HEC RAS 2D Methods Guidance: South Dakota Large Scale Automated Engineering

Transcription

1 HEC RAS 2D Methods Guidance: South Dakota Large Scale Automated Engineering January, 2017 Prepared by: Compass PTS JV a JV led by AECOM and CDM Smith 3101 Wilson Boulevard, Suite 900 Arlington, VA 22201

2 3101 Wilson Boulevard, Suite 900 Arlington, VA T: ; F: DOCUMENT HISTORY REVISION HISTORY Version Number Version Date Summary Changes Team/Author 0.1 April 22, 2016 Draft version for Compass PTS JV Review Compass PTS JV 0.2 May 11, 2016 Compass PTS JV Comments Incorporated Compass PTS JV 0.3 July 5, August 12, APPROVALS September 30, 2016 Extended Content, Clarifications, and Revisions Extended Content, Clarifications, and Revisions Extended Content Compass PTS JV Compass PTS JV Compass PTS JV 0.6 January 2017 Extended Content Compass PTS JV This document requires the approval of the following persons: Role Name Phone Title (CLIN/RMC) Review Date Approved Date Project Manager Brandon Banks Project Manager April 21, 2016 Regional Technical Coordinator Brian Murphy Principal Engineer April 22, 2016 CLIENT DISTRIBUTION Name Title/Organization Location Ryan Pietramali Risk Analysis Branch Chief, FEMA Region VIII Denver, Colorado Dawn Gladwell FEMA Region VIII Denver, Colorado Brooke Conner FEMA Region VIII Denver, Colorado Page i

3 Table of Contents 01 Introduction D Modeling Advantages and Considerations Data Sources Automated Engineering Approach Inputs Terrain Roughness Boundary Conditions Model Controls Mesh Area Cell Size Selection Time Step Selection Equation Selection and Other Computation Settings Internal 2D Mesh Connections Reasonability and Verification Checks for the 1% Annual Chance Event Multi-Frequency Analyses % Plus and Minus Analysis Model and Mapping Outputs Regulatory Approach Modeling Considerations Mapping Post-Processing CNMS Zone A Validation Deliverables Page ii

4 List of Tables Table 1: Typical Landuse-Soils-CN Matrix Table 2: Point Depth Calculation Example List of Figures Figure 1: Hydraulic Computations in RAS 5 Utilize the Underlying, Higher Resolution Terrain Data Figure 2: Typical Scenario for Applying Inflow Contributions from Large Drainage Areas Outside RAS5 2D Study Limits Figure 3: TP-40 Figure 15 Area-depth Curves Figure 4: Computation Log File Example Figure 5: Plot of ARC I, ARC 1.5, and ARC III versus ARC II Figure 6: Table 10-1 and Figure 10-4, Chapter 10 of NRCS s NEH Figure 7: NOAA Atlas 14 Precipitation Values Figure 8: 2D Mesh Refinement Figure 9: Internal Connections Figure 10: Boundary Creation Workflow Figure 11: Profile Example of Validation Results Page iii

5 01 Introduction This document provides recommendations specifically for using HEC-RAS (RAS5) rain-on-grid 2D modeling and mapping utilities for producing Large Scale Automated Engineering (LSAE) products in South Dakota, along with limitations and relevant challenges. The latest version of HEC-RAS fails to incorporate an infiltration model when applying direct rainfall. This functionality, along with GIS preprocessing and other utilities, are anticipated in future releases of the publically available software. This document addresses these current limitations, and focuses on engineering and GIS measures that ought to be considered for any application of RAS5 rain-on-grid modeling. These procedural recommendations can be organized into two categories: 1. The standard Automated Engineering Approach, which is consistent with the procedures and level of effort that would be expected for an Automated Engineering analysis 2. The Regulatory Approach, which takes into account additional recommended considerations and enhancements to the modeling inputs and mapping outputs in order to create and publish Zone A Special Flood Hazard Areas (SFHAs) While these recommendations provide additional detail to the process and methodology presented in FEMA s Guidance for Flood Risk Analysis and Mapping: Automated Engineering (May 2016), they are complimentary, and not in conflict, to that guidance. FEMA s Automated Engineering Guidance provides only a brief mention of 2D analyses, maintaining focus on parameter settings in 1D unsteady analyses. The recommendations provided for using RAS5 for 2D modeling are in alignment with corresponding 1D guidance, where applicable. The RAS5 User s Manual ( D Modeling Users Manual.pdf) was referenced extensively as a resource in developing the methodologies outlined. Extensive testing of the software in a variety of scenarios, along with USACE guidance, significantly informed the recommendations for 2D LSAE, herein D Modeling Advantages and Considerations One of the primary benefits of a rain-on-grid 2D simulation is the hydraulic modeling is essentially handled by the detail of the terrain model used, as opposed to being constrained by the placement of 1D cross-sections and associated 1D assumptions and methods. Therefore, at the very least, attenuation of flood waves propagating through topography of any watershed should be well represented, usually much more so than 1D sub-critical models. This benefit, of course, depends on the study area and problem at hand. It is worth noting that when modeling very large areas that require numerous models (1D or 2D), significant benefit can be found by maximizing the 2D model areas, thereby minimizing the number of models. A main factor in determining these sizes is the number of 2D mesh cells, tied directly to the cell size. The question of when to use 2D modeling is described well by Gary Brunner in Chapter 6 of the HEC-RAS 5.0 2D Modeling Users Manual. As with any flood modeling and mapping, detailed terrain data is paramount. One of the unique features of RAS5 is sub-cell detail of a 2D mesh. That is, the 2D mesh computational cell size can be much larger than the terrain cell size, while still capturing reasonable hydraulic detail. As shown in Figure 1, flows and water surface elevations (WSELs) calculated across each cell face utilize the underlying geometry of the high resolution terrain data, as opposed to an approximation entirely limited by the mesh cell size. Mapped output is also based on the terrain grid cell size, rendered from the computational mesh cell size. For this and other reasons, it is recommended that the best available terrain data always be used given the option. Page 1-1

