COBLDR20_BoulderSloughEastReviewComments March24, 2017

375 East Horsetooth Road, Bldg. 5 Fort Collins, Colorado 80525 Phone: (970) 226-0120 Fax: (970) 226-0121 www.acewater.com

March 24, 2017

Ms. Katie Knapp City of Boulder Public Works Department Utilities Division 1739 Broadway Boulder, CO 80306 RE: Boulder Slough East Floodplain Mapping Comments Response to “Boulder Slough East

Floodplain Mapping Study” (ACE Project No. COBLDR20)

Dear Katie:

Anderson Consulting Engineers (ACE) has reviewed and researched the “Review of Boulder Slough East

Floodplain Mapping Study” by Merrick & Company on March 6, 2017. In order to address these

comments, ACE has prepared this letter to clarify modeling techniques. No mapping or modeling

changes have been made at this time, but any changes requested by Boulder after reviewing these

comment responses will be incorporated into the final hydraulic modeling, flood hazard workmaps, as

well as documented in the final study report.

Comment #1 The numerous split flow paths leaving Boulder Slough were modeled using the 2D flow area technique within HEC-RAS. A 20-foot by 20-foot grid was selected within the 2D flow area. After reviewing the model results, this grid cell size is too large to accurately capture the details of the terrain without additional breaklines. The report documented the modeling used a 20' x 20' grid because the model took a long time to run (4 hours) and the 20' grid did not produce significantly different results than a previously run 50' x 50' grid. A test run by Merrick that included additional break lines along elevated (high ground) terrain features showed different results than the submitted model which impacted the spill flow values between the flow area and Boulder Slough. Figures 1 through 3 show examples of this condition. Please consider utilizing breaklines along terrain features where flow leaves the 2D flow area or provide justification for not utilizing this method. Adding breaklines had some impact on velocity, depth, and flow rates within the test run. For example, at Pearl Circle XS 560, the flow rates associated with the test run with breaklines is 120 cfs instead of 142 cfs. This is a decrease of 15%. Alternatively, please consider using a 2D flow area grid smaller than 20'x20' if utilizing the breaklines is not applicable.

• While a 20-foot grid size was used as the computational cell size, it should be noted that the

computation cells within HEC-RAS 2D modeling do not have flat/uniform bottoms. Each 20-foot

computational cell has an underlying “sub-grid” that represents the underlying terrain. The

attached section of the HEC-RAS 2D Manual (pages 1-3 to 1-5) summarizing the capabilities of

the HEC-RAS 2D modeling system, and how it allows for larger computational cell sizes than

other 2D models, while still adequately representing the underlying terrain.

COBLDR20_BoulderSloughEastReviewComments 2 March 24, 2017

• Since the 2D model was utilized only for determining flow splits, we only added additional

topographic data and breaklines where prominent topographic features separated distinct flow

paths; these included the upstream lateral weir and the BNSF Railroad embankment. While the

splits between individual split flow paths were found to vary between the 20-foot grid analysis

and 50-foot grid analysis, the overall spill between the Slough main channel and the 2D area is

largely dictated by the lateral weir defined along the Slough; therefore very similar total spills

were achieved for both the 20-foot grid and the 50-foot grid analyses (a total of approximately

498 cfs for the 20-foot grid compared to approximately 489 cfs for the 50-foot grid).

• We agree with Merrick & Company that additional terrain features need to be modeled to

correctly calculate water surface elevations. In this case, we focused on adding additional

terrain data as cross sections in the 1D model used to calculate water surface elevations. An

example of how this additional data was added in the 1D model is shown below in response to

Comment 1, Figure 1a-1c.

• Pearl Circle XS 560 is given as an example of an area where a relatively large 15% flow difference

was calculated in Merrick’s test run vs ACE’s 2D modeling due to the addition of breaklines. We

ran a test calculation in the 1D model to determine the impact of a 15% in reduction of flow to

the Pearl Circle SFP from XS 560 downstream on water surface elevations. This change

resulted in water surface elevation differences of less than 0.1 feet, as shown below in Table 1.

