Daniels, Jason Michael2023-08-102023-08-102021-08https://unbscholar.lib.unb.ca/handle/1882/37279Many models describing the relationship between landcover and diversity have been developed, but there is relatively little literature on how well these models perform on new data. Claims of increased understanding of ecological relationships require models that make more accurate predictions on new data than we could make before those studies were done. In this study I assess the spatial transferability of models using landcover variables to predict bird abundance and species richness from 3 different regions across the conterminous United States. Notably, final models for each dependent variable always varied across training regions. Most models performed relatively poorly on average when applied to test data and they generally performed worse with increasing distance from the training location, yet in nonlinear and complex ways. These results suggest that the relationships between bird abundance/diversity and landcover are real but the mechanisms driving these targets are changing in space. While improving the transferability of models is an important and useful objective, inherently many ecological models likely have spatial and temporal limits to their transferability.ix, 66electronicenhttp://purl.org/coar/access_right/c_abf2master thesisHoulahan, JeffBiology