6 Figure 1: Hydraulic Computations in RAS 5 Utilize the Underlying, Higher Resolution Terrain Data Transforming rainfall into runoff for a watershed, most of which experience drastic ranges of soil moisture and a number of other conditions across seasons, months, and days, is a challenging problem in flood hazard identification. Care must be taken in the development of the model inputs and outputs of LSAE applications of RAS5 in South Dakota, as well as in communicating uncertainty and limitations. Despite the limitations and engineering assumptions required for any rain on grid modeling, and RAS5 specifically, the Automated Engineering Approach remains more sophisticated than traditional LSAE hydrologic and hydraulic modeling methods, particularly for flat terrain, areas common to South Dakota. 1.2 Data Sources The following data sources are vetted and widely available, providing most of the information used in various phases of the analyses. Studies in areas that are not covered by these data sources must be supplemented with other, best available, data. Precipitation Data: NWS/NOAA Precipitation Frequency Data Server GIS Precipitation Data: PFDS in GIS format (including confidence limits) Soils Data: USDA/NRCS Web Soil Survey Soils Data by State: Land Use Data: National Land Cover Database 2011 Terrain Data: National Elevation Dataset Page 1-2

7 02 Automated Engineering Approach The following section details an Automated Engineering Approach recommended for any application of RAS5 rain-on-grid in South Dakota. Again, these recommendations provide additional detail to the limited guidance provided in FEMA s Automated Engineering Guidance (May 2016). These recommendations for using 2D in RAS5 are in alignment with the corresponding 1D guidance in that document, where applicable. 2.1 Inputs The primary model inputs for any RAS5 rain-on-grid analysis are detailed below, not least of which is terrain, along with land use and roughness, and finally boundary conditions Terrain Summary: Details: 2D Automated Engineering analyses in RAS5 use terrain data with a resolution generally no coarser than 10 feet. All terrain processing, particularly mosaicking of terrain grids, should be performed prior to importing to RAS5. For example, a single terrain.tif file should be imported into RAS5. A terrain model is essentially the only digital data required for implementing RAS5 rain-on-grid modeling. The HEC-RAS 5.0 2D Modeling User s Manual (Manual) lists over a hundred file formats that can be imported into RAS Mapper (Appendix B), though a GeoTIFF (.tif) is generally recommended for creating a terrain model. The best available terrain data should always be used. Specifying the spatial projection in RAS Mapper is required before creating a terrain model. RAS5 will appropriately project the terrain file if it is in a spatial projection other than that specified for the model, though it is currently recommended all terrain processing be performed prior to importing to RAS5, particularly mosaicking of terrain grids with non-factorable cell sizes. The greatest benefit of RAS5 may be the sub-grid detail of a 2D mesh cell, supported entirely by the underlying terrain model. It is recommended that terrain grids no coarser than 10 feet (1/9 arc second) be used, except where this level of precision is unavailable. If a FEMA approved terrain source for the area exists, and supports a 1 meter terrain grid resampled from detailed LiDAR, then a 1 meter terrain grid should be used as the terrain source for importing into RAS5 for modeling, despite the additional processing time, and increased mapping output file sizes. RAS5 is able to handle very large terrain rasters, at least more than 50 gigabytes. Except for the obvious extended processing time, importing a very large 5 foot resolution terrain model does not excessively tax the equation-solving simulation any more than importing a 10 foot terrain model of the same area. RAS5 computes hydraulic characteristic tables for all 2D mesh cells, for a given terrain model, and these tables are referenced for computations within and across cells during a simulation Roughness Summary: 2D Automated Engineering analyses in RAS5 will use vetted and widely-available soils and land use data as the source for estimating Manning s n-values in the 2D model, unless local data is available. Page 2-1

8 Details: The Web Soil Survey and the National Land Cover Dataset can be leveraged for developing Manning s n- value coverages, as well as Curve Numbers (CN), Lag Times, and other data supporting rainfall-runoff simulations. All Automated Engineering Approach studies should utilize the NLCD dataset for creating RAS5 roughness/land Cover grids, unless local, or otherwise more accurate data, are available Boundary Conditions The following section discusses boundary condition considerations for any application of RAS5 rain-ongrid for LSAE efforts in South Dakota. Again, these recommendations provide additional detail to the limited guidance provided in FEMA s Guidance for Flood Risk Analysis and Mapping: First Order Approximation (November 2014). The recommendations for using 2D in RAS5 are in alignment with corresponding 1D guidance, where applicable. Contributing drainage area to a study area (i.e. a 2D computational mesh) can be modeled in a number of reasonable ways. Rainfall-runoff modeling (e.g. HMS) is generally considered a detailed modeling method for computing excess precipitation. USGS gage analysis discharges, paired with dimensionless unit hydrographs, provides another detailed method with flexibility in verifying hydraulic model results. Upstream boundary conditions, such as an inflow hydrograph, are to be developed based on the procedures outlined below, and it should be noted that significant effort (as with a typical detailed rainfall-runoff model) should not be spent on developing rainfall-runoff based inflow hydrographs. Figure 2 displays a typical situation where inflow hydrographs were derived and applied to the computational 2D mesh (along the contributing area inflow lines). The 2D mesh shown in the figure covers the study area (county boundary), for which an excess precipitation should be applied. Figure 2: Typical Scenario for Applying Inflow Contributions from Large Drainage Areas Outside RAS5 2D Study Limits Page 2-2

