Purpose and Scope

The U.S. Geological Survey (USGS) study summarized in this report investigated the hydraulic effects of natural beaver dams and ponds during 2016–17 along two beaver-affected urban stream reaches in the Tualatin River Basin of northwestern Oregon (fig. 1). The two stream reaches were Fanno Creek at Greenway Park and Bronson Creek between Kaiser and Saltzman Roads (henceforth referred to as “Fanno and Bronson Creek study reaches”). The purpose of this study is to quantify the affect that beaver dams in these reaches have on hydrograph dynamics during storm events as well as channel hydraulics across a range of stream flows. Three related studies document current beaver-dam distributions and model the stream network’s capacity to support beaver dams in the Tualatin River Basin (White and others, 2025), investigate changes in water quality (Smith and others, 2025), and sediment transport and deposition (Doyle and others, 2025) along the Fanno and Bronson Creek study reaches. The overall purpose of this series of studies was to provide Clean Water Services of Washington County (the primary wastewater and stormwater management utility for the urban areas of Washington County, Oregon) with the data and insights needed to understand beaver-dam distributions and the physical and water-quality effects of beaver dams and ponds on urban streams in the Tualatin River Basin to inform beaver related restoration and management across the basin.

The goal of this study was to assess the effects of beaver dams and ponds on storm flow hydrographs and hydraulics along two urbanized and low-gradient reaches containing multiple beaver dams. To accomplish the goal, the study developed two-dimensional steady and unsteady hydraulic models to examine the effects of beaver dams on (1) storm flow dynamics for three different storm flow hydrographs with and without beaver dams and (2) stream hydraulics by modeling identical stream reaches with and without beaver dams. The effects of beaver dams in modifying storm flow hydrographs through attenuation and water storage were analyzed using two approaches: (1) a comparison of hydrographs to assess differences in travel time, streamflow magnitude, flow recession, and flow duration, and (2) a comparison of cumulative flow to assess differences in stored water volumes during the three storm events. To evaluate the role of beaver dams in altering hydraulic conditions in channel and floodplain areas, steady-state models were developed to simulate the study reaches with and without beaver dams for four streamflow magnitudes, ranging from base flow to relatively large storm flows. Hydraulic results from these steady-state simulations were output as rasters and analyzed using geographic information system (GIS) tools to examine differences in inundated area and water depth and velocity. A summary of the modeling approaches and objectives can be found in table 1, and the combined results from this study provide an initial quantitative framework to assess hydrologic and hydraulic changes to urban streams in the Tualatin River Basin that may result from beaver dams, including:

Tualatin River Basin and Study Reach Descriptions

The Tualatin River Basin drains 1,844 square kilometers (km²) of northwestern Oregon before joining the Willamette River just upstream from Willamette Falls (fig. 1). The Tualatin River headwaters begin in the eastern flanks of the Coast Ranges, which reach a maximum altitude of 1,057 meters (m) above sea level referenced to the North American Vertical Datum of 1988. Missoula Flood deposits blanket the lower Tualatin Valley in fine silt and clay (less than [<] 0.2 millimeters [mm] grain sizes), creating vast areas of low permeability soils (U.S. Department of Agriculture, Natural Resources Conservation Service, 2012). Land use within the basin is generally forested, agricultural, suburban, and urban areas from upstream to downstream, with the Urban Growth Boundary (UGB; fig. 1) restricting the extent of urban development in the lower valley. Beavers are native to the Tualatin River Basin and have been recolonizing forested and urban streams throughout the basin (Smith, 2017) after being nearly extirpated by hunting and trapping in the last 150 years.

The Tualatin River Basin averages 1.1 m of precipitation per year, with considerable differences between the highest elevation reaches and the lower valley (Franczyk and Chang, 2009). Most precipitation falls as rain during winter (Jung and Chang, 2012), and peak flows typically occur between December and February. Summers typically receive little precipitation, and streams tend to exhibit base-flow conditions between June and September.

Fanno Creek Study Reach

The Fanno Creek Basin drains 82 km² and is approximately 89 percent urban, 8 percent forested, and 3 percent other land uses (U.S. Geological Survey, 2016). Doyle and others (2025) and Smith and others (2025) provide more detailed descriptions of the Fanno Creek Basin and subreach modeled in this report. Fanno Creek tends to have stormflows that quickly rise and fall in response to rain events (McCarthy, 2000; Sobieszczyk and others, 2018).

