The first step in constructing the 2D hydraulic model was developing a model terrain. The USGS Coastal National Elevation Database (CoNED; Danielson and others, 2018) provides a 1-meter digital elevation model (DEM) for New Jersey and Delaware. An area surrounding the stream channel of the Millstone River approximately 0.8 miles from upstream to downstream at Blackwells Mills, encompassing the area of the weir, was extracted from the DEM and converted to units of survey ft (fig. 5). The cell size of the HEC-RAS terrain was set as 3 ft. This DEM is considered a topobathymetric (shows land above and below water) DEM; however, it incorporated an approximate channel bathymetry for the Millstone River in the study reach that did not reflect true channel depth and volume when verified against field-surveyed cross sections.

A channel was added to the terrain using functionality embedded within HEC-RAS that allows modifying an existing terrain using ground-truthed field data. Five field-surveyed cross-section profiles were obtained in October 2023 by USGS field staff in the area immediately around the Blackwells Mills weir and were used to define an additional terrain composed of just the stream channel bed in the area of the cross-sections. The HEC-RAS Mapper program within HEC-RAS allows a surface to be extracted from point cross-sectional data (interpolating linearly between cross sections) and synthesized into a DEM. This additional terrain was overlaid and merged with the main terrain to create a channel bed that was more representative of bed elevations and depth. It was outside of the scope of this report to update the channel bathymetry for the entire 0.8-mile reach, and the limited cross sections surveyed only provided updated bathymetry from about 400 ft upstream of the weir and about 300 ft downstream. Elsewhere throughout the model, the terrain was adjusted with an approximated channel using the “modify terrain” functionality within HEC-RAS. This tool allows the user to “excavate” a channel using a predefined longitudinal-profile line and selecting a channel depth and bank slope (U.S. Army Corps of Engineers, 2025b). Applied from the area downstream of the last-surveyed cross section to the downstream model extents, the tool lowered the channel bed to an elevation equal to or less than that of the surveyed elevations within the cross sections (fig. 6). The channel parameters, such as bank slope and channel width, were selected to match known geometries from the surveyed cross sections. Inclusion of a full-depth channel within the 2D model allows simulations to better represent streamflow conditions and for the model to correctly represent the tailwater (water downstream) condition. An additional terrain modification was necessary to correctly represent the weir sill and anti-scour skirt and is described in more detail further in the report.

Following terrain development, a model mesh was implemented within a 2D flow area that began in Millstone Township, approximately 2,400 ft upstream of the Blackwells Mills weir and streamgage location. The downstream end of the 2D flow area terminates approximately 2,400 ft downstream of the Blackwells Mills weir. A base cell size of 12 by 14 ft was chosen for the mesh after initial simulations of the model. This base mesh size was used to optimize computation time, while still providing adequate model resolution. The general relief in the vicinity of Blackwells Mills streamgage tends to be modest, without excessively steep gradients, hills, and ridges. In situations such as this, larger mesh cell sizes can be used with little loss in accuracy, while reducing computational time, whereas for higher gradient and steeper areas, finer cell resolution can be necessary to allow HEC-RAS to compute the hydraulic parameters within the mesh cells.

Breaklines are vector lines inserted into the model mesh that “force” mesh cells to align a side along them during mesh generation. These breaklines were incorporated at all roads throughout the 2D flow area, along the stream channel, and other areas of high ground around the streamgage area (fig. 7). Typically, they are used around terrain features, such as steep banks, ridges, embankments, and roads where sudden interruptions in streamflow may occur and may assist in having the model more accurately represent streamflow in these conditions. Breakline cell size was reduced or adjusted from the standard mesh cell size, where model simulations showed excessive iteration resolving a water surface and associated higher levels of water-surface error. Around the breakline associated with the weir structure, the mesh cell size was reduced to a maximum of 3 ft to allow for additional detail in this area. This mesh definition proved to be stable over the course of a variety of model simulations using the software option for Shallow Water Equations, Eulerian-Lagrangian Method (SWE-ELM) for computation (U.S. Army Corps of Engineers, 2025a). Very few mesh cells had issues resolving water-surface elevations with this mesh definition. The few cells that did have issues across simulations could not resolve water-surface elevations with errors of less than 0.1 ft, but these cells were in areas further from the weir. Several profile lines were delineated within HEC-RAS Mapper around the weir location to allow for more detailed analysis of internal model behavior such as transverse and longitudinal water-surface elevations around the structure. Additionally, reference points that represent the locations of upstream and downstream stage sensors at the streamgage were added to the model geometry to compare with observed values during model calibration.

