto Rico. All model discrimination values were relatively high (0.751 – 0.956) indicating good model performance for all species. Actual fern species richness correlated positively with predicted fern species richness demonstrating great utility of predictive modeling with Maxent to identify areas of high conservation prioritization in Puerto Rico’s recovering mid- to high-elevation forests.

Broader Impacts:

1 - Grand Valley State University, Biology, 1 Campus Drive, Allendale, MI, 49401, USA
2 - Grand Valley State University, Biology, 1 Campus Drive, Allendale, MI, 49401-1000, USA

Keywords:
pteridophyte
conservation
Puerto Rico
secondary forest
Maxent
species distribution modeling.

Presentation Type: Oral Paper:Papers for Topics
Session: 20
Location: Marlborough A/Riverside Hilton
Date: Tuesday, July 30th, 2013
Time: 9:30 AM
Number: 20005
Abstract ID:693
Candidate for Awards:None