The Fanno Creek study reach has an upstream boundary at the Hall Boulevard bridge and extends about 1 km from north to south through Greenway Park, a popular urban park frequented by bikers, commuters, and nearby residents (fig. 2) in Tigard, Oregon. The reach is bordered to the west by residential development and to the east by the Koll Center Wetlands. Clean Water Services added meanders and large woods to the Fanno Creek channel and planted native vegetation from 2008 to 2012 (Smith and Ory, 2005). Beavers colonized the restored reach and began building dams and lodges in approximately 2012 (Matthew Brennan, Clean Water Services, oral commun., 2016). During this study (2017–19), the reach had three large channel-spanning beaver dams and two smaller beaver dams in off-channel features. The three largest dams were each at least 1-m tall. One of the channel-spanning dams was 30-m long and extended more than 10 meters onto the floodplain, creating a large pond (fig. 2; approximately 7,940 square meters in wetted surface area in base-flow conditions, referred to as “south pond”). Once the dam was built, the south pond flooded the main channel, nearby paths, and riparian areas throughout the year, even during periods of low flows. The north pond and adjacent Koll Center Wetlands are off-channel features not created by beaver dams. Just downstream from the 30-m-long south pond dam, the stream backwaters into a relic channel as a result of the downstream most dam in the reach. There are at least two small stormwater drainages that enter the Fanno Creek reach. These were observed to have flowing water during storm flows, but the flow was not measured.

The Fanno Creek study area reach drains 27 km² (about 32 percent) of the overall Fanno Creek Basin, and its contributing drainage area has a mean basin slope of 5.7 degrees (U.S. Geological Survey, 2016). The study reach has a channel slope of 0.002 (fig. 2). Within the reach, Fanno Creek predominantly occupies a single thread channel flanked by streambanks typically rising 0.1–1 m above the low-flow water surface and is generally incised at least 1 meter deeper than the adjacent floodplain. From Hall Boulevard to the entrance to the south pond, about 800 m downstream, the main channel is narrow (2–3 m) and shaded by riparian vegetation, with a mean depth of 1.5 m during typical low-flow conditions. One channel-spanning beaver dam exists in this upper part of the reach, which backs water into the north pond. There are also two footbridges within this subreach; neither inhibit water flow at low flows, but the upstream-most footbridge is inundated at high flows, causing flow to backwater into the adjacent field. Within the south pond, the submerged main channel remains narrow and deep, whereas the ponded area is generally shallow (0.5 m on average) with limited riparian shade and some dead snags. Downstream from the south pond and to the end of the study reach, the main channel continues to be narrow (2–3 m wide), deep (about 1-m depth at low water), and shaded by riparian vegetation, with one large channel-spanning beaver dam. During some larger high-flow conditions, water flows from the south pond into the Koll Center Wetlands to the southeast (fig. 2). There is a final footbridge below the downstream-most dam; however, the footbridge did not inhibit flow at any observed flows (up to 6.8 cubic meters per second [m³/s]).

Bronson Creek Study Reach

The Bronson Creek Basin drains 12.7 km² of a rapidly urbanizing area. As of 2016, the basin was approximately 50 percent developed land, 23 percent forest and land shrubs, 18.6 percent impervious area, and 1 percent wetland (U.S. Geological Survey, 2016). The Bronson Creek study reach is approximately 700-m long (fig. 3) and drains about 5.6 km² of upland area with a mean basin slope of 8.96 degrees (U.S. Geological Survey, 2016). From 2004 to 2019, Clean Water Services planted vegetation and added large wood to the channel to improve water quality and enhance the stream and floodplain habitats (Smith and Ory, 2005). Beaver dam-building is believed to have begun sometime around 2010 (Matthew Brennan, Clean Water Services, oral commun., 2016).

