Using species distribution models to quantify climate change impacts on the rosy-finch superspecies: an alpine obligate
Anthropogenic climate change is forcing plants and animals to respond by shifting their distributions poleward or upward in elevation. An organism's ability to track climate change is constrained if its habitat cannot shift and is projected to decrease in geographic extent. Geographic distribut...
| Main Author: | |
|---|---|
| Format: | Text |
| Language: | English |
| Published: |
University of Utah
2016
|
| Subjects: | |
| Online Access: | https://dx.doi.org/10.26053/0h-b7n2-ymg0 https://collections.lib.utah.edu/ark:/87278/s6m07dq4 |
| Summary: | Anthropogenic climate change is forcing plants and animals to respond by shifting their distributions poleward or upward in elevation. An organism's ability to track climate change is constrained if its habitat cannot shift and is projected to decrease in geographic extent. Geographic distributions were predicted for the Rosy-Finch superspecies (Leucosticte atrata, Leucosticte australis, Leucosticte tephrocotis) for climate change scenarios using correlative species distribution models (SDM). Multiple sources of uncertainty were quantified including choice of SDM, validation statistic, emissions scenario and general circulation model (GCM). Species distribution models are traditionally validated using a threshold-independent statistic called the area under the curve (AUC) of the receiver operating characteristic. However, during k-fold cross-validation, this statistic becomes overinflated due to spatial sorting bias existing between the presence and background points. Therefore, a calibrated area under the curve (cAUC) that removes spatial sorting bias, was used to assess predictive accuracy. Boosted regression trees (BRT) and random forest regression trees (RF) had significantly higher predictive skill than did maximum entropy (MaxEnt) or generalized additive models (GAM; p=0.05). Predictions in space and time were made using the tree-based algorithms and applied to two representative concentration pathway (RCP) emissions scenarios, RCP2.6 and RCP8.5, for 15 and 17 GCMs, respectively, for 2061-2080. Despite variability in predictions, there is agreement between the projections for many regions suggesting that Rosy-Finches are extremely vulnerable to climate change. The short-term management implications of this study are the need for an immediate assessment of Rosy-Finches by surveying suitable habitats predicted by the species distribution models to enable population estimates to be made and field validation of species distribution model outputs. Such an approach would revisit historical breeding sites lacking coverage in the eBird and Global Biodiversity Information Facility (GBIF) datasets to determine if some populations have already been extirpated due to climate warming. Long-term management implications are to consider the use of large herbivores in alpine and tundra ecosystems in order to mitigate ecosystem response to climate warming and preserve foraging habitat for Rosy-Finches during the breeding season. |
|---|