Regional comparisons of phylogenetic diversity (PD) are difficult to interpret without a baseline for expected PD as a function of species richness (SR). We developed an SR-matched randomization method that preserves phylogenetic structure using mean pairwise distance (MPD). We partitioned the MPD into equal-width intervals and assigned bin quotas proportional to the observed frequencies. For each SR, we sampled species sets, computed the MPD, and accepted a set when its MPD fell within an MPD bin that still had a quota based on the observed frequencies. This bin-and-quota rule preserves the balance between clustering and overdispersion. Using the accepted sets, we estimated the expected PD as a function of the SR, and converted each region's PD at its SR to a standardized Z -score. We applied this to Korean vascular plants using 2.23 million records and presence–absence data for 162 administrative units. For the benchmarks, the MPD-constrained null model increased the proportion within the intervals from 79.6% to 92.6% in the observed units and from 85.9% to 97.1% in the evaluation units. MPD-constrained randomization tracked the observed PD–SR relationship more closely, conservatively maintained SR-specific uncertainty, and improved interpretability by excluding implausible draws. The southern and insular regions exceeded the expected PD, whereas high mountainous areas along the Baekdudaegan Mountains showed negative deviations. This framework is readily applicable to other taxa and regions to produce structure-aware, SR-specific expected PD baselines for fair mapping of PD hotspots and deficits, and support practical area prioritization alongside threats and connectivity. • New randomization approach preserves evolutionary relatedness among species. • Eliminates species richness bias for fair phylogenetic diversity comparisons. • Aligns randomized communities with the mean pairwise distance distribution. • Applied to Korean vascular plants, revealing distinct regional diversity patterns. • Outperforms existing null models in capturing phylogenetic diversity patterns.
Lee et al. (Tue,) studied this question.