2.3 Experiment setup and site maintenance
Before planting, each site was treated with a broad-spectrum herbicide (Roundup, active ingredient Glyphosate, SpecialistSales, Toowoomba, QLD, Australia) across the entire plot for the DRO and Cow Bay, or along the planting strips (~1m wide) across the slope at Thiaki. Saplings ~6 to 12 months (mean height 37 cm, range 5 to 107 cm) were planted at each of the three trial sites in late March to April 2022 at 3 m x 3 m spacing to provide for an even light environment as plants grew. This included ten replicate saplings each of the two provenances (lowland and upland) for 16 species planted at each of the three trial sites (n = 320 at each site). The order of plants was randomised within ‘blocks’ with 1 replicate plant from each of 16 species and provenance (upland or lowland) represented in each block (n = 32 plants per block x 10 blocks at each site). At planting, soil cores were augered (diameter 200mm, depth ~30 cm) with each hole receiving 50g of an organic fertilizer pellet (Organic Life, Terra Firma fertilizers, Beaudesert Qld, Australia), and the pre-soaked plants replanted and watered in immediately. At the Thiaki site, there was a lack of significant rainfall in the weeks immediately after planting, so seedlings were hand-watered every 3 days until the first rains. After seedling establishment, weed management consisted of periodic row spraying with a C4-targeted herbicide (Fuzilade Forte 128C, active ingredient Fluazifop-p, SpecialistSales, Toowoomba, QLD, Australia).
Measurements of growth and survival
Two months after the initial planting 98.4% survival was observed across all sites. Supplementary planting of additional replicates was carried out to maintain the complete planting matrix; these replacement plants were not included in our survival analyses. Over the experiment, total plant height (m) and plant health (described below) were recorded: at planting, and 4, 12, 17, and 24 months after planting. For multi-stemmed individuals, the total height of the leading stem was measured. Health was assessed qualitatively using an index of 0 to 5, with 0 = dead, 1 = main stem dead but with resprouting evident, and levels 2 to 5 representing; <25%, 25 to 50%, 50 to 75%, and plants with a health class of 1 or lower were classified as functionally dead. For analysis of height data, all functionally dead individuals were excluded.
To account for potential within-site variation in environmental conditions (i.e. edaphic or light conditions) known to impact tree growth across the experimental sites (Cheesman et al. 2018) we tested for spatial autocorrelation present in the tree growth rates, using Moran’s I. This test indicated significant spatial autocorrelation in growth data within each plot. As trees were not grouped by species in the plots, this autocorrelation is likely due to intra-site variation in soil properties such as rockiness or fertility. Therefore, to account for this spatial autocorrelation, we built a spatial autologistic term (Dormann 2007) using height data from the August 2023 campaign normalised for each site. We generated autologistic terms using focal windows with diameters of three, five, and seven trees, then selected the window size that addressed spatial autocorrelation while exerting the least influence on the model. All three autologistic terms accounted for spatial autocorrelation in model residuals, thus, we chose the term with a focal diameter of seven trees as it had the least influence on the model and was least likely to bias model parameter estimates (Figure S2).