9 Generally, determining Lag Time for a sub-basin can be time consuming, depending on the methodology used. For LSAE in SD, much less time should be spent on developing these rainfall-runoff inputs than for a detailed rainfall-runoff model. For example, simply estimating the longest flow path for a sub-basin (based on any available data, detailed or not) can be used in conjunction with an assumed flow velocity, to quickly derive a Lag Time. Other, slightly more sophisticated methods can also be used, such as the Simple SCS method, so long as the time required to do so is minimal. When using a simple rainfall-runoff model (such as HMS) for Automated Engineering analyses, the objective is to develop an excess precipitation hyetograph that is appropriate for input into the 2D model, as opposed to trying to match recurrence interval peak discharges computed from a gage analysis. This is because the simple rainfall-runoff model does not take into consideration the attenuation of flood waves propagating through the topography of the area. The HMS simulation (or other rainfall-runoff model), therefore, is performed in order to generate a hyetograph for the RAS5 Precipitation boundary condition that considers losses (infiltration), since RAS5 does not have this functionality in its current version. Gage analyses are critical in the 2D LSAE modeling process, and all USGS and other available peak streamflow gaged runoff data should be collected prior to any hydrologic or hydraulic modeling. Gaged rainfall data can also be useful, specifically when analyzed in conjunction with runoff data. Statewide and regional USGS regression analyses and reports coincident with any study area should be reviewed in detail, as these reports most often utilize gage records on flooding sources being modeled. Percent annual chance of exceedance (recurrence interval) discharge estimates in these reports are often developed from gage records by fitting the logarithms of annual streamflow peaks to a Log-Pearson III distribution, as described in Bulletin 17B and as modeled within the USGS s PeakFQ and the USACE s HEC-SSP software. The lower and upper confidence limits for all gage analyses should be 0.16 and 0.84, respectively, which correspond to a confidence interval of This also corresponds to a standard error of prediction in a recurrence interval estimate, defining the 1% annual chance Plus and Minus events used by FEMA for study validation purposes (and detailed in Section 3.3). It should be noted that PeakFQ provides a single entry for Confidence Intervals ; the value entered to determine the 1% annual chance Plus and Minus estimates should be 0.84 (which is actually the upper confidence limit desired) Upstream Summary: Details: Inflow hydrographs developed for significant drainage area outside of the 2D model mesh should be developed within HEC-HMS (HMS), using SCS CN (for losses) and Lag Time (for transform), or other approved methods. However, when gage data is available, this should drive the hydrologic modeling, whether maintaining a rainfall-runoff approach or applying the gage unit hydrograph approach. Areal reduction factors should be used to reduce the recurrence interval precipitation values for all precipitation events modeled within HMS. These will be based on the drainage area to which that rainfall is applied, as well as engineering judgment, particularly with regards to verification of 2D model results. The following outlines the rainfall-runoff approach for developing inflow hydrographs, for contributing drainage areas that are not part of the 2D model mesh, as well as excess rainfall hyetograph time-series for areas modeled by a 2D mesh: Page 2-3

10 1. Develop simple rainfall-runoff model (HMS) with area, CN and Lag, or some other loss and transform methods, specified for contributing areas to the RAS5 2D mesh. Or, if sufficient peak streamflow record, as well as observed event record exists for a gage at a contributing inflow location, develop recurrence interval inflow hydrographs from gage analysis information using a unit hydrograph approach (this unit hydrograph approach should be used in lieu of rainfallrunoff modeling, if appropriate records available, and is detailed later in this report section). 2. Determine a precipitation depth (from PFDS) and use a NOAA, SCS, or other reasonable storm distribution in the rainfall-runoff model (HMS). 24-hour storms are generally recommended, though shorter (or longer) duration storms can be considered, particularly when interested in matching up flooding along large mainstem reaches and incoming tributaries. 3. Use the simulated or gage-derived outflow hydrographs for upstream inflow hydrograph boundary conditions to the 2D mesh. Point precipitation values, or GIS polygons for recurrence interval rainfall durations, can be obtained from the PFDS and used to determine recurrence interval rainfall for modeling within HMS. Areal reduction factors should be considered using TP-40, or another approved or more appropriate methodology for these South Dakota study areas, and ultimately engineering judgment. The areal reduction applied can also be considered in the reasonability and verification process described later in the report. Figure 15 from TP-40 is provided in Figure 3 below, for reference: Figure 3: TP-40 Figure 15 Area-depth Curves Statewide soils coverages in gridded format, for developing initial CN s, should be obtained if the study area is large enough to require a prohibitive amount of individual county soil coverage downloads. Statewide soils coverages can also be obtained from NRCS s Web Soil Survey at this link: A Landuse-Soils-CN matrix can be used to determine CN s for a particular study area. Table 1 below displays a typical matrix for determining CN s for the intersection of soils, land use, and drainage areas, in order to develop a weighted CN for a drainage area. The CN values are sourced from the NRCS s TR- 55. This matrix is intended as a general guide, and should be adjusted based on engineering judgement for a given study area. Page 2-4

11 Table 1: Typical Landuse-Soils-CN Matrix LU_GridCode NLCD LU Description Hydrologic Soil Group A B C D 11 Open Water Developed Open Space Developed Low Intensity Developed Medium Intensity Developed High Intensity Barren Land Deciduous Forest Evergreen Forest Mixed Forest Shrub Scrub Herbaceous Hay Pasture Cultivated Crops Woody Wetlands Emergent Herbaceous Wetlands Inflow hydrographs should generally be used for all significant drainage areas or streams intersecting the study area boundary which would contribute to the 2D simulations. These hydrographs should be generated, at least initially, using rainfall-runoff modeling or pairing gage analysis information with the unit hydrograph approach. Further relevant discussion for inflow hydrographs, though specifically regarding excess rainfall, is provided in Section The unit hydrograph approach for inflow locations with sufficient peak streamflow and observed historical record should be used in lieu of a rainfall-runoff simulation. This approach involves the following procedure: 1. Perform a flood frequency analysis, using Bulletin 17B procedures (or other approved methods) to determine the magnitude of recurrence interval discharges of interest. 2. Utilizing observed event record, several large events should be selected, with preference given to relatively simple, single-peak hydrographs. 3. Convert the observed event hydrographs to dimensionless hydrographs by computing the time and discharge ordinates as t/tp and q/qp (where tp is time to peak, qp is peak discharge), and align the dimensionless hydrographs. 4. Determine an average q/qp and t/tp to derive an average dimensionless hydrograph which can be used for developing scaled hydrographs for recurrence interval discharges determined in the gage analysis. 5. Use these gage-derived outflow hydrograph(s) for upstream inflow boundary conditions to the 2D mesh. Page 2-5

