GWAS/GEA
The GAPIT3 model selection was performed to compare the efficacy of five models, namely General Linear Model (GLM), Mixed Linear Model (MLM), Compressed MLM (CMLM), Fixed and random model Circulating Probability Unification (FarmCPU), and Bayesian-information and Linkage-disequilibrium Iteratively Nested Keyway (BLINK) (Wang & Zhang, 2021). FarmCPU was selected based off model selection and for its false positive control (Liu et al., 2016). It was then implemented to test both environmental and phenotypic parameters.
From the GEA analysis, 573 adaptive loci passed the significance threshold of a false discovery rate of 0.1. 139 of these involved soil parameters, 393 involved climatic conditions and 41 of these were landscape specific. All loci and positions have been reported (Supplementary Information 2). The GWAS discovered eight adaptive loci for height and one for DBH. Of the loci involved with height, three were also adaptive loci for environmental parameters, namely minimum water vapour, forest fire risk and maximum solar radiation. As forest fire risk assessment is based on aridity this can be considered a variable.