From land to stream : an assessment of watershed-scale biogeochemical interactions at the stream-groundwater interface

The stream-groundwater interface (SGI) is typically studied at scales <1000 m, whereas watershed management needs to understand outcomes of stream-groundwater interactions at scales of tens of kilometers. As a ubiquitous reactive ecotone, the SGI plays a critical role in biogeochemical cycling across stream networks. Process-based models have examined these needed larger scales, but there is a distinct absence of field data to validate modeling efforts. Due to the paucity of these data, this dissertation sets out to be among the first efforts to sample across the SGI at the watershed scale. The three primary goals are to 1) identify the grain of measurements needed to assess the SGI across a stream network, 2) determine whether landscape biogeochemical signals are modified by the SGI at the watershed scale, and finally 3) to investigate whether the SGI acts consistently to modify biogeochemical inputs from the landscape. This work was done through the lens of common solutes found in streams, specifically focusing on dissolved organic carbon, an important driver of many stream biogeochemical reactions. In Chapter 1, I evaluate how to sample the SGI across a stream network. I test two fundamental sampling schemes, focusing on local heterogeneity (i.e., features or plots) compared to longitudinal heterogeneity (i.e., stream reaches). There was previously no clear guidance as to which kind of sampling scheme would be most appropriate. This was necessary because sampling in the SGI is time and labor consuming, and one must determine how to distribute a finite number of sampling points. These data were collected in two synoptic sampling campaigns in a third-order stream network in southwest Michigan. Here, I found that longitudinal sampling accounted for similar stream network variance as localized heterogeneity. Therefore, it may be useful to focus on longitudinal sampling as local sampling becomes redundant.In Chapters 2 and 3, I investigate, first, how different watershed delineations are used to understand landscape contributions to the biogeochemical signal of the stream water. I compared surface watershed and novel groundwatershed delineations to evaluate which areal delineation of the landscape would best predict stream biogeochemistry. Both delineations were then used to investigate how the biogeochemical signal propagated from the land into the SGI, and whether this signal was modified as it entered the SGI by way of spatially lagged linear models. I found that both watersheds were comparable, and therefore the groundwatersheds may be appropriate for lowland watersheds, with strongly upwelling groundwater. Further, I found that the landscape signal found in surface waters through linear modeling was modified as models were propagated into the SGI, given the decreasing performance of linear models in the stream subsurface.In Chapter 4, I evaluated the SGI effects from multiple watersheds on various sources of dissolved organic carbon and its molecular components. I used mixed-effects linear models to test if there was a consistent modification of dissolved organic carbon across a multitude of SGIs as compared to stream water alone. I found that for most optical properties tested, the interaction between the specific carbon source and the SGI sediments was important, functionally obscuring the effects of the sediment alone. The one exception to this was a proxy for humic substances called Peak T, for which the SGI sediment had a significant, identifiable effect. These results indicate that it may be difficult to make broad generalizations about the function of SGIs, where local heterogeneity might be an important consideration.