Using latent selection difference to model persistence in a declining population

Population persistence is a direct measure of the viability of a population. Monitoring the distribution of declining populations or subpopulations over time can yield estimates of persistence, which we show can be modeled as a latent selection difference (LSD) contrasting attributes of sites where populations have persisted versus those that have not. Predicted persistence can be modeled with predictor covariates to identify factors correlated with species persistence. We demonstrate how to model persistence based on changes in occupancy that can include adjustments for detection probability. Using a known historical distribution of the western grebe (Aechmophorus occidentalis), we adapted methods originally developed for occupancy modeling to evaluate how environmental covariates including emergent vegetation and human developments have affected western grebe persistence in Alberta. The relative probability of persistence was correlated with the extent of shoreline bulrush (Scirpus lacustris), which is important vegetation for nesting cover. We also documented that western grebe populations were less likely to persist on lakes in the boreal forest, primarily located on the northern boundary of the species' range. Factors influencing occupancy were different than those determining persistence by western grebes; persistence and occupancy were not correlated. Persistence was more likely on lakes with recreational development, reflecting reliance by grebes on the larger, fish-bearing waterbodies that also are attractive for lakeshore development. Unfortunately, the correlation with recreational development on Alberta's lakes puts grebes at risk for loss of brood-rearing habitats--primary threats to altricial birds--if steps are not taken to prevent disturbance to bulrush stands. Identifying factors related to the persistence of a species--especially one in decline--is a fundamental step in conservation management.