PS 86-172
Using species distribution modeling to understand the spread of the naturalized Honduran pine in Puerto Rico

Friday, August 9, 2013
Exhibit Hall B, Minneapolis Convention Center
Wilnelia Recart, Ecology and Evolutionary Biology, University of California - Irvine, Irvine, CA
James D. Ackerman, Department of Biology, University of Puerto Rico, Rio Piedras Campus, San Juan, PR
Wilfredo Falcón, Institute of Evolutionary Biology and Environmental Studies, University of Zürich, Zürich, Switzerland
Julio Lazcano, Department of Biology, University of Puerto Rico, Rio Piedras Campus, San Juan, PR

In the mid 1960s Pinus caribaea var. hondurensis (Honduran pine) was introduced across Puerto Rico to reforest abandoned agricultural lands and facilitate a forest products industry. The latter goal never materialized, leaving most pine plantations abandoned and unmanaged. Currently, the pine populations show heterogeneous recruitment varying from high to none. Why are there differences in recruitment amongst sites? What factors are shaping the expansion or containment of the Honduran pine? We hypothesize that the availability of mycorrhizal fungi shapes the local distribution and environmental factors such as elevation, temperature and precipitation shape the regional distribution. We expected to shed light on the possible causes of non-recruitment using species distribution modeling. For this we identified populations of Honduran pine and categorized them as recruitment or non-recruitment sites. We then modeled the predicted distribution of the species using the maximum entropy modeling algorithm software MaxEnt, a sample set of localities with pine recruitment and 5 layers of abiotic conditions. We ran 10 replicates with bootstrapping using 75% recruitment records for training and 25% for testing in each run. We used the area under the curve (AUC) as a parameter for the quality of the model.


We obtained 84 localities with either recruitment (n=51) or non-recruitment (n=33) of the Honduran pine. We used 20 localities with recruitment to model the probability of occurrence. Our AUC results for testing and training were > 0.90. The bioclimatic layer of “maximum temperature during the warmest month” was the one that contributed the most in the model. We found that the probabilities of occurrence at non-recruitment sites were significantly different from those at recruitment sites. We found that 90% of the sites with recruitment were within suitable areas. Approximately 50% of the non-recruitment localities are in areas of low suitability and the other half are in areas of high suitability. From these results we conclude that pine populations in Puerto Rico are experiencing differences in recruitment through the mediation of two different factors. The first factor is occurring in a suitable climatic area and the second is overlapping with suitable biotic conditions (e.g. mycorrhizal associations). Our predicted distribution shows that most of pine populations will continue to expand in the mountainous interior of the island, thus increasing the probability of the pines generating novel interactions with the flora and fauna of this region.