The upstream end of the model reach is approximately 0.6 km downstream from Laidlaw Road, the nearest road crossing. The reach extends southwest for roughly 0.7 km downstream to a point near the intersection of Northwest Meisner Drive and Northwest Henninger Lane (fig. 3). The reach is bounded by relatively steep valley slopes on either side of the roughly 70-m-wide floodplain. The channel slope in the Bronson Creek study reach is 0.005, as derived from Global Navigation Satellite System (GNSS) stream surveys (White, 2025). Flow through the modeled reach is strongly influenced by three large channel-spanning beaver dams, each of which is greater than 1-m tall. Unlike the Fanno Creek reach, none of these dams extend into the adjacent floodplain. These dams have been present since at least 2015, based on aerial imagery of the site. Three smaller beaver dams also affect flow and water levels to varying degrees. In the upstream-most 150 m of the reach, Fanno Creek occupies a single narrow (<2 m) channel, with flow typically confined within that channel. The first large dam creates a backwatered pond at low flow and diverts water into several channels at high flow. The backwater pond is separated from the main channel by a narrow (<0.5 m) notch with a small sill, which limits water exiting the pond. Approximately 100 m downstream from the first dam is the second large dam, which diverts most of the flow into a 50-m long side channel, which re-enters the main channel approximately 50 m upstream from the downstream-most dam. The main channel between the second and third beaver dam is relatively narrow (2–3 m) and flanked by mostly deciduous trees along the banks. Water levels in ponded reaches (areas immediately upstream from beaver dams) often are near the top of the channel bank during low-flow conditions, and the adjacent floodplains are frequently inundated, even at moderate flows. Most of the floodplain is dominated by dense stands of reed canary grass (Phalaris arundinacea), whereas much of the main channel is bordered by a thin ribbon of deciduous and coniferous trees. The fourth and final large dam in the model reach is approximately 60 m upstream from the downstream reach boundary but does not create side channels at observed flows. One small drainage enters from the southwest approximately halfway down the study reach.

Topographic Inputs

Accurate representations of channel bathymetry and floodplain topography are critical components of hydraulic modeling. Terrain models combining channel bathymetry and floodplain topography for the Fanno and Bronson Creek study reaches were created by collecting bathymetric and topographic survey data in the wetted channel and adjacent floodplain and combining those survey data with light detection and ranging (lidar) data collected in 2014 (Watershed Sciences, Inc., 2014).

Bathymetric and topographic survey data were collected in the Fanno Creek study reach in May 2016 and in the Bronson Creek study reach in March 2017. Surveys were completed during low streamflow conditions. In the Fanno Creek study reach, survey crews used inflatable kayaks to access the deepest parts of the channel; banks and floodplains were accessed on foot. In the Bronson Creek study reach, the channel was often too narrow to accommodate an inflatable kayak and too deep to wade; thus, a Seafloor Systems SonarMite single-beam echosounder was attached to an improvised hydro-board and pulled across the channel from the banks to survey the deepest parts of the channel. Both reaches were surveyed using real-time kinematic global positioning system (RTK GPS; Trimble R8) and Total Station (Trimble S6) surveys to create a point cloud density of approximately 0.35 pts/m² in the stream channel and adjacent floodplain. Horizontal precision of the measurements was typically less than 0.05 m, whereas vertical precision was typically less than 0.1 m, as calculated by back sight checks with total stations and RTK supplied precision metrics. Surveys extended 10–20 m beyond the channel to capture any potential changes since the 2014 lidar data collection. Other survey points included the locations, crest heights, and endpoints of beaver dams. To aid in model calibration, points were collected at the edge of the wetted channel to measure water-surface elevations and inundated extents. Edges of water points were also collected during high-flow surveys (at streamflows 6.8 m³/s and 1.67 m³/s at Fanno and Bronson Creek reaches, respectively) at each reach to calibrate high-flow dynamics of the hydraulic model. In total, 2,437 points were collected at Fanno Creek and 7,155 points were collected at Bronson Creek. The reach at Bronson Creek included several complex side channels, necessitating the greater number of points.

The subsequent point clouds from the field surveys were converted into Triangular Irregular Networks (TIN) using ArcGIS 10.7. Breaklines were added to the TINs to denote locations of slope changes noted during the field surveys, such as the top and bottom of banks. Breaklines force interpolation between two points and prohibit interpolation through the line. The finalized TIN was converted to a rasterized 0.5 square meter (m²) digital elevation model (DEM) using a linear interpolation. The DEM was then combined with the lidar data to create a terrain model (White, 2025), incorporating both the stream channel and adjacent floodplain (fig. 4).

