Ghosts of a Structured Past: Impacts of Ancestral Patterns of Isolation-by-Distance on Divergence-Time Estimation
Summary
Ingested 2026-04-21. 3 findings extracted and verified.
Findings worth citing
Finding 1 — When ancestral populations follow a linear stepping-stone model with low migration, SNAPP overestimates divergence times between end-deme-derived species by roughly 700% when TD/ND = 1.
The most dramatic overestimation occurred between end species when T D / ND = 1 at ~700% the true age. Importantly, this was not merely the result of a low T D / ND ratio, as the other migration regimes performed well. — p. 577
Why this is citable: This is the headline quantitative result of the paper, grounding claims that ancestral IBD can severely bias coalescent-based divergence dating. Useful for citing when critiquing or contextualizing SNAPP/BEAST divergence estimates in phylogeographic studies.
Counter / limitation: The 700% figure comes from a specific simulation scenario (m=0.001, TD/ND=1) with a fixed 10-deme linear stepping stone; empirical systems may not match these parameters, and the result is SNAPP-specific rather than a general property of all coalescent methods.
Topics: divergence_time_estimation, isolation_by_distance, coalescent_simulation
Finding 2 — Divergence-time overestimation due to ancestral IBD becomes severe specifically when the ratio of divergence time to ancestral population size falls below 10 and migration is low.
Divergence-time overestimation in SNAPP becomes severe when the divergence-to-population size ratio < 10 and migration is low. — p. 573
Why this is citable: Provides a concrete, citable threshold (TD/ND < 10) that downstream studies can use to decide whether their systems are at risk of ancestral-structure bias. Directly extends Rosenberg and Feldman’s (2002) TD/Ne ≈ 5 rule to spatially structured ancestors.
Counter / limitation: The threshold was derived from simulations with 10 demes and specific migration rates (0.1, 0.01, 0.001); empirical systems with different deme numbers, geometries, or continuous habitats may have different critical ratios.
Topics: divergence_time_estimation, isolation_by_distance
Finding 3 — Ancestral IBD inflates estimated effective population size relative to census size, with mean ancestral Ne reaching ~7370 versus Nc = 1000 at m = 0.001 when sampling end species.
However, when m = 0.001, π 12 for end species was 0.012948, an ancestral Ne = 7370. Center species, on the other hand, only increased to Ne = 1255. — p. 580
Why this is citable: Quantifies how ancestral spatial structure inflates Ne estimates derived from pairwise divergence (π12) relative to the known census size, showing mean Ne (~7370) can exceed Nc (1000) by roughly 7-fold when m=0.001 and end species are sampled. Relevant to any study interpreting coalescent-based Ne as a proxy for census population size, particularly in conservation contexts involving fragmented or range-edge populations.
Counter / limitation: The Ne value of ~7370 is derived by applying the panmictic expectation (equation 5: Ne = (π12 − 2TDμ) / 4μ) to data generated under a structured model; this is the very source of the inflation, not a failure of a general Ne estimator. Methods that explicitly account for population structure (e.g., structured coalescent models) would not produce this bias. Additionally, the specific TD/ND ratio used matters: the paper reports this figure for TD/ND = 50 with m = 0.001, so the magnitude may differ for other parameter combinations.
Topics: effective_population_size, isolation_by_distance, conservation_genetics
Read the paper
doi.org/10.1093/jhered/esaa042