Determining the Influence of Abiotic and Biotic Predictors on Ecological Niche Models

<p>Biotic variables, including those reflecting interactions such as competition and parasitism, are generally assumed to have negligible effects on broadscale species distributions and are typically excluded from ecological niche models (ENMs). However, the role such interactions play in shaping species distributions is increasingly recognized, sparking the development of methods for their integration into ENMs.</p> <p>Among the most common approaches are those that limit the ENM output of the focal species by the distribution of biotic interactors (post-processing) and those that incorporate biotic interactors as predictor variables. These endeavors hold widespread and critical usages, such as predicting the impacts of climate change on species distributions, invasive species control, and disease risk mapping. Despite the importance and growing number of methods and studies incorporating biotic interactors in ENMs, similar developments to evaluate whether model performance is in fact improved by their inclusion are lagging. This is also true of methods to guide appropriate biotic variable selection and inferences regarding their biological relevance.</p> <p>Here, I present three chapters that make methodological advances to evaluate model performance and biotic variable contributions when accounting for them in ENMs. The first chapter evaluates the impacts of accounting for biotic interactions via post- processing on model performance and explores its potential implications for ENM applications. In it, I demonstrate that accounting for biotic interactions via post-processing balances model omission and commission error rates for biological entities with presumed parapatric distributions (i.e., vegetation types). In the second chapter, I develop a novel null model test to enable comparisons of model performance between two or more ENMs built with different predictor variable sets by extending a null model framework proposed by Bohl et al. (2019). Here, I demonstrate how ENM performance among models built with different predictor sets may not differ significantly despite considerable differences in their geographic predictions and their underlying explanatory variables. The approach presented in this chapter provides a rigorous method for variable selection based on model performance, while pairing it with an assessment of variable contributions and ecological realism. Lastly, I evaluated if parasite ENM predictions are improved by including information on host availability (host ENMs and richness) as predictor variables, using the tool developed in chapter 2. Similar to results from the previous chapter, I found that biotic predictors did not impact model performance despite influencing the biological realism of parasite models. This approach constitutes an advance towards a comprehensive methodology for relevant biotic variable selection in parasite (and by extension pathogen) ENMs.</p> <p>Throughout this dissertation, I demonstrate how accounting for biotic interactions influences ENM performance, making them more biologically realistic. These advances promote model transferability for many ENM applications ranging from biodiversity assessments under climate change to disease risk mapping.</p>

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Biology

Grantor

The Graduate School and University Center of The City University of New York

Year dc:date.available

2023

Author dc:creator

• Johnson, Erica

Advisor dc:contributor.advisor

• Robert Anderson

Committee members dc:contributor.committeemember

• Phillip Staniczenko
• Elia Machado
• Maria Diuk-Wasser
• Sean Maher

• Bioinformatics
• Biology
• Ecology and Evolutionary Biology
• Ecological niche models
• Species distribution modeling
• Biotic interactions
• Vegetation models
• Parasite distributions
• Disease risk mapping

Repository record dc:identifier

https://academicworks.cuny.edu/gc_etds/5502

OAI identifier oai:identifier

oai:academicworks.cuny.edu:gc_etds-6617