Validating dispersal distances inferred from autoregressive occupancy models with genetic parentage assignments.

1 Dispersal distances are commonly inferred from occupancy data but have rarely been validated. Estimating dispersal from occupancy data is further complicated by imperfect detection and the presence of unsurveyed patches. 2 We compared dispersal distances inferred from seven years of occupancy data for 212 wetlands in a metapopulation of the secretive and threatened California black rail (Laterallus jamaicensis coturniculus) to distances between parent-offspring dyads identified with 16 microsatellites. 3 We used a novel autoregressive multi-season occupancy model that accounted for both unsurveyed patches and imperfect detection to quantify patch isolation using buffer radius (BRM) and incidence function (IFM) connectivity measures at 15 scales (1-10, 15, 20, 25, and 30 km). Connectivity measures were then fit as colonization covariates in occupancy models to estimate a model-averaged dispersal distance. 4 As predicted, colonization was more strongly related to connectivity at small spatial scales (< 10 km). AIC weights were greatest at 7 km for BRM and at 4 km for IFM. 5 Model-averaged dispersal distances (BRM = 7.46 km; IFM = 5.48 km) showed good agreement with the mean M(+/- SE) dispersal distance from 23 parent-offspring dyads (5.58 +/- 1.92 km), indicating reasonably accurate mean dispersal distances can be inferred from occupancy data when isolation strongly affects colonization.