|| Checking for direct PDF access through Ovid
Developing strategies that reduce the impacts of climate change on biodiversity will require projections of the future status of species under alternative climate change scenarios. Demographic models based on empirical data that link temporal variation in climate with vital rates can improve the accuracy of such predictions and help guide conservation efforts. Here, we characterized how population dynamics and extinction risk might be affected by climate change for three spotted owl (Strix occidentalis) populations in the Southwestern United States over the next century. Specifically, we used stochastic, stage-based matrix models parameterized with vital rates linked to annual variation in temperature and precipitation to project owl populations forward in time under three IPCC emissions scenarios relative to contemporary climate. Owl populations in Arizona and New Mexico were predicted to decline rapidly over the next century and had a much greater probability of extinction under all three emissions scenarios than under current climate conditions. In contrast, owl population dynamics in Southern California were relatively insensitive to predicted changes in climate, and extinction risk was low for this population under all scenarios. The difference in predicted climate change impacts between these areas was due to negative associations between warm, dry conditions and owl vital rates in Arizona and New Mexico, whereas cold, wet springs reduced reproduction in Southern California. Predicted changes in population growth rates were mediated more by weather-induced changes in fecundity than survival, and were generally more sensitive to increases in temperature than declines in precipitation. Our results indicate that spotted owls in arid environments may be highly vulnerable to climate change, even in core parts of the owl's range. More broadly, contrasting responses to climate change among populations highlight the need to tailor conservation strategies regionally, and modeling efforts such as ours can help prioritize the allocation of resources in this regard.