12 Flexibility in establishing desirable timing of inflows can be realized using this approach, which in turn may provide flexibility in reasonability and verification efforts, as discussed later. When modeling very large multi-model areas (multiple 2D models), although using outflow hydrographs from upstream rain-on-grid models as inflows for downstream adjacent models would perhaps be most ideal, this can be quite problematic specifically with regards to overall simulation time moving downstream. Furthermore, doing so often creates significant issues with reasonability efforts moving downstream. Yet another option would be to model large upstream areas using a separate RAS5 model with a coarse cell size. The outflow hydrograph from this simulation could be used as the inflow hydrograph for the downstream model/area of interest. However, an HMS model would still be required to develop excess precipitation for that upstream coarse model. This method has not been tested, but is likely a viable option for areas that may be more suited to this approach. It should be noted that significant effort should not be spent on developing rainfall-runoff models, specifically initial CN s, as these will end up being adjusted, often significantly, during the verification process Downstream At a minimum, all 2D meshes should have an outlet for flow to leave the system. In some cases, the outlet boundary condition line(s) can extend around a very large portion of the mesh. However, an ideal mesh will have a single obvious location for the downstream outlet. All 2D meshes should have at least a single outlet boundary condition, such as normal depth, which does not create any artificial backwater. Gage rating curves from gages at or near the outlet of a model are recommended for use as the model downstream boundary condition, whenever available Initial Conditions Initially wet conditions may need to be considered, especially for significant flood control structures for which the terrain model captures significant bathymetry that should be considered unavailable for flood storage during a significant event. Performing a very rough simulation, with a coarse time step, that is long enough in duration for volume remaining in the RAS5 mesh to empty (unless it ought to remain trapped in depressions or otherwise), could then be used as a restart file for a refined simulation. This restart file approach can be particularly useful for beginning a simulation with large (and small) flood control structures approximately full (unless the terrain model captures approximately normal pool, which is often the case for LiDAR-based terrain) Precipitation Summary: The Precipitation boundary condition for a 2D Automated Engineering analysis in RAS5 should be developed using a rainfall-runoff simulation, using SCS CN (for losses) and Lag Time (for transform), or other approved methods. Details: A single Precipitation boundary condition can be specified for any RAS5 2D mesh, and a single 2D computational mesh is generally recommended for any automated engineering application of RAS5 rainon-grid modeling. Precipitation data from NOAA s PFDS can be used directly for the Precipitation hyetograph boundary condition in RAS5, though doing so assumes all rainfall is converted to runoff. This Page 2-6

13 data can also be used as the Meteorological input for a rainfall-runoff simulation in order to derive an excess precipitation hyetograph for the Precipitation boundary condition of a RAS5 2D mesh (as well as inflow hydrographs for contributing drainage areas extending beyond that 2D mesh). GIS format of PFDS data, including confidence limits, is available, and should be used where appropriate for 1% plus and minus events (detailed in Section 2.5). Similar to the approach defined in Section for developing inflow hydrographs, the following outlines the rainfall-runoff approach for developing excess rainfall hyetograph time-series data for areas modeled by a 2D mesh: 1. Develop simple rainfall-runoff model (HMS) with area, CN and Lag, or some other loss and transform methods, specified for a sub-basin that represents the RAS5 2D mesh. 2. Determine a precipitation depth (from PFDS), select a reasonable areal reduction to the rainfall total(s), and use a NOAA, SCS, or other reasonable storm distribution in the rainfall-runoff model. 3. Use the simulated excess rainfall time-series as the Precipitation boundary condition for the 2D computational mesh. 2.2 Model Controls The following sections provide recommendations and discussion for RAS5 model controls Mesh Area Summary: Details: 2D meshes should be limited to approximately one million cells (this is discussed further in Section 2.2.2). The size and shape of each 2D model mesh should be selected in order to maximize the area that can be modeled using the rain-on-grid approach, while focusing on the location of useful peak streamflow gages for model inflow and outflow locations. This general approach limits the need to utilize additional RAS5 models and inflow hydrographs, while more importantly providing the best means for reasonability and verification measures. RAS5 2D meshes can be developed using GIS, applying smoothing and simplifying tools, which significantly reduces the need for making edits to the 2D mesh boundary within RAS5 to resolve mesh errors. Generalizing polygons is another available feature using GIS, however the process creates problematic sharp angles, which should be smoothed or simplified before using the vertices to define the X-Y coordinates of a RAS5 2D mesh. It is recommended that a very small 2D mesh be created within RAS5 (4 or 5 clicks), and then the 2D mesh vertices (with X-Y coordinates) be imported or pasted into the Storage Area Outlines table defining the spatial orientation of the mesh. It should be noted, the.dbf file associated with a shapefile can be opened within Excel, such as for copying and pasting coordinates directly into the Storage Area Outlines table for a 2D mesh boundary. If jagged edges exist for a 2D mesh boundary, significant mesh modifications will be required. This should be avoided by smoothing/simplifying a jagged-edged polygon before converting feature vertices to points. Not performing this step could be the difference in minutes versus a day for establishing an error-free RAS5 2D mesh. Also, it is recommended that 2D mesh errors be addressed by first modifying the mesh boundary, as opposed to moving mesh cells, and then proceeding with addressing the internal Page 2-7

14 cell errors. Using this approach should help eliminate the need to resolve repeated errors regarding cells along the mesh boundary that would occur if a new mesh is generated (such as if a new nominal cell size is desired) Cell Size Selection Summary: It is recommended that for 2D Automated Engineering analyses in RAS5, an initial nominal mesh cell size of 200 feet be used. The engineer performing the analysis may increase or reduce this during the course of model iterations if deemed appropriate or necessary for a particular study area, model stability, or any other reasonable factor. Details: For accurate rain-on-grid modeling, the underlying terrain model must sufficiently describe the topography of the study area. In RAS5, 2D mesh cell faces are essentially treated as cross-sections, while hydraulic tables are computed and stored for each cell. That is, the terrain model may consist of 1 meter cells, while the nominal cell size for a large 2D mesh covering 1,000 square miles can be significantly larger without sacrificing all hydraulic detail. How much larger depends in large part on the purpose of the modeling. For applications of RAS5 for Automated Engineering analyses in South Dakota, 200 feet is a reasonable starting selection for the nominal cell size of a 2D mesh. However, specific watershed topography and conditions, and useful gage locations (and therefore 2D mesh area), should drive the nominal cell size selection. Watersheds (of equal contributing drainage area) having hundreds or thousands of feet of elevation change from ridge to outlet may need finer cell sizes than those with only dozens of feet elevation change. For South Dakota LSAE, 200 to 300 feet nominal cell sizes have shown to provide sufficient resolution for obtaining stable and reasonable results. 200 feet is the general recommendation, though may vary based on work/model area. It is a general recommendation that 2D meshes be limited to approximately one million cells, as it has been shown that exceeding this number by a relatively small margin causes significant issues in RAS5, often resulting in memory overflow errors (seemingly regardless of hardware configurations). A 200 foot nominal cell size for a large (countywide) area, supported by a 10 foot terrain raster, is more than sufficient for accurately representing large rivers and streams in the study area, though it is likely the representation of very small streams would improve with a smaller cell size. Determining a mesh cell size is a balancing act the number of cells, along with the simulation interval, dictates in large part the run time, and the modeling and mapping accuracy. The runtime increases exponentially as the number of cells increases, and time step decreases. Care should always be given to producing stable modeling results, and little confidence given to results of unstable simulations. However, the mapping outputs for even an unstable rain-on-grid simulation can provide a great deal of information about the study area, such as to indicate where is it going to flood and what features, bridge and culvert crossings or otherwise, may require additional attention. The nominal mesh cell size should be fine enough to produce stable computations. The mesh cell size should be fine enough to produce floodplain mapping products comparable to those from conventional 1D methods. The overall shape of a 2D mesh will depend on the actual study area (e.g. within a watershed or county), the topography, and the extent of the terrain model. Ideally, models will be setup and performed on a watershed basis, so as to optimize the model s ability to fully utilize the benefits of a rain-on-grid approach, and to capture all the contributing runoff within the study area. In some cases, however, it may be appropriate or necessary to reduce the size of the 2D mesh and to use inflow hydrographs for particular flooding sources extending beyond the mesh boundary. Doing so will prevent Page 2-8