The weir structure was entered into the 2D model using the 2D connection editor (U.S. Army Corps of Engineers, 2025a). Dimensions for the weir, including the weir transverse profile, longitudinal profile, and the widths and dimensions of abutments, were obtained from a survey conducted by the USGS field staff in October 2023. Field elevations were compared to an as-built survey drawing (Work Progress Administration New Jersey, 1937) to check for any anomalies between the historical record and field observations. The vertical datum for elevations transferred from the plans was adjusted from the National Geodetic Vertical Datum of 1929 (NGVD29) to North American Vertical Datum of 1988 (NAVD88) for comparison to the field-surveyed elevations. Actual surveyed dimensions were used to construct the weir in the 2D connection editor, with the weir crest shape within HEC-RAS being set as “broad-crested.” This was the best application of the binary options of “ogee” and “broad-crested” that HEC-RAS provides, despite the Blackwells Mills weir being a nonstandard weir shape (Haro and others, 2017). The 2D connection editor for weirs is restricted to a single transverse cross section that describes the weir along with a single fixed downstream width and a single fixed weir coefficient (for broad-crested weirs). Due to the limited representation available within HEC-RAS to describe the weir profile, a terrain modification was necessary to correctly model the weir sill and anti-scour skirt. A raised rectangular polygon modification was inserted into the terrain below the width of the weir with an elevation set to surveyed elevations (fig. 8).

During model iteration, it was found that the fixed weir coefficient for the entire range of streamflows for the migration period was insufficient for matching observed water-surface elevations upstream of the weir. A weir coefficient for the 98 ft³/s streamflow regime was not efficient enough at the middle and high streamflow regimes of 251 and 487 ft³/s, respectively. This is likely because at the lower streamflow regime, the water surface at the weir has minimal interaction with the chamfered (angled) abutments. As the upstream water-surface elevation rises, the chamfered weir abutments tend to improve the efficiency of the weir, so the coefficient continues to increase until submergence takes effect. This necessitated having separate model geometries employing different weir coefficients that were used to simulate the 95-, 50-, and 20-percent-exceedance flows.

The alternative weirs for both the V-type modification and trapezoidal designs were implemented by modifying the weir profile within the 2D connection editor to match the design drawings. The transitional area between the original sill dimensions and the lowered sill area downstream of the notch was not modeled, with the sill and anti-scour skirt terrain modification being lowered by 4 in. across its entire width. This was due to difficulty defining the appropriate shape within the HEC-RAS framework. Similarly, the individual diffuser blocks, which are described in more detail in the report, are used within the 3D model and were not included within the 2D model for those same reasons. These modifications were more pertinent to observing the very complex flow depth and velocity-pattern changes from the 3D model, and the resolution of the 2D model would not have supported this level of detail. Iterations of the 3D model showed that inclusion of these blocks had negligible effects on the water-surface elevations on the upstream side of the weir where the 2D model was used for a detailed comparison against the existing stage-discharge relationship. Thus, their absence from the 2D model was considered acceptable.

