Material and Methods
Population Genetic Analyses
Sampling of needle and bud tissue was conducted across the range of both whitebark pine and western white pine. For whitebark pine, 28 sites were sampled and were combined into 6 regions (n=21 to 72, Richardson and others 2002). Western white pine was sampled from 15 sites (n= ~30) across the range. Plant tissue was used to extract genomic DNA using previously published protocols (Kim and others 2003, Richardson and others 2002).
For whitebark pine, two types of DNA markers were used: mitochondrial (mt)DNA and chloroplast microsatellites (cpSSR) haplotypes. These markers are uniparentally inherited enabling separate estimates of gene flow via pollen by cpSSR and via seed dispersal by mtDNA. Protocols for DNA typing and analyses are described by Richardson and others (2002). For western white pine, an amplified fragment length polymorphism (AFLP) approach was used to provide high-resolution markers that are typically biparentally inherited, (i.e., nuclear markers). Western white pine AFLP protocols and analyses are described in Kim and others (2003).
Plant-Climate Modeling
Whitebark pine and western white pine bioclimatic models used species presence/absence data for more than 118,000 point locations described geographically by latitude, longitude, and elevation. Climate was estimated for each of these locations using a spline climate model (Rehfeldt 2006). This procedure yielded a data set, which described species presence or absence based on 33 climate variables representing simple interactions of temperature or precipitation, or both, in relation to time, all of which have some relationship to plant responses. Random forests multiple classification tree analysis (Breiman 2001) was then used to identify climate variables important for predicting species presence and to build decision tree-based bioclimatic models.
Arc map software (ESRI Inc.) was used to project the models predictions on a map. For this approach, the spline climate model was used to estimate climate in each 1 sq km grid across the Western United States (more than 5.9 million pixels). Each pixel was then assessed using 100 independent decision trees. For each decision tree, a vote for or against species presence was cast. Pixels receiving vote tallies of 50 percent or greater were defined as having suitable climate space. These pixels are represented by the colored areas in Figures 1 and 2. These procedures are described in depth elsewhere (Rehfeldt and others 2006).
The response of a species contemporary suitable climate space under climate change was then assessed. The spline climate model was updated to the predicted climate of beginning in ca. 2030 using an average from the HadCM3GGa1 of the Hadley Centre (Gordon and others 2000) and CGCM2 of the Canadian Centre for Modeling and Analysis (Flato and Boer 2001) general circulation models. The procedure described in the previous paragraph was then repeated using the updated climate model to predict future climate for each pixel space.
Encyclopedia ID: p3567
