Author

David F. Ford

Abstract

Unionid mussels are a guild of freshwater, sedentary filter-feeders, which play a critical role in freshwater systems. Mussels are currently experiencing a global decline in both species richness and abundance, due to invasive species, human alteration of water systems, and climate change. In North America, which is considered to have the highest global diversity of bivalve species, native mussels are currently declining rapidly with at least 37 species considered to already be extinct. If extant mussel species are to be preserved, then it is vital that conservation efforts be prioritized towards areas in which they are likely to be found. This is often done through ecological niche models. Maxent for example uses the principle of maximum entropy on presence data and environmental variables to create a suitability score for a particular area, and is one of the most widely used of the ecological niche modeling programs. It has been used to make maps predicting the suitability scores for multiple species but very little ground-truthing to see if the maps are assigning the correct scores has been conducted. We ground-truthed the Maxent program’s suitability score from 30 samples for six threatened mussel species in the rivers of East Texas by visiting differently scored sites to determine if the Maxent suitability scores were reflective of actual abundances. Ground-truthing was done by sampling at 138 sites throughout East Texas. These sites had been assigned suitability scores from a previous study, and the mussels found at the site were compared to those that were predicted to be there by Maxent. The new maps created by Maxent were compared to the original maps to see if new occurrence points added to the predictive ability of the maps by looking at the test AUC and test gain values. The influences of new data on Maxent’s predictive ability for finding a particular mussel species at a site, and the number of mussels found at a site were also investigated by linear and logistic regression. Additional occurrence points were found to significantly improve the predictive maps for the Triangle Pigtoe, the Texas Heelsplitter, and the Southern Hickorynut, and all maps were found to accurately predict locations for mussels. Maxent’s predictive ability via their suitability scores was improved for all species with additional occurrence points. However, for almost all of the species looked at, there was a data cap, which was a point at which additional data no longer improved the models. This suggests that the amount of data necessary to make accurate maps may not be as large as originally thought, and when trying to conserve an organism this could be important.

Date of publication

Spring 5-6-2013

Document Type

Thesis

Language

english

Persistent identifier

http://hdl.handle.net/10950/110

Included in

Biology Commons

Share

COinS