Using HEC-RAS 2D hydraulic modeling necessitates defining a land-cover feature class to delineate Manning’s n-values for the entire study reach; more on this specific process can be found in U.S. Army Corps of Engineers (2025a). However, the primary difference is that n-values are spatially applied to the model reach and not by cross section, as they would be in a one-dimensional hydraulic model. Usually, a polygon feature class representing various land covers is defined within a geographic information system to provide n-values across the study reach. Each defined land use or land cover of the land cover feature class is assigned a Manning’s n-value approximating the values defined from field investigation or literature. In this case, a land use and land cover (LULC) feature class for the year 2020 was downloaded from the New Jersey Department of Environmental Protection (NJDEP) geographic information system portal (New Jersey Department of Environmental Protection Bureau of Geologic Information Systems, 2024). This feature class, clipped to the HEC-RAS 2D mesh extent, had 13 different LULC categories covering the entire model terrain. It was necessary to assign roughness coefficients (how much resistance a surface offers to flow of water) based on the standard classification of each of the LULC categories. Following guidance in the HEC-RAS 2D Modeling User’s Manual (U.S. Army Corps of Engineers, 2025a), each of the 13 categories was assigned a base Manning’s n-value (table 1). Once the base value is assigned to each individual land use or land cover, the value can be forced to a different value during calibration or can be overridden through application of a specific n-value calibration region, which can be drawn into the model using the HEC-RAS Mapper application (U.S. Army Corps of Engineers, 2025b). To better represent the channel downstream of the weir as observed during field surveys, the base n-value of 0.035 for streams and canals was altered downstream, by use of a calibration region, to a value of 0.040 (fig. 9). This additional roughness is representative of woody debris and vegetation at the channel edges. This value superseded any adjusted values shown in table 1 in the areas covered by the calibration region. The value of 0.035 for the “streams and canals” land use category was raised to a values of 0.045, as the calibration region polygon did not match exactly with the edges of the gridded land-use cover raster. This prevented an unrealistic drop in the roughness near the edges of the calibration region.

The downstream end of the model used normal depth as a boundary condition. Using normal depth as the downstream boundary condition assumes that the channel beyond the downstream model extent is flowing under comparable conditions to the downstream end of the model itself. An estimated or calculated slope of the channel is then entered within the normal depth field in the unsteady flow editor in HEC-RAS. In the case of this model, normal depth was approximated by plotting the furthest downstream cross sections longitudinally and computing the bed slope of the channel. The average slope using this method was computed to be 0.0041 and was input within the normal depth field. The upstream end of the hydraulic model begins approximately 2,400 ft upstream of the Blackwells Mills weir, and it was reasonable to use the computed streamflow hydrograph from the Blackwells Mills streamgage directly as the upstream boundary condition as there are no tributary inflows that would augment the flow computed at the streamgage between the streamgage and the upstream end of the model (U.S. Geological Survey, 2025).

For the initial simulations, the observed published-streamflow hydrograph from October 14 through October 17, 2014, was entered as the upstream boundary condition. This hydrograph was entered at a 15-minute interval from midnight on October 14, 2014, through 11:45am on October 17, 2014. The simulation was run at a 2-second computation interval, which was successful at suppressing iterative model errors and maintaining model stability. Maximum Courant numbers, which are used to evaluate time steps of the model, peaked between 2 and 3 for the mesh cells around the weir and immediately downstream of the weir. This was considered acceptable as the streamflow hydrograph is not highly dynamic and changes relatively slowly over time. HEC-RAS can tolerate high Courant numbers due to the implicit solution scheme (method used to calculate future streamflow) it uses, provided the flood wave (depth and velocity) being modeled is not changing rapidly (U.S. Army Corps of Engineers, 2025a).

As calibration efforts for the model continued and the previously described weir coefficient issue was revealed, a different approach to running the model was necessary. Because the weir coefficient did not change during the simulation, what the authors describe as “quasi-steady” flow simulations had to be iterated through the hydraulic model. This was accomplished within the unsteady flow data editor in HEC-RAS by implementing a 24-hour static hydrograph at a 15-minute interval at the 95-, 50-, and 20-percent-exceedance streamflow regimes of 98, 251, and 487 ft³/s (U.S. Army Corps of Engineers, 2025a). This method permitted the use of different weir coefficients at these streamflow rates.

