Models of a species’ distribution and models of a species’ spatially explicit density are valuable tools for conservation. They allow researchers to estimate changes in distributions, densities, and populations, based on changing environmental conditions. To trust such estimates, however, the quality of models is exceedingly important. Model performance can be affected by the scale at which the environment is characterized, species-specific characteristics such as prevalence and habitat specialization, or the quality of data collected. High-quality models of current distributions and densities can be used to assess the current status of species over a study extent. With environmental data extrapolated to the past or future, such models can be used to predict changes in species status through time.
In Chapter 2, I developed and assessed multi-scale models of species distribution and abundance using the boosted regression tree algorithm. Multi-scale species distribution models consistently performed as well or better than best single-scale models. This multi-scale framework greatly reduces the number of models necessary to account for species’ response to environmental scale and avoids potential mischaracterization of important environmental scales. Hurdle models of abundance, however, performed poorly throughout the study. Fine-scale abundance is more difficult to model accurately than occurrence. Alternative methods may be more appropriate.
In Chapter 3, I examined the comparative roles of species’ prevalence, sample size, and habitat specialization, on the performance of species distribution models. Although all three affected model performance, the role of prevalence was minor compared to the role of habitat specialization. I found a species-specific quadratic effect of sample size on model performance where generalist species required larger minimum sample sizes for accurate models. I then used my findings to recommend minimum sample sizes based on habitat specialization.
In Chapter 4, I examined the use of occurrence and abundance data from the eBird database for fine-scale species distribution and density modeling. I compared the results of distribution and density models built on eBird data to the results of models built on highly standardized surveys from the Oregon 2020 project. To convert observed abundances in eBird to densities, I used estimates of detection probability from the Oregon 2020 data. Overall, the results of distribution and density models from the two datasets were very similar. I explore the reasons for differences and give guidelines for how to best use eBird data in fine-scale distribution and density modeling.
In Chapter 5, I modeled current and historic distributions and densities of seven species of bird in the Willamette Valley, Oregon. I used reconstructed 1850s habitat data to hind-cast densities and distributions to pre-European-American settlement throughout the study area. I estimated current and historic populations for each species as well. I found population declines exceeding declines in suitable grassland and oak habitats for grassland and oak specialists. I found relatively stable populations and suitable habitat for riparian and edge specialists and increases in population and suitable habitat for species associated with coniferous forests. Hind-casted distributions and densities can provide a new baseline for the effects of anthropogenic habitat change on bird populations in the Willamette Valley.
My research aims to better the understanding of species distribution and density modeling. I address important topics from the scale at which environmental variables should be characterized to the minimum sample sizes for accurate models and the species’ characteristics that affect that sample size. I build guidelines for fine-scale distribution and density modeling with the popular rapidly growing citizen science eBird database. Finally, I use fine-scale multi-scale models to estimate historic distributions and populations that provide a new baseline for change since European-American settlement.