Rebecca M. Diehl

Simplification of channel routing models results in errors in streamflow predictions and uncertain flood forecasts. Conceptual frameworks have been proposed to prescribe a heterogeneous routing scheme that incorporates greater complexity where necessary, but their implementation is constrained by the need for hydraulic data that is not widely available. To address this limitation, we use reach‐scale classifications to identify the least complex routing configuration appropriate for each reach. Focusing on seven reach types that differ in topographic characteristics hypothesized to influence routing behavior, we demonstrate that dimensionless scaling parameters, which are indicative of the appropriate flood wave approximation, are distinct among reach types. These results suggest that topographically defined reaches, more readily mapped at broad scales than hydraulic processes, can enable operationalization of heterogeneous routing schemes. We demonstrate regional proof of concept and identify the optimal routing configuration in Lake Champlain Basin, VT. Spatially variable routing across a river network strongly affects model outcomes; we observed a 108% difference in reach‐scale flood peak attenuation between the simplest and most complex configurations. By improving representation of flood routing dynamics in broad scale models, this framework enables efficient modeling schemes that represent natural landscape variability, supporting more reliable flood hazard assessments.

Fluvial sediment process domains (a.k.a. sediment regimes) conceptualize rivers as discrete segments with characteristic and variable tendencies to source, move, and store sediment, providing a useful categorization of river-network sediment connectivity with implications for river management. However, reaches are traditionally classified by expert assessment of field-collected data, preventing extensions to large domains. Here, we document that hydraulic properties automatically extracted from reach-averaged cross sections derived from high-resolution Digital Elevation Models (DEMs) encode the geomorphic signature of sediment regimes. We leveraged a Vermont, USA, dataset of field-labeled sediment regimes and used supervised machine learning to classify reaches by DEM-derived features. Random forest classifiers attained cross-validated balanced accuracies of 0.73-0.92 on a coarse-to-fine partitioning of the dataset, distinguishing reaches based on valley confinement, channel incision, and regime characteristics, although classification failed in confined settings. Prediction of sediment regimes from DEMs advances regional-scale assessments of river hazards, habitat, and water quality.

Dissolved oxygen (DO) regulates the dominant biogeochemical processes in floodplains and is an important water quality indicator. However, predicting DO dynamics with data driven methods in floodplains is challenging due to data scarcity, limiting our understanding of the efficacy of floodplain restoration for clean water objectives. This study applies domain adaptation transfer learning (TL) to a long short‐term memory (LSTM) model to generate floodplain DO predictions. First, a LSTM model was trained on a data‐rich river “source domain” and then used to predict floodplain DO. The trained river model was used to initialize a new TL LSTM model which was finetuned to the floodplain “target domain,” where the same type of monitoring data were scarcer. A third LSTM model was trained only on the floodplain data, and performance was compared across the three models. The TL model outperformed the river model and performed slightly better than the floodplain model (TL model—root mean squared error (RMSE): 2.79; floodplain model—RMSE: 2.90; river model—RMSE: 4.40). Shapley additive explanation (SHAP) values revealed that while the floodplain model relied more heavily on site‐specific attributes, the TL model encoded relationships with dynamic drivers, capturing process‐informed behavior from both river and floodplain domains. Our findings suggest that TL produces models that generalize better across sites and are more robust to variable conditions, offering both predictive skill and process insight. Our modeling framework offers a scalable and interpretable solution for data‐scarce environments, with broad applicability across water resources and Earth system sciences.

