Collection sites and common gardens
To establish the common gardens, 16 populations of Populus fremontii were collected throughout Arizona, encompassing the environmental variation experienced by the Sonoran Desert ecotype, as well as three populations located on the Mogollon Rim within the Colorado Plateau region of northern Arizona (Fig. 2). These populations group genetically with and have been alternatively identified as the Mogollon Rim (Blasini et al. 2020) or Utah High Plateau ecotype (Ikeda et al. 2017; Supplemental Table 1). This sampling design does not include the third described ecotype of the Central California Valley (Ikeda et al. 2017). Cuttings were taken from individual tree genotypes located over 20 m away from each other to ensure independent genotype sampling. Clonal replicates from 12 trees per population were planted in the summer and fall of 2014 in each of the three common garden sites after rooting in the greenhouse for approximately four months.
The three replicated experimental common gardens span broad elevation and climatic gradients, resulting in extreme climatic transfers for some populations. The northernmost garden represents the cold edge of the species’ climatic range. It is located adjacent to Canyonlands National Park, Utah and is maintained by The Nature Conservancy’s Dugout Ranch. The middle Arizona garden is located adjacent to the Agua Fria River in Agua Fria National Monument and is maintained by the Arizona Game and Fish Department. The southernmost garden is in Yuma, Arizona near Mittry Lake, and is maintained by the Bureau of Land Management. These gardens span over a 1500 m elevation difference, a 12°C mean annual temperature range (10.7°C in Yuma, 17.2°C in Agua Fria, and 22.8°C in Canyonlands), and a precipitation difference of ~350mm (Supplemental Table 1). Each common garden was planted with 4,096 trees. These trees were arranged into four replicated blocks to account for within-garden environmental variance, with each block made up of 16 randomized population-level plots. Each population plot had 64 trees, made up of three to six replicates of the 12 genotypes collected for that population. Plots were arranged in a randomized 8 x 8 grid, with trees spaced 1.85m in each cardinal direction. The garden was designed using population plots instead of fully randomized by genotype to assess population-level effects on dependent community members such as arthropods and mycorrhizae, as well as ecosystem-level traits like carbon flux.
In order to examine the relationship between climate and traits, we downloaded 30-year normals (1961-1990 means) for 21 abiotic climate variables for each of the 16 provenance sites and the three common gardens using the program ClimateWNA (Wang et al. 2012). Because variation in both temperature and precipitation in the Southwest are very strongly correlated with elevation, these current climate variables are excellent proxies for the climates that trees have experienced during their local evolutionary histories (r > 0.985 for correlations between current MAT and MAP (WorldClim 2, Fick & Hijmans 2017) and those variables estimated from 6,000 or 22,000 years ago (WorldClim 1.4, Hijmans et al. 2005). To create a multivariate climatic index representing the environmental variation found throughout the 16 provenances, the ClimateWNA variables plus elevation, latitude, and longitude, were combined in a principal component analysis (PCA) using labdsv (Roberts 2007) and vegan (Oksanen et al. 2016) packages in the R statistical language (R Core Team 2014).