Characterizing riparian ecosystem composition and structure in lower Michigan.

The ecology of river, lake, and wetland ecosystems is strongly affected by the routing of their source waters. Relative contributions of land surface runoff and groundwater help shape seasonal hydrography and physical conditions in rivers and riparian areas. As the interface between rivers and terrestrial landscapes, riparian ecosystems are uniquely situated to influence, and be influenced by, hydrologic variation. Sharpening predictions of surface and subsurface contributions to local riparian ecosystems across broad heterogeneous landscapes is a major challenge facing ecologists seeking to understand both watershed and riparian dynamics. The primary objective of this research was to characterize the range and spatial variation of riparian ecotypes throughout Lower Michigan using readily available digital maps and an extensive inventory of riparian forests. Utilizing a terrain-based model of groundwater delivery and regression analysis, I accounted for significant proportions of variance in stream gauge baseflow yields (48--54%, N = 58), summer stream temperatures (23--40%, N = 171), and local groundwater accrual rates (59--65%, N = 48). Using network summaries of groundwater delivery and water table proximity to augment mapped agricultural and wetland information, I demonstrated that accounting for variation in riparian hydrology improved regressions of nutrient export across Lower Michigan (N = 290) between 6--20%. As a proxy for variation in riparian site conditions, I sampled forest composition from 94 locations to describe the range of riparian ecotypes and their spatial distribution. Based on direct and indirect ordinations as well as autecological interpretations of tree species occurrence, I developed a causal hypothesis of soil saturation and flood dynamics using macroclimate, local landscape, and catchment influences on hydrologic character using covariance structure analysis. The causal model exhibited an excellent fit with the observed data (chi2 = 18.11, p = 0.382), accounted for nearly 65% of observed variation in the probability of correct riparian classification, and indicated that distinct hydrologic factors were important determinants of each riparian ecotype. Map-based models explained large portions of natural variation in the compositional and structural character of riparian areas. I therefore conclude that relatively simple, yet process-specific interpretations of existing spatial data sets have enormous potential for predicting complex, yet ecologically relevant hydrologic processes at the land-water interface.Ph.D.Biological SciencesEarth SciencesEcologyEnvironmental scienceHealth and Environmental SciencesHydrologic sciencesUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/131222/2/3057891.pd