CONSEQUENCES OF SPATIAL AND TEMPORAL CLIMATE VARIABILITY FOR SPECIES DISTRIBUTION MODELING

Our understanding of how species will respond to global change is still limited. Species distribution models (SDMs) are used to generate hypotheses regarding the potential distributions of species under different environmental conditions. However, that species observations and climatic variables are not measured at the same spatial and temporal resolution still hinders our ability to forecast species range shifts and expansions in response to global change. One of the possible consequences of this data mismatch is the observed discrepancy between realized climate niches in species' native and invasive ranges. In the first chapter of my dissertation, I address this issue by evaluating niche similarity between native and invaded ranges for 10 species using a combination of monthly and inter-annual climatic variability data. My results suggest that some species' ranges may be constrained by one aspect of climatic variability in the native range but a different one in the invaded range. A second issue, also a consequence of the spatio-temporal mismatch, is that weather station data are often spatially interpolated to match the species observations without any uncertainty assessment. The second chapter evaluates and quantifies the effects of three complementary aspects of uncertainty present in weather station data interpolations. I examine the influence of topographic heterogeneity, interannual climatic variability, and distance to weather station on SDM performance for 20 well observed North American breeding birds, and show that topographic heterogeneity has the highest contribution to omission errors, or false absences. A third consequence of the spatio-temporal data mismatch is the inability of global simulations to capture local manifestations of climate change. This inability can limit the capacity of SDMs to produce accurate simulations for species whose distributions depend on small-scale climate phenomena. While changes in global climatic patterns are projected using global climate model (GCM) simulations, local climatic trends are not always well represented by GCMs, or by simple downscaled projections derived from GCMs. In the final chapter of my dissertation, I use interpolated, fine-scale historic climate records in a novel approach to estimate the sensitivity of SDM's to locally coherent changes in temperature and precipitation at larger scales, using coast redwood (Sequoia sempervirens) as an example. Overall, the results of this thesis confirm the importance of establishing an appropriate relational basis in time and space between species and climatic observations. Historical collection records should be thoroughly analyzed and integrated with historical climatic time series to gain a better understanding of species' response to climate variability in the past, thereby informing the selection of appropriate spatio-temporal scales of climate variability for projections under present and future conditions