Beaver dams are hydraulically complex, variably porous structures that, unlike earthen dams or other solid features, cannot accurately be represented by a simple topographic mound in a hydraulic model. After evaluating several approaches, the simplest and most effective method was to represent each dam as a staircase-shaped notch in a solid channel-blocking topographic mound (fig. 5). The elevation of the center and deepest notch was lower than the measured dam crest to facilitate some through-flow during low flows. The progressively higher notches to either side were raised 0.05–0.10 m from each previous step. The exact elevation of each notch, as well as associated roughness, was iteratively adjusted in calibration model runs by comparing pool inundation to measured inundation. Edits were made to the terrain model using standard editing tools in ArcGIS 10.7. This method of representing each beaver dam produced a more accurate representation of stage-discharge dynamics compared to a simple topographic mound because more flow passes over and through the dams as flows increase, which would be expected with actual beaver dams.

To simulate the hypothetical scenarios in which the Fanno Creek and Bronson Creek study reaches do not contain beaver dams (figs. 2 and 3), the final TINs from the with-dam scenarios were copied and edited to remove the existing dams. This was accomplished by interpolating a linear channel gradient through the beaver dams, beginning from the depositional mound of sediment just upstream from each dam and extending downstream below the plunge-pool. No edits were made to other geomorphic features that may be different without beaver dams because these differences were assumed to be negligible. TINs where beaver dams were removed were rasterized identically to TINs with beaver dams. Final DEMs without beaver dams are henceforth referred to as the “no dam” DEMs.

Computational Mesh

Two-dimensional hydraulic models that simulate water movement require underlying topography to be spatially discretized into a computational mesh to apply the equations of flow. Final DEMs were imported into the HEC-RAS model framework as terrain files on which this computational mesh was created. HEC-RAS 2D uses an unstructured mesh approach, whereby model computational cells can be of varying size, shape, and orientation. An initial mesh was created by bounding the study extent with rectilinear cell dimensions of 2 × 2 m. A polygon was then manually digitized around the main channel, dams, and pools for a more refined sub-grid of 0.5 × 0.5 m. HEC-RAS leverages sub-grid bathymetry to enable coarser meshes, while still capturing detailed hydraulics. However, sub-grid bathymetry can result in spurious flow through topographic features if meshes are not properly aligned. To help prevent such errors, breaklines were added as necessary to features such as banks and levees. Breaklines force cell alignment and closure, thus ensuring that cell faces, which dictate hydraulic flow-through, have accurate representations in the model. Identical meshes were used for both with-dam and no-dam DEMs on each reach.

Roughness

All surfaces exert friction on flowing water, and the extent and magnitude of this friction are important inputs for hydraulic modeling. This friction is often referred to as “roughness,” and HEC-RAS uses Manning’s n values to simulate roughness. Once the computational mesh was created, Manning’s n roughness values were assigned to it. This was done by manually digitizing polygons around distinct and relatively homogenous features in the channel and floodplains, such as open channel, grassy fields, and dense vegetation. Roughness coefficients for each feature were then assigned initially from modeler experience and common guidelines (Barnes, 1967). Roughness values were then used to calibrate the model and were adjusted iteratively within reasonable ranges to improve model fit. Final calibrated roughness values can be found in table 2. Roughness values were held identical for the with-dam and no-dam model runs, except for roughness values at dam locations. At these locations, roughness in the no-dam scenarios was assigned the same value as the surrounding channel.

Boundary Conditions

Water can only enter and exit the hydraulic models at specified locations, referred to as boundary conditions. Boundary conditions require parametrization to accurately represent flow dynamics at those locations. The upstream boundary conditions for both study reaches rely on streamflow hydrographs, which were obtained from a related study (Doyle and others, 2025) that deployed temporary pressure transducers to measure water stage from May 2016 to November 2017. Refer to Doyle and others (2025) for details of the data collection and the subsequent development of the stream-discharge relations.