15 individual models from becoming unnecessarily-large for the sole purpose of performing a rain-on-grid analysis for the entire contributing drainage area (regardless of the actual study area). In these cases, inflow hydrographs should be developed, as previously described, using rainfall-runoff modeling or applying the unit hydrograph approach to gage data, and then input into the RAS5 model. It should not be expected that any 2D model simulation be a one-and-done iteration, RAS5 included. Adding layers of complexity is part of the 2D modeling process. Even if a modeler is satisfied with an early run at 200ft cells and a 30 second time step, additional iterations may well be warranted Time Step Selection Summary: 2D Automated Engineering analyses in RAS5 will use a computational time step required for stable model results. 30 seconds is a reasonable selection for a 200 foot cell model, though may vary without sacrificing accuracy for a work area model. Details: The Manual describes that Courant numbers as high as 3.0 can be sufficient when using the Full Momentum equations, while values as high as 5.0 can provide sufficient accuracy and stability when using the simplified Diffusion Wave equations. Likewise, there are instances where ensuring a Courant number less than 1.0 is required for accuracy and stability, even when using the more stable Diffusion Wave equation set. Generally, the computation interval should be small enough such that the time required for water to move through any cell is greater than that interval. Most importantly, the time step used must be sufficient to produce stable results, which can be identified rather quickly by viewing stage and discharge hydrographs within a 2D mesh. Methods for providing these stage and discharge hydrographs within a mesh are described in a subsequent section. A general rule for 2D LSAE modeling in South Dakota is that a time step (in seconds) of a simulation not exceed 0.15 multiplied by the nominal 2D mesh cell size (for a 200 nominal cell size, this equates to a simulation interval of 30 seconds). A balance between accuracy (and precision) and run times for a large model is an inherent goal in the RAS5 2D modeling process. Reducing the time step may be a solution for resolving areas of a model producing suspect results. However, say a model is producing accurate results at every location in a watershed, except for a few locations where hydraulic features, say waterfalls, exist. Reducing the time step to resolve errors at these features, which exist in varying intensities in any watershed, at the expense of taxing run times that gain no more accurate results is not recommended. 30 seconds is a reasonable selection, though may vary, so long as reasonable results are simulated for a particularly work area model Equation Selection and Other Computation Settings Summary: 2D Automated Engineering analyses in RAS5 should primarily use the Diffusion Wave equations. In certain instances, however, the Saint-Venant equations may be used if deemed necessary by the engineer performing the analysis. Model conservation of volume errors less than 0.1% will be considered acceptable. Hydrograph and Model Output intervals will initially be set at values significantly larger than the time step. The engineer performing the analysis should adjust these as necessary. Page 2-9

16 Details: A rough initial simulation is recommended (i.e. using a very coarse time step), in order to define an end of simulation time that captures the full outfall hydrograph. Once a Terrain.hdf file is created within RAS Mapper, a 2D mesh can be created within the Geometry Editor, boundary and model setting conditions can be set (in the unsteady flow and plan file settings, respectively), and a RAS5 rain-on-grid simulation can be performed. In all cases, 2D modelers are tasked with balancing hydraulic resolution with computational runtime. Six significant factors in this balance are 2D mesh area, nominal 2D mesh cell size, computation interval, simulation duration, output settings, and equation selection. It is recommended that a rough initial run be made, with a coarse time step (e.g. 5 minutes) in order to determine the time at which the outfall discharge hydrograph no longer contributes significantly (e.g. to a downstream model). Then, the simulation duration can be set such that the simulation ends shortly after the peak is experienced in the outflow hydrograph. The shape of the outflow hydrograph from a model using a proper time step, beyond the peak, can then be approximated. This hydrograph which can be used for the inflow hydrograph of a downstream model will consist of the trimmed-just-after-peak portion as well as an approximated tail. The start of a simulation should generally begin when either inflow hydrograph runoff begins emptying into a 2D mesh or the excess rainfall hyetograph applied to a 2D mesh begins. The simulation run times of models moving downstream will grow dramatically if each of these models must have a full simulation such that the entire outflow hydrograph is captured. Using the aforementioned approach can alleviate, perhaps minimize, the length of simulation runs, which translates to very significant savings in computation times throughout the project. The Full Momentum, or better known as the Saint-Venant equations for shallow flow, can be specified for 2D computations. It should be noted, using these full shallow water equations significantly increases runtime. A simplified set of equations, known as the Diffusion Wave equations, are also available, which are typically suitable for flood modeling, where inertial forces tend to dominate frictional and other forces. Mixed Flow Regimes are to be expected with any unsteady hydraulic modeling, so as depths reduce from subcritical to supercritical conditions, the Saint-Venant equations become increasingly more accurate. However, the Manual makes note that the simplified equation set is usually sufficient for purposes such as these flood inundation applications. For the approaches described herein, the Diffusion Wave equations were considered to provide a sufficient balance of model accuracy, stability, and runtime. The Manual mentions a conservation of volume error for a particular study of 0.011% as very low. This relative value and the description in the Manual applies to a much smaller study area than is addressed as part of a LSAE analysis. It is, therefore, recommended that a volume accounting error of 0.1% not be exceeded. The volume of error for a model can be viewed in the Computation Log File (.bcoxx file), as shown below. Page 2-10