Table 1 - Comparison of Base Flood Elevations Along Pearl Circle SFP

Comment #1 - Figures 1a-1c Figure 1a-1c. Results of HEC-RAS 2D Boulder Slough East model near northeast corner of downstream boundary. 1a shows results using a 20' grid. In this case, flow is "leaking" through the high point instead of flowing over the terrain. Figure 1b shows the results when a breakline is added along the top of this ridge. The upstream water surface is now higher and the flow goes over the ridge. Figure lc illustrates the difference in water surface elevation between the two runs.

COBLDR20_BoulderSloughEastReviewComments 3 March 24, 2017

• While a 2D breakline was not used at this location, we agree that it is important to model this constriction; and it is represented in the 1D modeling as Cross Section 148 on the Pearl Circle SFP. ACE’s 1D model calculated a water surface elevation of 5255.8 feet, NAVD at this location in the 1D model, slightly higher than either the ACE 2D model or the Merrick 2D model with added breaklines. Figure 1a&b Response - Comparison Plan View

COBLDR20_BoulderSloughEastReviewComments 4 March 24, 2017

Figure 1c Response - Comparison Section View

Comment #1 - Figure 2 Approximately 70 cfs of the Pearl Circle spill leaves the model and flows into Goose Creek. The spill occurs between Pearl Circle XS 1051 and 771. This spill may be partially caused by the "leaking" described in Figure 1. The decrease in spill flow is reflected in the 1D Pearl Circle flow rates, however this spill is not shown on the Floodplain Workmaps.

• A boundary condition was set along this berm in the 2D model; we noted approximately 63 cfs

leaving the system at this point in our discharge map included as Appendix B.2. Given that the

flow leaving at this location almost immediately enters the Boulder Creek (and Goose Creek)

floodplain, we chose to not add an additional flow path to account for the 63 cfs. An AO1 zone

would be appropriate for helping to tie the detailed floodplain for the Pearl Circle SFP to the

Boulder Creek Floodplain at this location, and will be added to the final floodplain workmap.

COBLDR20_BoulderSloughEastReviewComments 5 March 24, 2017

Figure 2 Response – Downstream Boundary Condition

Figure 2 Response – Discharge Map Appendix B.2 Showing 63 cfs Leaving to Boulder Creek

COBLDR20_BoulderSloughEastReviewComments 6 March 24, 2017

Comment #1 - Figure 3 2D results at downstream end of Walnut Spill Flow Path. The report and floodplain workmaps indicate that the Walnut St spill flow ends in Boulder Creek approximately 800' southwest of the BNRR South Spill Flow path. However, the 2D results show most of the Walnut spill flowing northeast before connecting to Boulder Creek at the same location as the BNRR spill. Additionally, the model does not fully model the ridge separating the floodplain from Boulder Creek.

• The area in question is embedded within the Boulder Creek Floodplain, and was included in the 2D model to facilitate a proper tie in with that flooding source. While it is agreed that an additional split flow path or different flow path alignment would more adequately describe flow northeast through the parking lot, the Boulder Creek floodplain extends generally to the edge of the parking lot, encompassing this area. We also agree that if we were to model the flow entering the Boulder Creek Main channel, we would need to include the ridge shown for the boundary condition. In this case, we were using the average slope of the Boulder Creek overbank to get a normal depth downstream boundary condition for the Walnut SFP and BNRR South SFP; this area is well-embedded within the Boulder Creek Floodplain, therefore we stopped short of modeling flow into the Boulder Creek main channel. A screenshot with a Boulder Creek Floodplain transparency shown on top of the Boulder Slough East Floodplain Workmap is shown below.

Figure 3 Response – Boulder Creek Floodplain Shown as Transparency Over Boulder Slough East Floodplains

COBLDR20_BoulderSloughEastReviewComments 7 March 24, 2017

Comment #2 The terrain data used for the hydraulic analysis was developed from a 2013 LiDAR survey provided by the City of Boulder. From the LiDAR survey a bare-earth DEM was developed. From viewing the contours provided with the floodplain mapping it appears that buildings filtered out in creating the bare earth file were not extracted correctly. Many of the buildings have several feet of elevation change within the building pad which do not appear consistent with the project photographs. Due to the issues with filtering out the buildings the floodplain delineations are shown to only be partially within some of the building footprints. The elevation change within the building footprint may also impact the 2D flow area analysis by misrepresenting the storage effects of the buildings. Please see Figure 4 for an example of this condition. Please consider reprocessing the LiDAR data to more accurately filter the features removed to create the bare-earth DEM. Alternately please consider using blocked obstructions if utilizing a smaller 20 flow area grid as suggested in the previous comment. For the sake of clarity, a discussion on whether the intent of the modeling is to allow for flow through buildings should be added to the study report.

