EXTINCTION SELECTIVITY IN PERMIAN AND TRIASSIC GASTROPODS: MULTIVARIATE ANALYSES OF GEOGRAPHIC RANGE, SPECIES RICHNESS, BODY SIZE, AND ABUNDANCE

We used a global gastropod database spanning the Early Permian through Early Jurassic to investigate the influences of geographic range, species richness, body size, and abundance on genus-level extinction risk. We found that range and richness are inversely associated with extinction risk, whereas size of the smallest species in the genus is positively associated. The sizes of the largest and median species in the genus are not associated with extinction risk. Because geographic range, species richness, and minimum size are significantly correlated with one another, we used multivariate logistic regression to estimate their independent associations with extinction risk. The inverse association of geographic range with extinction risk differs little between univariate and multivariate regression models (odds of extinction are halved with each additional province occupied), whereas the inverse associations of species richness and minimum size are attenuated in the multivariate model, suggesting that some of their association with extinction risk reflects their correlation with geographic range. Abundance in regional assemblages is inversely associated with extinction risk in the univariate regression model, but positively associated with extinction risk in the multivariate model including geographic range. Geographic range is more strongly inversely associated with extinction risk during background intervals than during intervals of mass extinction, consistent with findings for other taxa and other intervals. The limited associations of body size, species richness, and abundance with extinction risk contrast with the predictions of simple models and the distribution of extinction risk in extant taxa. This discrepancy may reflect limitations in the resolution of paleontological data or the importance of rare disturbances in causing extinction that may limit the applicability of simple ecological models over geological timescales.