Regional stream classification frameworks and biotic homogenization in streams of the Midwest: An evaluation using landscape-scale patterns in fish communities

Abstract

This dissertation includes three studies that explored landscape-scale patterns in reference stream fish communities in Kansas, Missouri, Nebraska and Iowa (U.S.A.) to inform the design of classification frameworks for streams, and to elucidate important mechanisms that shape stream ecosystems in this region. The development of regional stream classification frameworks is a work in progress. Much debate addresses whether more terrestrially or more aquatically– defined frameworks better explain landscape-level ecological variation in streams. Chapter One addressed this and other issues by comparing the classification strengths (CS) of 8 a priori and a posteriori classification frameworks for streams at two scales. CS was based on the difference between mean within-group fish community similarity and mean among-group similarity (Sorenson and Bray-Curtis similarity indices) within each framework. The a priori frameworks included: Strahler order; Hydrologic Unit Codes (HUCs) or “watersheds”; two terrestrial approaches (Bailey and Omernik ecoregions), and one aquatic approach (Maxwell et al. 1995). The a posteriori frameworks included: geographic proximity groupings, non-spatial random groupings, and groupings based on fish community similarity. Chapter two explored the relative importance of physical vs. ecological boundaries to stream fish dispersal by constructing and testing hybrid ecoregion-bywatershed frameworks using the same CS analysis as above. iv To understand the influence of humans on Chapter One and Two results, and to quantify homogenization of stream fish communities, Chapter Three evaluated the change in beta similarity among watershed fish faunas in Kansas before major human modifications to streams (pre-1958) and after modifications (post 1988). Chapter one showed that the framework based on geographic proximity produced the highest CS values at both scales compared to other classification frameworks. Chapter two indicated that the hybrid ecoregion-by-watershed frameworks did not effectively classify stream fish communities at the scale tested here. Chapter three revealed that past watershed faunas in Kansas were more regionally distinct, and have become 8.2% more similar to each other since major stream modifications. Results from both Chapters one and two indicated that geographic proximity is one of the most influential forces on landscape-scale patterns in stream fish communities, which highlights the importance of historical species distributions on modern stream biota.