The initial simulations from the HEC-RAS 2D hydraulic model were set to utilize the streamflow hydrograph that aligned with October 2014 when a temporary downstream water-level logger was installed at the Blackwells Mills weir. The logger provided the ability to use the approved streamflow hydrograph as a boundary condition, with the hydraulic model having the ability to freely compute upstream and downstream water-surface elevations that were able to be compared to observed values on both sides of the weir. Although this iteration of the model reproduced generally accurate upstream and downstream water-surface elevations when compared to observed values, the simulated upstream water surface tended to rise between approximately 0.1 and 0.3 ft higher than what was measured when the flow was above about 200 ft³/s (fig. 10). The rise noted at higher streamflows was assumed to be a result of the fixed-weir coefficient for the broad-crested weir setting in HEC-RAS. The model was unable to adjust the weir coefficient as the depth of flow above the weir increased. Also, the chamfered abutment walls increased the contraction efficiency of the section, which was assumed to be the primary cause of simulated water-surfaces not precisely matching observed elevations.

To accurately calibrate the model simulation to the existing stage-discharge relationship at the Blackwells Mills streamgage, it was necessary to run the 95-, 50-, and 20-percent-exceedance streamflows as target streamflows in the previously described quasi-steady flow manner (refer to the “Boundary Conditions and Model Simulations” section of the report). This provided a mechanism where a specific weir coefficient for each target streamflow could be utilized to precisely relate the model to the existing conditions at the streamgage. This provided the most accurate starting point for employing and evaluating the alternative weir designs within the 2D hydraulic model framework. Applying the weir coefficients in the manner described in the “Boundary Conditions and Model Simulations” section of the report, the simulated upstream water-surface elevations matched the rating-derived water-surface elevations associated with the targeted streamflow regimes within 0.00 ft (table 2). To quantify how the addition of the alternative weir designs would affect streamflows lower than the 95-percent-exceedance streamflow of 98 ft³/s, two additional model simulations were performed at streamflows of 10 ft³/s and 50 ft³/s. These simulations used the same weir coefficient as the 98 ft³/s simulation and remained in agreement with the rating-derived values. Figure 11 displays the stage-discharge relationship generated from the 2D HEC-RAS simulation-overlaid values obtained from stage-discharge relationship (rating) number 31.0 at the Blackwells Mills streamgage, which was in use at the time of publication of this report. Figure 11 also demonstrates that the model calibration was excellent in the three target-streamflow regimes, with only minor differences of 0.03 ft in water-surface elevations at extreme low streamflows of 10 ft³/s. A simulated difference of 0.03 ft at an extremely low streamflow of 10 ft³/s was considered acceptable for the simulation. The difference at this low streamflow can be affected by minor construction imperfections in the weir crest shape or minor variations in surface-area roughness that are not explicitly represented in the computer-simulated weir design. Additional context for the rarity of extreme low flows is represented by the historic period-of-record low flow of 5.0 ft³/s being set more than 100 years ago in 1923 (U.S. Geological Survey, 2025).

To supplement the 2D hydraulic model’s comparison against the observed surface-water-elevation values at the Blackwells Mills streamgage, a comparison of modeled velocities downstream of the weir versus velocities obtained from discrete streamflow measurements was performed as a further check on the model’s ability to correctly represent field conditions. During periods of low flows, streamflow measurements were performed by wading with an acoustic Doppler velocimeter (ADV) between 100 and 150 ft downstream of the Blackwells Mills weir. A tagline was strung perpendicular to streamflow across the stream and the profile was divided into 25–30 individual subsections, where point velocities were collected for 40 seconds. Within the measurement data, these individual subsection velocities can be viewed from where they were collected in the measurement profile. Additional information on the application of USGS procedures for making streamflow measurements can be found in Turnipseed and Sauer (2010).

