Recombination rate

Variants were called using GATK version 4.0.7.0 (DePristo et al., 2011) for one representative individual per region (Table S1). Briefly, variants were first called for each individual with HaplotypeCaller in BP-RESOLUTION mode, then combining those GVCF files for each sample into a single one using CombineGVCFs per scaffold of interest, and finally joint genotyping with GenotypeGVCF. The default filter of GATK (--phred-scaled-global-read-mismapping-rate 45; --base-quality-score-threshold 18; --min-base-quality-score 10) was used. Recombination rates for 100 kb non-overlapping windows along the genome were estimated using the iSMC approach from (Barroso et al., 2019). We fitted an iSMC model with 40-time intervals and five categories of recombination rates to the samples from each population and optimized parameters in composite likelihood fashion (Barroso & Dutheil, 2021). We then obtained recombination landscapes of single-nucleotide resolution by performing posterior decoding in each diploid using the estimated parameters, and computed a consensus map for each sample by averaging over (for each site) the posterior estimates of rho = 4*Ne*r from all diploids. The final map of 100 kb resolution was obtained by further averaging the single-nucleotide estimates over 100 kb in non-overlapping windows.