17 Figure 4: Computation Log File Example Within RAS5, *.hdf files are generated for each run of each plan. Depending on the Computation Settings, these files can grow quite large, and hinder runtime performance. If only maximum depths or elevations are of interest, Hydrograph- and Mapping Output Intervals are irrelevant, and should be set to very large values compared to the time step. However, useful time-series information of a simulation should be available for pertinent final simulations. It is recommended that the interval for hydrograph outputs be set large enough to minimize unnecessary hard drive data writing, while fine enough to capture useful (stable) stage and flow hydrographs. An acceptable value may be hours or more for a first run. Output Options can be accessed from the Options menu of an Unsteady Flow Analysis Plan window. Ultimately, the final Hydrographand Mapping Output Intervals become part of the 2D product, so all of these settings should be considered before post-processing Internal 2D Mesh Connections Summary: Details: Stage/Discharge comparison lines should be added into the 2D Automated Engineering model in RAS5 at stream gage locations, and along small and large drainage area flooding sources to provide a reasonable method of comparison for a study area. Reservoirs, particularly those known to provide significant flood control storage, should have internal connections defined. Reasonable results at such embankments must be supported by modeling, as well as published information. Detailed discussion on the use of internal connections and breaklines for enhancing 2D LSAE models is provided in Section 3.1, including the use of offset breaklines and internal connection structures for modeling transportation and other flow obstructions. The following section is meant to provide basic information regarding internal connection functionality within RAS5 that should be considered for any application in South Dakota 2D LSAE modeling. Stage and discharge hydrograph time-series data at various locations within a 2D mesh are valuable for comparing RAS5 rain-on-grid results with other available data. In order to provide locations for reporting stage and discharge hydrographs, RAS5 Breaklines should be created (described below) and converted to Internal Connections. Terrain profiles along Internal Connections can be copied to weir geometry for the Connection, and filtered to less than 500 points, as required. It is worth noting, some GIS software, Page 2-11

18 both proprietary and open source, provide plugins for importing 2D modeling data, alleviating the need for the breaklines to obtain discharge hydrographs, though these are not discussed here. The weirs created in this process behave more like 1D cross-sections, tied to the 2D mesh cells. Since these features are really just locations of measure, a minimal width of the weir (1 foot, maybe less) is recommended (though this is merely a schematic setting), along with a minimal ground exchange weir coefficient of 0.0 to 0.2, unless otherwise appropriate. The Normal 2D Domain Equation should be selected for these features. Breaklines can be defined quickly, within a 2D mesh, by importing georeferenced polyline shapefiles via the GIS Tools of the Geometry editor. These breaklines can be converted to Internal Connections with ease. (The number of stage and discharge hydrograph locations does not have to be extensive, but should be sufficient for validating the rain-on-grid results. Stream gages in the study area are optimal locations to add these discharge comparison locations. As a rule of thumb, one stage/discharge hydrograph comparison line ought to be added into the mesh say for every 50 square miles of model area, though this value will vary for all study areas. Internal connections should be generated for significant reservoirs within a 2D mesh. Simple spillway geometry can be obtained from terrain data, or perhaps other published information, such as normal pool elevation. Storage of an embankment that would not be available during a flood event should be excluded from any flood control storage considerations. A restart file of a simulation, having these such features full of water and unavailable for flood attenuation, can be used for initial conditions of a simulation (as discussed in Section ). Generally, published normal pool elevations should be similar to terrain elevations within significant water bodies, though all of these data should be considered in hydrologic and hydraulic modeling. It is not uncommon to see unreasonable results at dams during a 2D simulation, depending in large part on the orientation of 2D mesh cells, including erroneously high water surface elevations due to a lack of release from the dam. In such cases, gate and culvert openings should be considered within the internal connection defining and representing the dam. 2.3 Reasonability and Verification Checks for the 1% Annual Chance Event Summary: Details: Multiple iterations of the 2D Automated Engineering analysis in RAS5 may be run until WSELs and discharges are deemed reasonable by the modeling engineer. Verification measures should always be implemented first and foremost with the most robust available data (i.e. systematic gage records). Further verification measures should be driven by the quality and availability of pertinent data. There will often be several instances where reservoirs, particularly those providing significant flood control storage, can be considered as verification points within a model, whether or not sufficient gage record exists or detailed information is available. Verification checks for locations without gage estimates should use the best available data, including model-backed effective AE and A Zone flooding, regression equation estimate ranges, and even adjacent watersheds with available gage estimates. Following is a simplified approach for checking for reasonable results from any RAS5 rain-on-grid application, given the lack of variable rainfall or infiltration functionality for a RAS5 2D mesh in the Page 2-12

