Species distribution modeling in a rare endemic herb: Multiple models suggest that dispersal limitation and past historical processes limit range expansion

Background/Question/Methods

As anthropomorphic climate change continues, many species ranges will shift (polewards or upslope) tracking ideal climate conditions. In some species, however, lack of genetic variation and/or dispersal barriers may prevent these range shifts from occurring, likely resulting in species extinctions. Species with small population sizes and restricted ranges are especially vulnerable. A first step in identifying what limits a species range is to determine if all suitable habitat (i.e. the entire niche space) is occupied. Unoccupied niche space may indicate dispersal limitation. Many European species may have had their ranges constricted by past glaciation, and are currently unable to disperse to fill their entire contemporary suitable habitats. I am interested if the small-ranged species endemic to the Southern United States show the same pattern. I focus on Phacelia fimbriata, Diphylleia cymosa, and Trillium vaseyi, all of which have ranges restricted to the cool mid/high elevations of the southern Appalachian Mountains. Here, I generate species distribution models (SDMs) using MaxEnt, MaxLike, boosted regression trees (BRT) and generalized linear models (GLM) to test if these species fill their entire contemporary niche space, and to conduct a paleo-niche analysis. Further, I compare and contrast the performance of these four SDM approaches.

Results/Conclusions

All four SDM methods predict suitable niche space for P. fimbriata in geographic areas much further north of its current range. This includes areas in the northern Appalachian Mountains of New York and Canada. A large area of highly suitable niche space exists in the central Appalachian Mountains; however, it is separated from the current range by an area of highly unsuitable niche space. Areas of highly suitable niche space in more northern areas are small and patchy, and are surrounded by areas of highly unsuitable niche space. Significant predictor variables include elevation, precipitation in the wettest month, precipitation in the driest month, and maximum temperature in the coldest month. These results suggest dispersal limitation in P. fimbriata. The SDM approaches MaxEnt and boosted regression trees (BRT) significantly outperformed generalized linear models (GLM). Preliminary results for D. cymosa and T. vaseyi are inconclusive.