The downstream boundary condition for both models was a stage-discharge relation based on the hydraulic normal depth, whereby the model assumes steady uniform flow at the boundary and uses the Manning equation to calculate depth as a function of the energy slope. The energy slope for the normal depth was estimated from the topographic survey data, and a sensitivity analysis was conducted to assess the effect of errors in slope. Because both reaches have large beaver dams less than 50 m upstream from the downstream boundaries, the extent of hydraulic changes caused by changes in normal depth slope was small. In both reaches, the normal depth of the downstream boundary was much less than the height of the nearest upstream dam.

The Koll Wetlands border the southeast extent of the Fanno Creek model reach. At the highest modeled flows, water exits the south pond and enters the wetlands. To account for this outflow, an additional normal depth boundary condition was added, using the local topographic slope connecting the Fanno reach and the wetlands (about 0.0001), as measured from lidar data. Bronson Creek is bounded by relatively steep valley walls, and thus all water flowing through the stream channel exits through the downstream end of the modeled reach.

Continuous stage measurements were taken near the downstream model boundary at both reaches (figs. 2 and 3), and thus a time series of stage could have potentially been used as the downstream boundary condition instead of applying a normal depth boundary condition. However, this approach presented several limitations, including the following:

A normal depth downstream boundary condition allows for substantial flexibility by estimating the rate at which flow exits the model at any location where flow is known to occur. Due to the potential uncertainties associated with a normal depth boundary condition, all time series of modeled discharge at the downstream end of the model were computed at a location 15 m upstream from the boundary, or about four times the channel width.

Ultimately, the primary goal of this analysis was to understand how the same reach responds with and without dams, not necessarily to simulate precise conditions of a given storm event. Thus, there are uncertainties in how much water enters the modeled reaches, but this uncertainty is included into both dam and no-dam scenarios, and the resulting differences are not sensitive to potential errors in streamflow. For example, if the peak streamflow of a modeled storm event was 20 percent higher than what was used as an input into the hydraulic model, the model will consequentially underestimate actual inundation from that event, but the relative difference between the dam and no dam model under this scenario is not sensitive to these differences because inputs to both scenarios are identical.

Model Runs

Two types of streamflow scenarios were simulated utilizing the two sets of terrain models (with-dams and no-dams) to quantify the role of beaver dams in altering hydraulic conditions (table 1). First, three storm events from the measured period of record at each reach were chosen to evaluate if and how beaver dams affect peak flow magnitude and attenuation, velocity distributions, and water storage in the channel and floodplain (fig. 6). These storms vary in the number of peaks, peak flow magnitude, and storm duration. For the Fanno Creek reach, events included the following:

The various magnitude events were chosen independently for each site; thus, the dates of each modeled event are unique. No formal recurrence interval analysis was conducted on these storms because the period of record was short (May 2016 to November 2017). However, a cursory analysis of this limited record suggests the small storm event occurred about 10 times, the moderate storm event occurred 3–5 times, and the large event occurred 2–3 times during this 19-month period of record. An informal analysis of the gage record at USGS sites 14206900 (Fanno Creek at 56th Ave.) and 14206950 (Fanno Creek at Durham) suggests that 2016 and 2017 were both relatively wet winters (U.S. Geological Survey, 2025), with more numerous and larger storms than most years, though neither year or gage reached historical high flows. Each storm event was modeled both with and without dams.

In addition to modeling the three storm events for the two sets of terrain models (with-dams and no-dams), four steady-state hydrographs were developed to support direct comparisons between hydraulic conditions produced by identical discharges between with-dam and no-dam models (table 1). This direct comparison of hydraulic conditions is more challenging when comparing model output from the storm event scenarios because the changes in travel time between with-dam and no-dam conditions and the flashy nature of the hydrographs make comparison of equivalent hydrologic events difficult. Streamflow conditions of base-, low-, medium-, and high-flows were chosen for steady-state flow simulations (table 3). Streamflow values for these steady-state flow simulations were chosen at each reach based on a cursory analysis of the hydrographs spanning the study period. Base flow was identified as the stable summer streamflow, and low-, medium-, and high-flows were identified by categorizing storm events into three generalized bins that spanned measured peak streamflows associated with storm events. The magnitudes of these steady-state simulations are similar to respective peaks of the storm flow events, with the notable exception of at Fanno Creek, where the high-flow steady-state flows were about half the magnitude of the large storm peak flow.