19 current version of RAS5. Although this procedure is suggested for any application of RAS5, it is also recommended for the Regulatory Approach for converting to Zone A SFHAs (as discussed in Section 3.1). It should be noted that the Automated Engineering Approach may not warrant more than a single iteration or two to reach satisfactory reasonability for the Automated Engineering Approach. It should also be noted the following verification measures are significantly more sophisticated than any such measures performed for typical 1D LSAE methods. Gage analysis estimates of the 1% annual chance stage and discharge essentially capture attenuation effects of the watershed draining to the gage. This is a severe simplification, as storm orientation, soil moisture, and much more is also captured by gage sampling. The hydrologic input in the RAS5 2D Automated Engineering analysis is rainfall. When using a simple rainfall-runoff model (such as HMS) for Automated Engineering analyses, the objective is to develop an excess precipitation hyetograph that can be input into the 2D model, as opposed to trying to match simulated recurrence interval peak discharges to those computed from a gage analysis. This is because the simple rainfall-runoff model does not take into consideration the attenuation of flood waves propagating through the topography of the area. The HMS simulation (or other rainfall-runoff model), therefore, is performed in order to generate a hyetograph for the RAS5 Precipitation boundary condition that considers losses (infiltration), since RAS5 does not have this functionality in its current version. The more important reasonability checks occur by comparing the WSELs and flood boundaries from the 2D model with the associated stage and discharge from a gage analysis, observed data, existing/effective WSELs and boundaries (primarily from model-backed Zone AE and Zone A studies), or other information that may be available. It should not be expected that the 2D model results will compare too closely with effective Zone A boundaries, as the outdated nature and associated deficiencies with many effective Zone A studies are widely known. The 2D Automated Engineering analysis represents a more sophisticated and credible approach, and as such should be expected to vary from many of these older, effective Zone A floodplains. However, model-backed Zone A studies should indeed be used for verification of model results, and significant effort should be made to tie into effective, model-backed floodplain boundaries. Regarding gage estimates to be used for verification locations, consideration must be given to the unregulated or regulated nature of the systematic record. This is also true for using verification locations which can be considered unregulated that are downstream of regulated gage records. Care must be taken to ensure proper comparisons are being made between gage estimates and 2D model results. All reservoirs, particularly those providing significant flood control, should also be considered for points of verification. Such embankments will often have significant published data, often more than enough to make reasonable assumptions for normal pool, for example, or perhaps documented values for such. Also, it is common that any LiDAR-based terrain model captures the approximate normal pool of a significant reservoir, depending on the operation of the reservoir and time during which the LiDAR was collected. Consideration should be given to water bodies within a model area, specifically as to whether they should provide significant flood protection during a significant event. When effective data, specifically effective WSEL s are not available for use in verification, regression equation estimates can be used as a guide for determining if further adjustments are needed to improve verification results. Specifically, the range of discharges computed using the standard error range of applicable regression equations can be compared with 2D model results at verification locations. However, it will not be uncommon for discharges to vary wildly, while actual WSELs would not. If effective detail data is available, then these WSEL s and boundaries should generally take precedence Page 2-13

20 over all other measures, except for gage estimates with sufficient record (at least 10 years of systematic record, preferably 20 or more). The following reflects the minimum approach that will be followed to perform checks to verify that the results from the 2D RAS5 model are reasonable: 1. Compare resulting RAS5 1% annual chance WSELs, flood extents, and peak discharges with best available data at the Stage and Discharge comparison lines discussed in Section Using engineering judgment, adjust Curve Numbers and/or Lag Times within the HEC-HMS model and regenerate an excess precipitation time-series for use as the Precipitation boundary condition in RAS5. If the unit hydrograph approach is employed for deriving inflow hydrographs, the timing of these can be adjusted (as well as magnitude) based on engineering judgment. 3. Rerun the RAS5 model with the updated variables. 4. Repeat Step Proceed when RAS5 elevations (and peaks) are reasonably close to the best available data (engineering judgment shall always be weighted more than general recommendations). It is worth noting another option for adjusting RAS5 results is to use Manning s n Land Cover data, though this has proven to be relatively ineffective in this regard. 2.4 Multi-Frequency Analyses Summary: Once the 1% annual chance event has been run and results checked, the 2D Automated Engineering analyses in RAS5 will be run for the 10%, 4%, 2%, and 0.2% annual chance events. Details: Once the 1% annual chance event RAS5 results have been verified, as described in Section 2.3, precipitation values corresponding to the 10%, 4%, 2%, and 0.2% events should be developed similarly to the 1% precipitation values. These precipitation data should then be applied to the rainfall-runoff model resulting from the 1% verification, without adjustment to other parameters (i.e. CN and Lag). The outputs from the rainfall-runoff run for these frequency events should be applied directly to the RAS5 model, as described in the previous section. If the gage-based unit hydrograph approach is used for inflow hydrographs, the scaled hydrographs should be used directly for all percent annual chance events. It should be noted that ratios can be applied, within both HMS and RAS5, in order to determine the precipitation input into HMS (and therefore excess precipitation for a 2D mesh directly) and hydrograph (for an incoming drainage area) inputs into RAS5. Utilizing this option should produce nearly identical results as developing the full range of precipitation inputs into HMS and excess precipitation input into RAS % Plus and Minus Analysis Summary: The 1% plus and minus analyses for a 2D Automated Engineering analysis in RAS5 should take into account error bands on the precipitation estimates provided by NOAA. Errors in runoff Curve Numbers applied within a rainfall-runoff model for deriving excess precipitation hyetographs should also be Page 2-14

21 considered. The 1% plus and minus analyses when using the gage analysis unit hydrograph approach should be determined directly from the 16% and 84% lower and upper confidence limits, respectively, of the Bulletin 17B frequency analysis of the record for the 1% event (i.e. the 68% confidence interval). Details: Procedures for estimating the discharge for the 1% annual chance plus and minus events for a gage analysis or using regional regression discharge estimates (i.e. based on a regional gage analysis) are usually well-defined and straightforward. Prediction errors in these instances are readily available via the statistical methods employed when undertaking a Bulletin 17B gage analysis. Procedures for quantifying uncertainty for deterministic models, such as rainfall-runoff or rain-on-grid methods are not well-defined. Specifically regarding rainfall-runoff modeling such as with HMS, the USACE s EM , Risk- Based Analysis for Flood Damage Reduction Studies is perhaps the most definitive guidance. However, it is lacking. Procedures described in EM quantify uncertainty in prediction essentially using Bulletin 17B guidelines for a gage analysis. Discharge estimates, including the 50% event are used, and an equivalent years of record value to be user-selected from a table is also required. The table differentiates this value by different levels of rainfall-runoff-routing model complexity and agreement with observed data, generally in increments of 10 years, with 10 years being the smallest value. It is worth noting, the selected equivalent years of record is dramatically sensitive to the 1% plus and minus results. The rainfall-runoff modeling defined for 2D LSAE in this guidance is merely a simplification of a rainfallrunoff (no routing) simulation to address the lack of infiltration modeling functionality in the current version of RAS5. No routing or attenuation is considered, and model parameters are often revised significantly during the RAS5 model reasonability and verification process. For 2D LSAE in South Dakota, Antecedent Runoff Condition (ARC) II CN s should be used for the 1% event and ARC III CN S for 1% plus rainfall-runoff event estimate. The 1% minus event should be computed using CN s assuming ARC 1.5 conditions, halfway between ARC I and ARC II conditions (Figure 5). Figure 5: Plot of ARC I, ARC 1.5, and ARC III versus ARC II Page 2-15

