Theses and Dissertations at Montana State University (MSU)
Permanent URI for this collectionhttps://scholarworks.montana.edu/handle/1/733
Browse
1 results
Search Results
Item Non-native species distributions in space and time : integrating ecological theory and predictive modeling(Montana State University - Bozeman, College of Agriculture, 2012) Brummer, Tyler Jacob; Chairperson, Graduate Committee: Lisa J. Rew.; Bruce D. Maxwell, Megan D. Higgs and Lisa J. Rew were co-authors of the article, 'Maximizing the utility of landscape scale species distribution models' in the journal 'Biological invasions' which is contained within this thesis.; Bruce D. Maxwell, Subhash R. Lele and Lisa J. Rew were co-authors of the article, 'Detection error in plant surveys: to correct or not to correct' in the journal 'Journal of applied ecology' which is contained within this thesis.Invasive plant species are perceived as a problem globally, but management occurs locally. Theoretical developments concerning the distribution of plant species and invasions have generally focused on coarse scales, with relatively little work performed at finer scales relevant to local landscape management. The majority of such model predictions are static reflections of current conditions. As species invasions are a temporally variable process, the need for tools to predict invasions through both space and time are vital. Thus, this thesis explored distribution models to predict non-native species occurrence, investigated sources of uncertainty in these models, and quantified the key drivers of non-native species metapopulation dynamics. Sampling methodology and sample size requirements to inform logistic regression models used to predict invasive species realized distributions were evaluated with simulation and empirical data from two multi-species surveys. Transect sampling was the most efficient way to generate species occurrence data and consequently landscape scale species distribution models. Empirical and simulation modeling results indicated minimum sample sizes of between 0.06 and 0.23% of the study area to maximize model predictive ability, independent of site characteristics. However, landscape scale species distribution models were more predictive at sites with steeper environmental gradients and for species at their ecophysiological range limits. Detection error in the plant species surveys was also quantified, as well as its effects on predictions and uncertainty of species distributions. Detection error did not practically alter predictions, nor model based uncertainty estimates if the detection rate was greater than 87%. However, at lower detection rates care needs to be taken when interpreting response variables and prediction certainty. Finally, multi-season repeat survey data were used to investigate the key spatial and temporal drivers of non-native species colonization and extinction. The drivers of non-native species colonization and extinction were sometimes simple but other times resulted from complex interactions between dispersal, disturbance, habitat and temporal climatic variation. Overall, these results demonstrate the need for reliable species occurrence records and monitoring data to fully characterize species distributions at present and their dynamics resulting in future distributions.