Date Approved

12-2014

Graduate Degree Type

Thesis

Degree Name

Biology (M.S.)

Degree Program

Biology

First Advisor

Gary K. Greer

Second Advisor

Shaily Menon

Third Advisor

Robert Hollister

Abstract

Prairie fens contain high levels of floral biodiversity, including 19 state threatened or endangered plant species, and are classified as rare and vulnerable communities by the Michigan Natural Features Inventory. The objective of this thesis was to develop multiple-regression (MR) models that reliably predict total, native, and invasive floral species richness for use by conservation organizations. Floral biodiversity surveys were conducted in eight southern Michigan prairie fens during the 2012 growing season. Simple linear regressions between fen size and biodiversity were used to optimize sampling strategy and effort (i.e., number of transects and plots per transect) in surveys conducted in 12 additional prairie fens in the 2013 season. Prairie fen characteristics including proximity to neighboring fens, size, shape, depth to water table, elevation, and land cover of the environmental matrix within a 250m buffer zone around each prairie fen were included as independent variables in the developed models. Nine sets of MR (MR families) were developed to predict total, native, and invasive floral species richness. The first MR family contained all normally distributed variables (p ≤ 0.05) to optimize the independent variables (i.e., find the minimal set of independent variables to generate a robust MR). As these models were not significant, the subsequent MR families were developed using ordinated independent variables. Correspondence Analysis (CA) was performed for each latter model family to retain as much variation as possible. The loading scores of each significant CA axis were used as independent variables for the eight subsequent MR families. Multiple Regression Family 2, the simplest of the ordinated model families, produced significant models and was the most reliable with the highest R2 adj and highest model fit compared to the other MR families. These models were most influenced by the land cover of the area immediately surrounding a prairie fen and, notably, did not include fen hydrology. We recommend managers consider the current land cover surrounding a fen, namely the presence of forested areas that negatively impacted diversity when developing management strategies.

Included in

Biology Commons

Share

COinS