Model Calibration

Hydraulic models require calibration of parameters to align modeled conditions with those measured in the field. Models simulating current with-dam conditions in both reaches were calibrated by comparing modeled inundated extent and water-surface elevation (WSE) to measured values. WSE values and inundated extent across each model reach were surveyed at low flow and during a storm event, as described previously. Where possible, data were collected along both banks; however, dense vegetation in some areas prevented data collection. Models were primarily calibrated by iteratively adjusting local hydraulic roughness coefficients to improve model fit. Final coefficients were constrained so that they would not exceed the typical range for the feature represented, as documented by previous studies and published references (Barnes, 1967). For example, the main channels at Fanno and Bronson Creeks were assigned Manning’s roughness coefficients of 0.04 and 0.05, respectively, which is within the expected range for low-gradient and silty-bottom streams. The topographic representation of dams was then adjusted to ensure that the model simulated flow characteristics over the dam as closely as possible. For example, if the pool behind a dam was significantly smaller than the measured pool, the staircase pattern of the dam was raised to reduce the amount of flow going through the modeled dam. If the inundated area of a modeled pool upstream from a beaver dam was similar to measured conditions at low flow but too large at high flow, the outermost area (shallowest part of the notch) of the dam was lowered to allow more flow-through at higher flows. These determinations were driven by modeler’s judgement. No calibration was performed on the hypothetical no-dam scenarios.

Analyses

A primary goal of this study was to assess the extent to which beaver dams influence hydrograph dynamics during storm flows. Model output of hydrographs and cumulative flows were compared from three measured storm events and terrains representing with-dams and no-dams conditions.

Analyses of streamflow hydrographs can be used to evaluate the rate at which water enters and exits the modeled reach; however, small differences in streamflow can result in large differences in the total volume of water exiting the reach when integrated over time. Hydrographs for the three storm events were compared between with-dams and no-dams to assess differences in flow magnitude, flow duration, and flow recession. Streamflow was evaluated at several locations in each model to compare upstream and downstream hydrographs both within individual model runs, as well as between with-dam and no-dam model runs.

Comparing the total amount of water, or cumulative flow, that entered and exited each reach can be used to determine how beaver dams affect the storage of water throughout the duration of a storm event. Time series were generated whereby cumulative flow was calculated as the integral of streamflow. Time series of cumulative flow exiting each model reach (output flows) were subtracted from cumulative flow entering the model reach (input flows). Under steady-state conditions, where inflows equal outflows, there is effectively no change in transient storage in the model reach. When inflows exceed outflows, such as during storm events, the difference of cumulative flow between inputs and outputs is indicative of an increase in transient water storage provided by the channel and floodplains, including beaver ponds. Comparing input and output cumulative flow facilitates analysis of how much water is stored throughout the model reach at a given time in the analysis.

Outputs of the steady-state models at four flows were compared to make comparisons of depth, velocity, and inundated area between with-dam and no-dam scenarios to assess how beaver dams affect the hydraulic characteristics and area of potential aquatic habitats. Although these variables are interrelated, each are also individually important for characterizing aquatic habitats and ecosystems. Rasters of each parameter were exported from HEC-RAS as a geo-referenced Tagged Image File Format (GeoTIFF) file and analyzed in the R Programming Language (R Core Team, 2024). Water depth and velocity rasters from the with-dam scenario were summarized by assessing the distribution of each variable and then comparing those distributions to results from the no-dam model scenario. Total wetted area was calculated for each inundated area polygon to compare the total amount of water present in each model run.

