I'm wondering how Dxy is computed using allele frequency data from pool-seq.

For example here:

"Allele frequencies based on SNPs were estimated with snp-frequency-diff.pl, filtering with the same thresholds a described for FST estimates. We computed the measure of divergence DXY for bi-allelic loci (~99.2% of the total allele estimates) as the average number of nucleotide differences for each gene based on the formula using allele frequencies from Smith and Kronforst (2013) defined as Dxy = 1/n Σ pix(1-piy) - piy(1-pix)." (Schirrmann et al.,2018).

The snp-frequency-diff.pl gives you the difference in allele frequency between two populations, not the actual frequency within each population.