6. Differences by the additional data treatment
Since data obtained in eDNA metabarcoding are different in nature from conventional tissue-based approaches and contain erroneous sequences, previous studies have indicated that data filtering or the conversion of relative read counts to semi-quantitative rankings are effective in population-based analysis (although their scope is mainly phylogeography). To investigate their effectiveness in landscape genetics analysis, D PS andG ST were calculated from the eDNA dataset in which the following treatments were performed and compared to tissue-based statistics: treatment 1, haplotypes with a low frequency (\(p_{i}<0.01\)) in each population were removed from the data (replaced to 0 reads) (Tsuji et al. 2023); treatment 2, haplotypes with less than half of the proportion of the most predominant haplotype in each population (\({p_{i}<max\ p}_{i}/2\)) were removed from the data (Tsuji et al. 2023); treatment 3, haplotype frequency was converted into the following semi-quantitative rankings: rank 1 if \(p_{i}\leq 0.5\) ; rank 2 if \(0.5<p_{i}\leq 0.75\); rank 3 if\(0.75<p_{i}\leq 0.9\); rank 4 if \(0.9<p_{i}\) (Turon et al. 2020).