Two measurements that noted clear channel conditions with the weir clear of debris that had flows near the streamflow rates for each of the 95- and 50-percent exceedance intervals were used to compare against model-simulated velocities. A comparison against the 20-percent exceedance interval could not be performed, as streamflow measurements were performed upstream of the weir for these moderate flows and thus could not be effectively compared against velocities downstream of the weir. The subsection-mean velocities from the streamflow measurements were plotted across the stream against simulated velocities from the 2D hydraulic model simulation run at corresponding streamflows (fig. 12). The simulated velocities generally plotted higher against the observed velocities, with some areas of the cross section in close agreement. Variation at the edges and within the profile likely exists due to irregularities between the underlying DEM and the actual channel, because the depths in this area of the model were interpolated between surveyed cross sections as opposed to directly measured depth observations within the streamflow measurements. The actual position of the profile line used to extract the velocity data from the simulation results was inserted into the model mesh 150 ft downstream of the modeled location of the Blackwells Mills weir. The measurement locations tend to be rough approximations of actual cross-section positions in relation to the streamgage versus an exact measurement, so the location of these data extracted from the model simulation may also serve to create variation in the comparison. Overall, the simulated velocity profiles show a reasonable match against field-obtained values, with the simulation marginally overpredicting stream velocities.

A sensitivity analysis for Manning’s n-values and the downstream boundary condition was completed to identify possible areas of concern with the model’s precision. Errors in the selection of Manning’s n-values for the land-cover feature class associated with the 2D-model geometry could result in the simulation incorrectly computing water-surface elevations for a given streamflow. In addition, errors in the selection of a normal depth value for the downstream boundary condition could result in variation in the simulated water-surface elevation, although the errors would typically become smaller further from the location of the downstream model boundary.

Adjustment of the downstream boundary condition was accomplished by raising and lowering the normal depth value incrementally by 10 percent within the unsteady flow data editor (U.S. Army Corps of Engineers, 2025a). Adjustment of these values yielded very little change in overall water-surface elevations within the area of the Blackwells Mills weir (fig. 13). This demonstrates that the downstream boundary condition was not sensitive to this parameter and any errors in the estimation of this boundary condition likely only present negligible effects on the final model results. Some differences would be expected to be viewed in the areas close to the downstream model boundary.

Adjustment of Manning’s n-values was accomplished by raising and lowering the values incrementally by 10 percent within the land-cover layer in HEC-RAS. Adjustment of these values yielded very little change in overall water-surface elevations (fig. 14). Differences of less than 0.1 ft in maximum-simulated water-surface elevations showed that the model was not particularly sensitive at the Blackwells Mills streamgage location to n-value selection and application. Larger differences would be expected in the areas downstream of the weir, as the weir dimensions and geometry have a much larger effect on water-surface elevations immediately upstream of the weir versus the channel-roughness coefficients.

A computational fluid-dynamics model of the weir was developed using FLOW-3D HYDRO to evaluate passage conditions across a range of weir configurations and streamflows. Simulations were conducted for three weir designs: the existing weir, the modified V-type weir, and the modified trapezoidal-type weir. Each configuration was simulated under three streamflow scenarios representing the 95-, 50-, and 20-percent-exceedance streamflows (98, 251, and 487 ft³/s, respectively) during the migration season, resulting in a total of nine simulations.

Model geometry was constructed from topographic and bathymetric data compiled from multiple sources. Bathymetric survey data collected by the USGS in October 2023 (Niemoczynski and others, 2026) included channel cross sections upstream, downstream, and across the weir structure. To supplement gaps along the floodplain and embankments, Quality level 2 light detection and ranging (lidar) data (U.S. Geological Survey, 2022) were incorporated. All elevation data were imported into CloudCompare version 2.13.1 (CloudCompare, 2024), aligned to the NAVD88 and North American Datum of 1983 datums, merged into a single point cloud (3D dataset), and interpolated into a 5 by 5-ft raster surface. This raster served as the terrain input for the FLOW-3D HYDRO model.