• ACE noticed the same issues with contours and notified the City of Boulder staff; this was

discussed on a teleconference prior to submitting the floodplan report. There are several areas

where the contours within building footprints do not match the bare earth DEM data, but the

DEM itself appears to be accurate. Since the contours were used only for display purposes, this

does not introduce errors in the modeling and mapping. Due to the rather obvious nature of at

least one area of contour errors, it is recommended that the City of Boulder provide updated

contours for the floodplain workmaps, to avoid comments from FEMA reviewers.

• Regarding the bare earth DEM, there do seem to be very small “bumps” represented within

building footprints. This seems to occur where the terrain shows different elevations on various

sides of the same building, and is likely being caused by interpolation across the building area

where the LiDAR data has been filtered out. These small bumps do show up in the

modeling/mapping based on the DEM (they are evident in the HEC-RAS cross sections). In

general, this likely does not present a problem from a modeling perspective, as most of the

active conveyance is located away from buildings. The buildings themselves are modeled using

the City of Boulder’s building outline layer (or in the case of a couple of newer buildings,

digitized from the aerial photo).

• After reviewing available 2D modeling literature, there are two common methods for modeling

buildings; a) using LiDAR data that includes the buildings, or b) using extremely high n-Values

(n=10) in building locations in order to allow water to be shown on the bare earth topography,

while also greatly restricting the actual conveyance allowed through these buildings. We

decided to use the methodology of “bare earth” LiDAR data with extremely high n-values

assigned to the building footprints, since this methodology is more analogous to our typical 1-

dimesional modeling techniques, with ground data being based on bare-earth and buildings

added in as blocked obstructions. This technique also allows use of the same bare-earth

topography for both mapping and modeling purposes. Our methodology for this use of high n-

values to represent the flow obstruction presented by buildings was outlined on page 9 of the

report; additional text will be added to the final report to further clarify this modeling technique

for future reviewers.

COBLDR20_BoulderSloughEastReviewComments 8 March 24, 2017

Comment #2 - Figure 4 The blue floodplain limits to not match up well with the building footprints.

Possible solutions:

1. Re-run the bare earth filtering algorithm to get better building footprint definition. 2. Digitize building sand add artificial elevations so flow is blocked. 3. Allow flow through buildings by digitizing the buildings but creating available pathways for

flow into the buildings, essentially creating a three sided box.

• As mentioned in the response to Comment 2, buildings were modeled using high-n values (n=10) in the 2D model. and as blocked obstructions in the 1D model, based on the City of Boulder’s building footprints layer. The floodplain mapping is then based on the bare earth DEM, which is consistent with methodology used in previous floodplain mapping studies conducted by ACE for the City of Boulder. From a mapping perspective, the small “bumps” in the bare earth DEM within building footprint areas could have a small impact on where the floodplain maps on the “bare earth” within a building footprint, but there are no cases in the study reach where it makes a difference between a building being either in or out of a floodplain.

HEC-RASRiver Analysis System

2D Modeling User's Manual

Version 5.0 February 2016

Approved for Public Release. Distribution Unlimited. CPD-68A

Chapter 1 Introduction

1-3

consider a river that is modeled in 1D with the area behind a levee is modeled in 2D (connected hydraulically with a Lateral Structure). Flow over the levee (Lateral Structure) and/or through any levee breach is computed with a headwater from the 1D river and a tailwater from the 2D flow area to which it is connected. The weir equation is used to compute flow over the levee and through the breach. Each time step the weir equation uses the 1D and the 2D results to compute the flow allowing for accurate accounting of weir submergence, at each time step, as the interior area fills up. Additionally, flow can go back out of the breach (from the 2D area to the 1D reach), once the river stages subside.

