SNP data for 23 forest tree species

SNP calling was performed on uniquely aligned reads using GATK v4.5.0.0. HaplotypeCaller was run in GVCF mode for individual samples, followed by joint genotyping with GenomicsDBImport and GenotypeGVCFs. SNPs were extracted using SelectVariants and filtered with VariantFiltration based on quality metrics (e.g. QD < 2.0, MQ < 40.0). SNPs from paralogous regions were removed using HDplot (Hmax=0.6, RAF range 0.2–0.8). Genotype-level filtering with bcftools retained only biallelic SNPs with sufficient depth (DP ≥ 6), genotype quality (GQ ≥ 20), and coverage. Samples with >80% missing data were excluded. Invariant sites were extracted separately and filtered by depth before merging with SNPs into V0 VCFs. These files were used to identify outliers and duplicates via PCA (snpR) and admixture analysis (sNMF, LEA). Relatedness was assessed with KING and PLINK to remove likely handling errors or redundant samples, unless biological duplicates (e.g. clones) were suspected.