The weir structures were modeled in OnShape version 1.196 (OnShape, 2025) using 2023 field-survey data (fig. 15A, B, C). The files (stereolithography format [file extension .stl]) representing each of the three configurations were exported from OnShape and imported into FLOW-3D HYDRO. The computational domain extended approximately 30 ft upstream and 50 ft downstream of the weir. A subcritical inflow boundary condition was applied upstream, and a fixed water-surface elevation was defined at the downstream boundary using results from the 2D HEC-RAS model. All simulations used a uniform mesh with 0.125-ft grid spacing, the Re-Normalisation Group method turbulence model (Flow Science, Inc., 2026a), and a first-order momentum advection scheme (Flow Science, Inc., 2026b). Surface-roughness height was set to 0.1 ft for natural bathymetry and 0.02 ft for the weir structures.

A calibration simulation was performed using the existing weir geometry to confirm that the FLOW-3D HYDRO model produced streamflow estimates consistent with measured data. This test run was configured with fixed headwater and tailwater elevations based on the 2D HEC-RAS model output at the upstream and downstream boundary locations. The model-calculated streamflow rate was 252.4 ft³/s, compared to a measured streamgage value of 251 ft³/s at the site. The simulated discharge was within 1 percent of the observed value, which was considered an acceptable level of agreement. Following this verification, the project team proceeded with the full set of nine design simulations.

Due to a numerical instability identified in FLOW-3D HYDRO version 2024R1 that affected model performance at low flows, the 95-percent-exceedance streamflow simulations were run using version 2025R1 (Flow Science, Inc., 2024, 2025). The remaining six simulations, representing the 50- and 20-percent-exceedance streamflow conditions, were completed using version 2024R1 prior to the discovery of the issue. Because each simulation required between 3 and 6 days to complete, the higher streamflow scenarios were not rerun in version 2025R1. All simulations were executed on a high-performance workstation using a 24-core FLOW-3D HYDRO license. Results were postprocessed in FLOW-3D POST version 2024R1 and 2025R1 (Flow Science, Inc., 2024, 2025). Model outputs included velocity fields, water-surface elevations, and flow depths for all nine simulations. These outputs were used to evaluate weir performance and to assess the potential for improving fish passage at the site and are available within the data release associated with this report (Niemoczynski and others, 2026).

After calibrating the computer models to the existing weir design, the weir design in both models was changed to match the modified V-type design and modified trapezoidal designs described earlier in this report. The weir designed with a modified V-type shape in the center of the span was input into the computer models, and simulations were run with flows of 98, 251, and 487 ft³/s corresponding to the 95-, 50-, and 20-percent exceedance flows. The V-type center weir crest design was limited to 4 ft wide but extended almost to the bottom of the existing weir in the center, stopping only 2 in. above the anti-scour pad skirt elevation. The weir design with the trapezoidal-shaped opening in the center of the weir crest span was limited to 3 ft wide and also extended almost to the bottom of the existing weir, stopping only 2 in. above the anti-scour pad skirt. For the 2D model, the weir design and modifications were an approximation of the more detailed design that was able to be employed in the 3D model due to constraints encountered within the HEC-RAS software. Both designs generated similar upstream water-surface-elevation results in comparison to the existing conditions. Table 3 shows the changes in upstream water-surface elevation for the 2D model. At the median streamflow of 251 ft³/s, the simulated upstream gage pool only dropped about 0.02 to 0.03 ft because of the changes to the opening. Simulations of the 95-percent-exceedance streamflow of 98 ft³/s simulated an upstream gage-pool elevation drop of about 0.03 to 0.04 ft using either the 2D or 3D models. At the higher flows of 487 ft³/s, both computer models simulated a 0.01 to 0.02 ft drop in the gage-pool elevation.

It should be noted that the changes in the gage pool were evaluated because the downstream water-surface elevation is not controlled by the weir and, therefore, would theoretically not change with minor modifications to the weir crest shape.