Model Limitations

The two-dimensional hydraulic models created in this study facilitate comparison of channel hydraulics produced by storm events under with-dam conditions as well under theoretical no-dam conditions where no beaver dams are present along the study reaches. While this approach helps to characterize the specific role of beaver dams in influencing flow dynamics, it has three important limitations. First, the model does not allow for any infiltration of water into the subsurface. Enhanced hyporheic exchange associated with beaver dams has been demonstrated in other studies (Briggs and others, 2013; Majerova and others, 2015). However, the sediment composition at the Fanno Creek and Bronson Creek study reaches consists primarily of silts and clays, indicating the streambed has limited porosity and infiltration capacity, thus minimizing any errors associated with this model limitation. Nonetheless, an unknown but small amount of infiltration likely occurs along each reach, which could lead to small changes in streamflow dynamics. Similarly, because water can only exit the models at specified boundary conditions, it cannot infiltrate into the groundwater or evaporate. Thus, if a cell is wetted during high flows but is hydraulically disconnected or stranded on the floodplain as flows decrease, the model will represent this cell as pooled water. While this pooling occurs along these reaches, it is likely that the model overestimates this abandonment because water on the landscape can utilize flow-paths that are finer than the resolution of the DEMs. This may cause the models to exaggerate how much water is stored on the floodplains and by extension, the potential attenuation of peak flows. The exact amount of this erroneous storage is unknown but is likely small relative to the volume of water flowing through the reach. Second, the Fanno Creek study reach has two small footbridges that may affect channel hydraulics at high flow. These bridges were not included in the model because modeling software did not support two-dimensional bridge infrastructure at the time of model development. Two of the three bridges in Fanno Creek obstruct flow at observed flows of 3.8 m³/s, resulting in greater inundated areas than simulated. While this difference reduces the accuracy of replicating observed conditions during these flows, it also prevents misappropriating temporary storage due to bridges instead of dams. Finally, beavers create broader changes in stream channels and adjacent riparian areas than just building dams in the channel. For example, beavers source dam materials from local vegetation and dig access channels through ponded areas, altering floodplain roughness, and their dams result in an accumulation of sediment in the channel and the floodplain. These and other effects of beaver activity on hydraulics were not evaluated in this study; however, many secondary affects would be limited to where beavers colonize and build dams, not where human-made structures, such as beaver dam analogs, are built.

Model Calibration

Overall, simulated water-surface elevations along both the Fanno Creek and Bronson Creek model reaches compared reasonably well to measured water-surface elevations over a range of streamflows (table 4; figs. 7–10). Patterns of inundation were also well represented by the hydraulic models, though the models generally produced slightly lower inundated area than observed. In both reaches, model residuals were greater at low flow than at high flow (table 4). The median low flow WSE error at Fanno and Bronson Creeks were −0.24 m and −0.12 m, respectively, whereas median high flow WSE errors were −0.11 and 0.05 m, respectively (table 4).

Simulated wetted area in the south pond at Fanno Creek was slightly less than measured conditions at both base and high flow (fig. 8). The shallow slope of the bathymetry of this pond, as well as the dam creating it, makes its volume and surface area sensitive to changes in WSE; thus, a small under representation of WSE can result in relatively large differences in wetted area. The floodplain in the northwest part of the Fanno Creek study reach shows a lower simulated inundated area than was measured at high flow. This is partly attributable to the footbridge downstream from the north pond; this bridge has a lower deck that diverts water onto the floodplain at high flows.

Comparisons of modeled and measured WSEs for the Bronson Creek study reach show close agreement along the main channel and mid-reach side-channel during low flow (fig. 10). Modeled WSE at high water is overestimated compared to surveyed high water marks, where the error mode is about 0.03 m. Due to access difficulties, the inundated areas near the northern off-channel pond and side channel downstream from the second large beaver dam were not measured, which increases the uncertainty of this evaluation of model results under high flow conditions.

Hydrograph Comparison

Comparing input hydrographs with their downstream counterparts for both with-dam and no-dam models facilitates an analysis of differences in travel time, flow magnitude, and storm flow duration and recession (table 5; figs. 11–16). For the smallest modeled storm events, peak streamflow magnitude was reduced slightly by 2.0 percent (−0.027 m³/s) and 4.8 percent (−0.031 m³/s) along the Fanno and Bronson Creek study reaches, respectively (figs. 11 and 14, table 5). However, peak streamflow magnitudes increased slightly with dams for both reaches under moderate and large storms. Fanno Creek peak flows increased by 1.1 percent (0.039 m³/s) and 0.9 percent (0.079 m³/s) at moderate and large flows, respectively (figs. 12 and 13, table 5), whereas Bronson Creek peak flows increased by 4.6 percent and 2.8 percent (0.052 and 0.059 m³/s) (figs. 15 and 16, table 5).

