Forecasting climate induced distribution shifts for the pinions and junipers of the Western US

Background/Question/Methods           Pinion and juniper vegetation types are estimated to cover 15% of the five U.S. interior western states of Utah, Nevada, Arizona, New Mexico, and Colorado. Six species of junipers and three species of pinions compose these vegetation types in a mosaic of overlapping distributions arranged primarily across a gradient of precipitation seasonality. Paleoecological studies continue to illuminate distribution dynamics over the last 10,000+ years. Individual species have been found to respond uniquely to climate changes, causing shifts in the co-occurrence of species. This is reflected in the differential mortality among species caused by recent droughts. Widespread mortality events like the recent one apparently happen every few centuries, causing distributions to contract to "core areas" where woodlands may persist for millennia, and from which they expand when conditions are favorable. Our work focuses on three objectives, first to characterize the bioclimatic conditions of "core areas" with distribution models for each species, second to apply climate change scenarios to the distribution models, and third to gauge shifts in the co-occurrence of species. The utility of three contemporary classification techniques was compared, including Classification Trees, Random Forests, and MaxEnt. Distribution models were developed using forest inventory data.

Results/Conclusions           The application of climate change scenarios to the distribution models resulted in the detection of areas becoming suitable to colonizing populations, as well as areas no longer suitable for the regeneration of existing populations. Model utility varied, with Random Forests and MaxEnt producing the best maps, but with CART generally producing the most ecologically interpretable models. All models indicated climate change scenarios will alter the co-occurrence of pinions and junipers. Shifts in co-occurrence varied from decreasing for all scenarios to increasing for all scenarios with some increasing or decreasing depending on scenario. Differences in the projected co-occurrence of species appear to primarily be a result of differences in projected precipitation seasonality among the scenarios.