dc.description.abstract | The influence of land cover on stream ecosystems has increasingly been the focus of
research in a variety of fields, including ecology, hydrology, and engineering. Human-altered
land cover can increase overland storm runnoff and be a source of nutrients, chemicals and
sediment to streams, which can negatively affect biota. Urbanization has been recognized as a
particularly influential form of land cover, and low levels of urban development (e.g., suburban
land) has become the focus of an increasing number of studies, as it is predicted to be a more
prevalent form of land-cover change over the next century. The coastal plains of the southeastern
US are a relatively understudied region; however, some research has indicated that these lowland
streams may be influenced by land cover to a lesser degree than those from more frequently
studied and higher gradient landscapes. Current trends and predictions in land-cover change and
human population growth suggest that streams and rivers will become increasingly influenced by
human activity. Thus, research is warranted to further investigate the influence of low levels of
urban development on streams and how this influence may vary regionally.
Studies on the influence of land cover on stream ecosystems are complicated by multiple
issues, including the following: 1) experimental manipulation of whole watersheds is generally
impractical, therefore most studies have been observational, 2) land cover is generally expressed
as proportions; thus, land-cover classes are likely to be correlated, 3) whole-watershed
replication is frequently low relative to the number of variables considered, thus constraining
statistical analyses, 4) land cover is typically not the direct cause of biological degradation, and
5) real-world data are not ideal and anomalies (outliers) are generally present and not always realized by the investigator. Some of the above issues are problematic when using traditional
statistical methods, including ordinary least-squares (OLS) regression. Small sample sizes,
correlated predictor variables, and outliers separately and in combination contribute to unreliable
estimation and prediction by OLS regression. Whereas no true solution to these problems exists,
most ecologists often do not consider alternatives that may, to some degree, address or alleviate
these analytical challenges.
Overall, my dissertation provides important information regarding appropriate statistical
choices for analyzing real-world data which usually should not be assumed to conform to
assumptional requirements of traditional methods. Nowhere are these issues more evident than in
land-cover or ecological studies in general where small sample sizes are commonplace, predictor
variables are frequently highly correlated, and outliers are likely present. In addition, my
dissertation contributes to research on the influence of land cover on stream ecosystems in the
understudied southeastern coastal plains, suggesting that impervious surface cover ≤ 11% likely
influences hydrology and physicochemistry of streams. This information is particularly
important because low-levels of urban development are predicted to be the most prevalent of
land-cover change along the Gulf Coast in the foreseeable future. Last, my dissertation
importantly highlights specific differences in macroinvertebrate taxonomic and trait composition
that exist in lowland coastal plains versus those in highland regions. Macroinvertebrates are the
most frequently used taxonomic group in stream biomonitoring, and my research reinforces the
body of literature suggesting that regional bioassessment metrics are needed to accurately
identify impairment specific to each region. | en_US |