When considering alternative weir designs, one primary focus of the design selection was to minimize degradation of the existing stage-discharge relationship at the Blackwells Mills streamgage while also preserving the sensitivity of the streamgaging weir. Sensitivity of a hydraulic control can be described as the relative change in streamflow per minimum unit of an independent variable, in this case, stream elevation or stage. Similar to many USGS streamgages, the Blackwells Mills streamgage can accurately measure stream stage to 0.01 ft (Buchanan and Somers, 1982). The present shape and condition of the weir allows each unit of stage (0.01 ft) to compute an average streamflow change of 3.5 ft³/s from flows of 10 to 487 ft³/s. Looking specifically at the lower streamflow regime, the weir can compute an average streamflow change of about 1.6 ft³/s from flows of 10 to 98 ft³/s. After implementing the alternative weir designs within the HEC-RAS 2D model, the changes in hydraulic-control sensitivity were evaluated. Table 4 shows the existing weir sensitivity and the changes in sensitivity that would be expected with the modified V-type and trapezoidal weir modifications. From 10 to 487 ft³/s, the average change in streamflow per unit of stage decreased, showing that the weir modifications marginally improved sensitivity over this range. Moving into the low streamflow regime from 10 to 98 ft³/s, the weir modifications show a greater improvement to control sensitivity with about a 10-percent reduction in average change in streamflow per unit of stage. This demonstrates that the alternative weir designs would be just as effective at computing streamflow as the existing weir, with potential improvement in the range of typical migration season streamflows.

In addition to evaluating control sensitivity, the 2D model was used to evaluate the overall change to the stage-discharge relationship, particularly where the weir modifications would no longer affect the existing relationship (fig. 16). As previously stated, the difference in stream stage between the existing weir and both alternative designs at a streamflow of 487 ft³/s was about 0.01 to 0.02 ft. Transposing this difference in stage to predicted streamflow, at 487 ft³/s, the alternative designs would be underpredicting streamflow by between 1.2 and 2.5 percent using the currently existing stage-discharge relationship. As streamflow increases beyond 487 ft³/s, the differences will continue to diminish as the weir quickly becomes submerged at these streamflows. This demonstrates that the weir modifications would negligibly affect areas of the established stage-discharge relationship above the 20-percent-exceedance interval migration flow.

The model simulation results also provided information about reduced velocity conditions along the crest of the weir and the downstream side. For the existing weir, the model-simulated velocities along the crest generally ranged from about 5 to 7–8 ft/s, but on the downstream side along the top of the weir skirt, the simulated velocity was in the range of about 10 ft/s for median and lower streamflow conditions—similar to what was estimated in the Haro and others (2017) study. American shad can overcome velocities of 10 ft/s over short distances, with alewife and blueback herring being capable of sprint speeds of 8 ft/s (Haro and others, 2017). One of the goals for the weir modifications was to reduce the water velocity across the skirt along with providing increased depth. The V-type and trapezoidal modifications with a lowered downstream skirt and individual diffuser blocks increased the depth at the skirt and reduced the average velocity across the skirt and sill downstream of the modified areas to about 7 to 8 ft/s. The maximum-simulated velocities in the model outputs in the area of the modification were about 8.5 ft/s in the area of the sill, versus maximum-simulated velocities exceeding 10 ft/s without the weir modifications. Another potential benefit of the weir modifications realized from the modeling was the disturbance of the more unnatural shallow straight-line flows (figs. 24 and 25) that typically exist without the modifications. These flow disturbances, shown in figure 26, could act as attraction flows for the fish species noted within this study. Many dams evaluated for fish passage are typically much higher and may require more complex fish ladders or ramped designs than the Blackwells Mills weir, which is only about 14 in. high in the center section from the crest to the lowest part of the skirt. Several discussions with NJDEP focused on design ideas to balance streamflow accuracy and sensitivity along with improved passage conditions near the current center of flow led to the approach of utilizing a modified opening design in the center of the weir as the most balanced approach.