In all modeled scenarios, the presence of beaver dams results in a short delay in the initiation of stormflows as well as a slight elongation of the receding limb of the hydrograph. At Fanno Creek, only the small storm showed a delay in timing of the peak streamflow magnitude as a result of dams; however, this delay was less than 15 minutes. The moderate storm event at Fanno Creek showed nearly identical timing in peak streamflow magnitudes, while the presence of dams during the large storm event caused the peak streamflow magnitude to occur roughly 30 minutes sooner than in the no-dams scenario. Peak streamflows at Bronson Creek were delayed roughly an hour for the small and moderate storm events with dams, although dams resulted in peak flows occurring roughly an hour sooner in the large storm event.

Cumulative Flow Comparison

Comparing the total volume of flow that has entered and exited each reach (henceforth referred to as “cumulative flow”) facilitates a comparison of results across the duration of the storm event. Model results show that reaches with beaver dams store considerably more water than their non-dammed counterparts (figs. 17–23). The magnitude and duration of these changes, however, varies by storm magnitude and reach.

Instantaneous streamflow can be used to track changes to streamflow peak magnitudes; however, differences in the cumulative volume of streamflow period provides a broader measure of beaver dams influence water storage. Cumulative flow between the with-dam and no-dam models shows that beaver dams increase storage of water in the channel and the floodplain during all storms, with the exception of the largest modeled storm at Bronson Creek. Differences in peak flows and volume are summarized in table 5.

Cumulative streamflow for the Fanno Creek study reach shows a consistent trend of water storage by beaver dams throughout all modeled storms. The presence of beaver dams results in considerably more water storage throughout the modeled events, storing 0.22–3.1 percent more water throughout the storm compared to the no-dams scenarios. During peak storm flows, dams resulted in 914–1,249 m³ more transient storage (figs. 17–19; table 5). During moderate and large storms in the with-dams scenario, some water in the floodplain becomes hydraulically disconnected as flows recede, resulting in a net positive amount of transient storage at the end of the modeled period.

Cumulative flow results and analyses from the Bronson Creek study reach are complicated by the presence of a large off-channel pond in the upper and northern part of the reach (figs. 3 and 9). This pond is inundated during base flows in the with-dams scenario but is primarily dry during equivalent flows in the no-dams scenario. The narrow channel connecting this pond to the channel also contains an approximately 0.5 m tall sill, which acts to trap water behind it, once inundated. The pond becomes inundated during all storm events, however, and effectively traps water in the no-dam scenarios, resulting in some hydraulically disconnected water at the end of the storm events and a net positive difference of inflows minus outflows. Because the off-channel pond is primarily dry before storms in the no-dams scenario, more space is available for off-channel and floodplain storage of stormwater. Even though some of that off-channel storage is already filled prior to the arrival of the storm flows in the with-dams scenario, more water enters the pond under the with-dams scenario than under the no-dams scenario, but the same amount of water that enters the pond exits the pond by the end of the storm simulation (fig. 20).

Steady-State Comparison

Comparing depth, velocity, and inundated area between model scenarios with and without beaver dams at a variety of steady-state streamflows facilitates evaluation of the broad hydraulic effects of beaver dams on streams, which ultimately can manifest in changes to aquatic habitats. Model results of Fanno Creek and Bronson Creek study reaches show large differences in stream hydraulics between the with-dam and no-dam models but, as with the results of the storm flow analysis, these differences diminish as flows increase.

The presence of beaver dams results in more than four-times the inundated area at low flow compared to the no-dam scenarios (figs. 24 and 26). However, as flows increase, the inundated area of dammed reaches grows slower than in the no-dam models, resulting in progressively smaller differences between the two conditions. At the highest modeled steady-state streamflow, total inundated area for the with-dam scenario was less than twice as large as in the no-dam scenario.

The differences in inundated areas are evident when comparing distributions of water depths and velocities (figs. 25 and 27) and suggest that differences between with-dam and no-dam model scenarios are greatest at low and moderate flows, but differences diminish as flows increase. Overall, median depths and inundated area are always greater with dams than without, while median velocities are always lower with dams than without. This aligns with results from the hydrograph findings that stream dynamics converge at the highest modeled flows.