Date of Award
Doctor of Philosophy (PhD)
The water temperature of global river networks (also referred to as ‘stream temperature’ or ‘river temperature’) is an influential control on numerous aspects of water quality and riverine ecology, impacting rates of solute processing, dissolved oxygen content, and habitat viability for aquatic ecosystems. River water temperatures arise from the complex interplay of hydrological processes, atmospheric forcings, anthropogenic disturbances, making river thermal regimes challenging to understand and predict at the reach, regional, and global scale. In the absence of widespread water temperature observations, models are commonly used to simulate aspects of water temperature variability by integrating the influence of basin-specific controls and heat fluxes into and out of river systems. In addition to their role as a critical water quality parameter, water temperatures can also be leveraged as a practical tool to probe hydrologic interactions between stream channels and the underlying subsurface. This dissertation explores three diverse applications of water temperature modeling: 1) tracing groundwater-surface water interactions around stream restoration structures using water temperature observations; 2) leveraging machine learning to infer continental-scale drivers of river thermal behavior; and 3) predicting water temperatures at high spatial and temporal resolutions with coupled temperature-hydrologic models.The first chapter of this dissertation uses water temperature heat tracing methods, in combination with other field observations, to characterize hyporheic exchange induced by beaver dam analogue restoration structures. Beaver dam analogues are process-based restoration structures designed to mimic the effects of natural beaver dams and stabilize degraded and incised river reaches. Despite their frequent application, the influence of these structures on groundwater-surface water hydrology remains unclear. Vertical heat tracing, measurements of hydraulic head, and analyses of porewater biogeochemistry were used to investigate hydrologic behavior associated with three beaver dam analogues installed on Red Canyon Creek, WY, USA. These analyses demonstrated that while the restoration structures had a negligible effect on overall stream chemistry, beaver dam analogues were capable of producing heterogeneous and localized regions of hyporheic exchange. These results highlight the effectiveness of using water temperatures to trace vertical heat flow and related groundwater-surface water interactions in tandem with other field-based observations. Given the demonstrated impacts of water temperatures on river water quality, it is critical to better understand how the dominant controls on river thermal regimes vary in time and across broad spatial scales in order to design more effective watershed management strategies. Machine learning models are well suited to this objective, as they can generate accurate predictions of environmental processes while revealing key interactions between variables in large datasets. In the second chapter of this dissertation, a suite of random forest models was used to predict metrics of river temperature variability across the conterminous US using watershed characteristics extracted from a publicly-available dataset. Variable importance metrics were then interpreted to infer the underlying controls on river temperatures. Regional climate forcings tended to most closely control river temperature magnitude, though those forcings were mediated by the influence of hydrological processes, watershed characteristics, and anthropogenic disturbances. Results from the random forest models underscored the challenge in predicting aspects of water temperature variability at continental scales, particularly when river thermal regimes are disrupted by dams and reservoirs. The presented machine learning approach to river temperature prediction illustrates how large environmental datasets can be leveraged to provide discerning insight into the drivers of hydrologic and thermal processes. To supplement predictions of water temperatures at point locations along the river network, deterministic energy balance models are often applied to provide spatially distributed and temporally continuous water temperature simulations. Deterministic water temperature models function by quantifying radiative, turbulent, and advective heat fluxes into and out of a river at the air-water and water-streambed interfaces. While such water temperature models are often applied within single catchments, many watershed management applications require high resolution predictions of temperatures at a broader spatial extent. The third chapter of this dissertation focuses on the development of a coupled hydrological-water temperature energy balance model in a single test basin with the potential for expansion to the full conterminous US. Using forcings and outputs from the National Water Model, a continental-scale hydrologic model implemented by NOAA and NCAR, several water temperature model configurations of increasing complexity were tested to evaluate tradeoffs between performance and computational efficiency. Modeling efforts demonstrated that the National Water Model can be effectively leveraged to provide high-quality predictions of hourly water temperatures throughout a river network, though critical challenges remain in expanding coupled water temperature models to continental scales.
Wade, Jeffrey, "Stream Temperature as a Tracer of Interactions Amongst Hydrological Processes, Atmospheric Exchange, and Human Activity" (2023). Dissertations - ALL. 1709.