Floodplain Geomorphons for the Lake Champlain Basin of Vermont

High Water Marks July 2024 Vermont Floods - UVM Floodplains

Floodplains can significantly impact the routing of flood waves across the landscape, however, their representation in broad‐scale water resource and flood prediction models is limited. To identify hydraulically relevant floodplains at scale, we developed a workflow to automatically extract reach‐averaged topographic features from high resolution (1‐m) LiDAR‐derived topographic data. These features were identified from departures in the relationship between hydraulic geometry and flood stage and hypothesized to define and characterize a zone within the floodplain that disproportionately dissipates energy and attenuates floodwaters, called the Energy Dissipation Zone. We applied the workflow in the topographically diverse Lake Champlain Basin in Vermont, USA, and used a K‐medoids analysis to cluster reaches into distinct feature‐based types that were expected to uniquely route hydrographs. In total, we identified eight clusters of reach types: two that were pre‐sorted because of the presence of a waterbody or limited floodplain access and six that reflected variability in reach‐averaged mesoscale floodplain features that describe the size and shape of the Energy Dissipation Zone. Reach types had distinct impacts on the attenuation of synthetically derived hydrographs, evaluated using the Muskingum‐Cunge method. From these clusters, we propose a Hydraulic Floodplain Classification, which is comparable to other geomorphically defined systems but novel in its focus on the landscape potential to influence flood routing. The automated workflow is repeatable and has the potential to improve the functionality of continental floodplain mapping efforts. Identification of hydraulically effective zones has implications for improved watershed management to meet flood resiliency goals and to improve flood predictions and warnings.

Proper characterization of river flow is essential for the development of structural and non‐structural measures to reduce flood damages, restore ecosystem functions, and manage environmental contaminants in riparian zones. The duration of flood events is an important feature that drives riverine processes and functions such as erosion, geomorphic adjustment, habitat suitability, and nutrient and water quality dynamics. Despite this, most flood characterization methods focus on relating the magnitude of annual‐maximum discharges to frequency, without addressing the duration of flood events. We investigated event‐specific discharge‐duration dynamics at 33 USGS stream gages within the US state of Vermont. Building on the method of Feng et al., 2017, https://doi.org/10.1061/(asce)he.1943‐5584.0001573, flood events from 15‐min discharge timeseries were extracted using an automated threshold method. A statistical model was fit at each gage for both frequency of discharge exceedance and conditional duration of discharge exceedance. This Duration‐Over‐Threshold model estimates the arrival rate of a discharge threshold, q, being exceeded for a given duration, d. Fitted model parameters were compared to basin and channel physiographical characteristics to develop regional regression equations and examine potential watershed processes underlying the duration dynamics. Model parameters summarizing event duration were best predicted by drainage area, mainstem slope, and soil depth/type. The regional regression equations enable design event estimation in ungaged catchments of the study region, which may be used to improve the predictive capacity of hydraulic and ecosystem models, outline a range of potential geomorphic trajectories, or inform emergency management plans and flood damage rating curves.

High Water Marks July and December 2023 Vermont Floods - UVM Floodplains

The National Park Service (NPS) is charged with maintaining natural riverine resources and processes in its parks along the Yampa River and downstream along the Green River. This mission requires information on how proposed water withdrawals would affect resources. We present a methodology that quantifies the impact on natural riverine and riparian features of Dinosaur National Monument based on alternative withdrawals that vary in volume and timing. This methodology uses a reverse quantification and develops tools to enable the NPS to ensure that if withdrawals must occur, the adverse impacts would be minimized by prescribing or constraining the timing, magnitude, and duration of withdrawal. The reverse quantification, well-suited for unregulated rivers such as the Yampa, strives to protect all flows minus extractions from daily flows based on three parameters: 1) a minimum flow, below which water diversion does not occur; 2) the percentage of the flow above the minimum that is diverted; 3) the maximum daily flow that is diverted. We apply 350 flow extraction scenarios, each defined by a unique set of parameters, to the 99 historic annual hydrographs of daily flows (water year (WY) 1922-2020), and to the more recent 20 years (WY 2001-2020). We also consider how hydrologic year type (wet to dry) influences the flow volume extracted and impact to the resource. Recognizing the seasonal differences in flow and ecological and geomorphic response, we divide each year into four distinct seasonal periods and use relations from the literature between flow, channel change, riparian vegetation and fish behavior, physiology, and habitat to define hydrograph and resource metrics used to evaluate impacts to the resource. While our analysis demonstrates that all withdrawals will damage the resource, extractions during the Early Runoff Period (March 15 – April 30) are least detrimental and extractions during the Summer Baseflow Period (July 16 – October 31) are most detrimental. We find that most aspects of the resource are more sensitive to increasing extractions during drier years than during wetter years. Recent decades have seen a shift towards more frequent drier years, resulting in less water in most periods. As a result, our analysis suggests that extractions in recent decades would have had a greater impact on the resource when compared to similar extractions during the full historical record. Finally, we demonstrate how the NPS may use these results to develop limits on extractions for resource protection.