COS 102-10
Conservation assessment for an endemic Malagasy rodent, Gymnuromys roberti (Family Nesomyidae), using ecological niche models and forest cover data

Thursday, August 13, 2015: 11:10 AM
322, Baltimore Convention Center
Maria Gavrutenko, Biology, City College of New York, City University of New York, New York, NY
Beth E. Gerstner, Biology, City College of New York, City University of New York, New York, NY
Steven M. Goodman, Association Vahatra, Antananarivo, Madagascar
Robert P. Anderson, Biology, City College of New York, City University of New York, New York, NY
Background/Question/Methods

Highland tropical biodiversity already threatened by deforestation may be particularly vulnerable to future climate change. This conservation assessment focused on Gymnuromys roberti, an endemic rodent species restricted to highland humid forests in Madagascar, currently considered “Least Concern” by IUCN. To reevaluate the species’ conservation status, we combined ecological niche models (ENMs) with MODIS forest cover data and current protected area maps to estimate abiotically areas with sufficient forest cover. We built Maxent ENMs using spatially filtered occurrence localities and climatic data, and evaluated them using a cross-validation approach. We selected a model of optimal complexity by minimizing average omission rate and subsequently maximizing average AUC. We built the final model using optimal settings and all filtered localities to identify areas of high suitability. We clipped this prediction using two deforestation tolerance thresholds (minimum and stricter) derived from forest cover values at recent occurrence localities, matched by year of collection. Maps of the national protected area system were used to assess how much of the species’ habitat is presently protected. We also projected this model to the year 2050 using several climatic estimates to assess if these protected forest areas are likely to remain climatically suitable in the near future.

Results/Conclusions

Optimal model settings (linear/quadratic features with regularization multiplier 1.0) resulted in a prediction indicating that the suitable area is naturally fragmented, with unsuitable conditions dividing the species’ range into several allopatric populations. A disjunct patch of suitable conditions without known occurrence records was identified as a target for future sampling efforts (near Mandritsara district in northwestern Madagascar). A substantial portion (24%) of the abiotically suitable area does not have the minimum forest cover necessary to sustain the species’ populations. Moreover, we inferred from recent occurrence records that the species more frequently occurs in areas with values of forest cover at or above the stricter deforestation threshold, which encompasses only 52% of the abiotically suitable area (32,000 km2). Overall, only 24% of the abiotically suitable area is forested above the strict threshold and is currently protected by national parks (15,000 km2).  Future model projections indicated that a significant portion of the currently available habitat will not remain climatically suitable in the future, with major contractions in the southern and northern parts of the range. In the final stage of our analyses, we will include these areal estimations into reassessing the species’ conservation status following IUCN criteria.