5. Unstructured or Structured Computational Meshes. The software was

designed to use unstructured computational meshes, but can also handle structured meshes. A structured mesh is treaded the same as an unstructured mesh, except the software takes advantage of cells that are orthogonal to each other (i.e. this simplifies some of the computations required). This means that computational cells can be triangles, squares, rectangles, or even five and six-sided elements (the model is limited to elements with up to eight sides). The mesh can be a mixture of cell shapes and sizes. The outer boundary of the computational mesh is defined with a polygon. The computational cells that form the outer boundary of the mesh can have very detailed multi-point lines that represent the outer face(s) of each cell. The computational mesh does not need to be orthogonal but if the mesh is orthogonal the numerical discretization is simplified and more efficient.

6. Detailed Hydraulic Table Properties for 2D Computational Cells and Cell

Faces. Within HEC-RAS, computational cells do not have to have a flat bottom, and cell faces/edges do not have to be a straight line, with a single elevation. Instead, each Computational cell and cell face is based on the details of the underlying terrain. This type of model is often referred to in the literature as a “high resolution subgrid model” (Casulli, 2008). The term “subgrid” means it uses the detailed underlying terrain (subgrid) to develop the geometric and hydraulic property tables that represent the cells and the cell faces. HEC-RAS has a 2D flow area pre-processor that processes the cells and cell faces into detailed hydraulic property tables based on the underlying terrain used in the modeling process. For an example, consider a model built from a detailed terrain model (2ft grid-cell resolution) with a computation cell size of 200x200 ft. The 2D flow area pre-processor computes an elevation-volume relationship, based on the detailed terrain data (2ft grid), within each cell. Therefore, a cell can be partially wet with the correct water volume for the given water surface elevation (WSEL) based on the 2ft grid data. Additionally, each computational cell face is

Chapter 1 Introduction

1-4

evaluated similar to a cross section and is pre-processed into detailed hydraulic property tables (elevation versus - wetted perimeter, area, roughness, etc…). The flow moving across the face (between cells) is based on this detailed data. This allows the modeler to use larger computational cells, without losing too much of the details of the underlying terrain that govern the movement of the flow. Additionally, the placement of cell faces along the top of controlling terrain features (roads, high ground, walls, etc…) can further improve the hydraulic calculations using fewer cells overall. The net effect of larger cells is fewer computations, which means much faster run times. An example computational mesh overlaid on detailed terrain is illustrated in Figure 1-1.

Figure 1-1. Unstructured computational mesh with detailed sub-grid terrain data.

Shown in Figure 1-1, is an example computational mesh over terrain data depicted with color shaded elevations. The computational cells are represented by the thick black lines. The cell computational centers are represented by the black dots and are the locations where the water surface elevation is computed for each cell. The elevation-volume relationship for each cell is based on the details of the underlying terrain. Each cell face is a detailed cross section based on the underlying terrain below the line that represents the cell face. This process allows for water to move between cells based on the details of the underlying terrain, as

Chapter 1 Introduction

1-5

it is represented by the cell faces and the volume contained within that cell. Therefore, a small channel that cuts through a cell, and is much smaller than the cell size, is still represented by the cell’s elevation volume relationship, and the hydraulic properties of the cell faces. This means water can run through larger cells, but still be represented with its normal channel properties. An example of a small channel running through much larger grid cells is shown in Figure 1-2. The example shown in Figure 1-2 has several canals that are much smaller than the average cell size used to model the area (cell size was 500 x 500 ft, where the canals are less than 100 ft wide). However, as shown in Figure 1-2, flow is able to travel through the smaller canals based on the canal’s hydraulic properties. Flow remains in the canals until the stage is higher than the bank elevation of the canal, then it spills out into the overbank areas.

Figure 1-2. Example showing the benefits of using the detailed subgrid terrain for the cell and face hydraulic properties.

7. Detailed Flood Mapping and Flood Animations. Mapping of the inundated area, as well as animations of the flooding can be done inside of HEC-RAS using the RAS Mapper features. The mapping of the 2D flow areas is based on the