Precipitation exclusion array
We conducted our study using an array of 25 paired control and precipitation exclusion sites that were established by the US Geological Survey across Southeastern Utah (central Colorado Plateau) in 2010 to understand the impacts of forecast increased aridity on Colorado Plateau drylands (Hoover et al . 2015, 2017; Winkler et al . 2019). Our set of 25 paired plots (with each pair representing one site) are a subset of the 40 sites in the USGS study, which we selected based on presence of well-developed biocrusts. The other 15 sites had insufficient biocrust cover to be suitable for our research (i.e., less than 15% cover of moss and lichen). The ca. 4500-km2 study region from which the experimental sites were selected spanned a range of plant communities, soil depths and parent materials, soil textures, elevations, climates and an aridity gradient (Table 1, Supporting Information). The climate in the ecoregion is characterized as a cold arid to semi-arid dryland with bimodal precipitation.
At each site, pairs of similar vegetation and soil patches were identified that represented a target plant community and treatments were assigned randomly. Each site consisted of two paired 150 x 200 cm plots: a control and a drought treatment. Control plots received ambient precipitation while drought plots experienced 35% precipitation reduction year-round using passive precipitation removal shelters from 2010 to 2018. These shelters were 230 × 310 cm and sufficiently wide to provide a large buffer around the plots that also received the drought treatment; their slanted roofs were composed of V-shaped plexiglass strips that served as troughs to intercept rainfall and snow (Figure S1.). These strips were connected to a gutter system that directed water away from the plot. Furthermore, vinyl flashing was installed vertically around the plot, below and a few cm above ground, but beneath the shelters to hydrologically isolate them (Hoover et al . 2015).