22 A review of the area indicated the 1% minus event is of lower magnitude than the 10% annual chance event using hydrologic methods other than a gage analysis. It is recommended herein that ARC 1.5 be used for the 1% minus event. It is also recommended the discharge estimates simulated (via rainfallrunoff modeling) for the 1% minus event and used for input into a RAS5 2D model, should be based on the combination of the rainfall and CN adjustments, and generally should not exceed the simulated estimate for the 10% annual chance event. Engineering judgment should also be part of final determination for the CN adjustment, and deviations from the recommendations documented. The unit hydrograph approach involves gage analyses, for which CN and rainfall adjustments do not apply. Applying these guidelines, a reasonable and standardized approach for deriving a band of uncertainty, comparable to other methods and appropriate for validation purposes herein, is made available. The conversion from a normal Antecedent Runoff Condition (ARC II) to a wet (ARC III) or dry (ARC I) condition is provided in Table 10-1 of the Hydrology section of the NRCS s National Engineering Handbook, and included in Figure 5, below. The same information is also provided in graphical form in Figure 5. Figure 6: Table 10-1 and Figure 10-4, Chapter 10 of NRCS s NEH-4 It is recommended rainfall inputs be adjusted, whether directly on a 2D mesh or within HMS over a subbasin, to model 1% plus and minus events, as described in the example below (Table 2). Using a rainfallrunoff simulation for developing an excess precipitation hyetograph to be used as the Precipitation boundary condition in RAS5, or for inflow hydrographs for upstream boundary conditions, adjustments to CN s should also be considered, in order to achieve ranges in rainfall-runoff simulated peak discharges similar to that of a gage analysis in the study area, or regression equations used in the study area. Because RAS5 can only use a single precipitation value for each 2D flow area, point rainfall values from the NOAA PFDS can be applied (either directly from the PFDS or via an area-weighted determination from PFDS GIS data). In addition to recurrence interval precipitation estimates, NOAA Atlas 14 provides 90% confidence intervals of reported precipitation values, as shown below. Page 2-16

23 Figure 7: NOAA Atlas 14 Precipitation Values On a normal distribution curve, 90% confidence intervals correspond to +/ standard deviations (i.e. 5% on each tail). The 1% plus and minus events are defined to be one standard deviation above and below the 1% event. It should be noted that the error bands for the selected rainfall depth should be used on the precipitation value used in modeling post-areal reduction. Here is an example of the calculation process for a point depth of 5.56 inches: Table 2: Point Depth Calculation Example 1% Precip: 5.56 in 90% Upper Limit: 7.02 in 90% Lower Limit: 4.34 in Difference (3.29 st. dev.): 2.68 in 1 st. dev.: 0.81 in 1% Plus Precip: 6.37 in 1% Minus Precip: 4.74 in 1% plus and minus event precipitation totals can likewise be determined using a GIS format of PFDS data, specifically of confidence limits. Using GIS data instead of point rainfall totals differs only in that GIS data provides a spatial understanding and area-weighted determination for a point depth estimate. Precipitation Data: NWS/NOAA Precipitation Frequency Data Server GIS Precipitation Data: PFDS in GIS format (including confidence limits) 2.6 Model and Mapping Outputs Geometry and plan run data are stored in the.g01.hdf and.p01.hdf files, respectively. The size of the plan.hdf file depends in large part on the Computation Settings, as described in section These include the Hydrograph, Mapping, and Detailed Output Intervals. Summary: Page 2-17

24 Details: Depth and WSEL grids will be exported for each flood event analyzed utilizing the Sloping (interpolated) rendering feature. Velocity grids will be exported for the 1% annual chance event. For the Automated Engineering Approach, the Hydrograph Output Interval should be fine enough to detect and confirm model stability considering any stage and discharge hydrograph from a simulation. The Mapping Output Interval is used for dynamic mapping (animating in RAS Mapper) of model results. This interval must be equal to or larger than the Hydrograph Output Interval, and the interval to use depends primarily on how interested a modeler is in a visualization of the simulation. The finer the interval, the more data is written to the plan run.hdf file on disk, having a significant impact on runtime. A great deal of information can be provided by visualizing the event, so any final run should consider a Mapping Output Interval that provides that utility, while also balancing data writing and storage. The Detailed Output Interval for an Automated Engineering Approach can be set to a very large interval, unless detailed information of the computations is important for the study. Depth and WSEL grids for the 10%, 4%, 2%, 1%, 0.2%, 1% plus, and 1% minus events should be exported, and included within the default model subfolders, for any multi-frequency RAS5 rain-on-grid model. The Sloped rendering mode should be selected from the Render Mode options window and used for these mapping exports. Inundation boundaries can also be exported from RAS Mapper, though this has proven troublesome for large studies. Fortunately, these can be generated with ease from the grid results using GIS tools. Generating the flood boundaries is an optional task when utilizing the Automated Engineering Approach. Page 2-18

25 03 Regulatory Approach There are considerations to be made for converting the Automated Engineering Approach products to a level suitable for regulatory purposes. This section discusses pertinent enhancements to the models developed in the Automated Engineering Approach that should be considered, primarily in the context of 2D LSAE efforts. However, the modeling and reviewing engineers retain the final decision on the extent to which applying these enhancements would actually result in a noticeable improvement in model results. Currently, the greatest limitation of RAS5 rain-on-grid is its inability to model infiltration of rainfall for any given condition. The Base and Regulatory Approach each address this limitation, via rainfall-runoff modeling for determining excess precipitation resulting from a recurrence interval flood event. For the Base and Regulatory Approach for large scale modeling, the use of a rainfall-runoff model, simple or complex, should be in response only to the current lack of infiltration modeling in RAS5. There are, however, opportunities for extending the understanding of runoff response for any watershed that is not covered here. Greater engineering detail of specific areas within a RAS5 LSAE model is a likely interest of stakeholders. Of course, more detailed models can be produced for these areas. However, the LSAE approach can be maintained by refining a 2D mesh in RAS5, for example, using breaklines, internal connections, internal structures, and manual cell configurations at targeted locations. Improving the definition of risk for particular areas within a work area model seems an appropriate measure for upgrading a 2D LSAE model to a product suitable for regulatory purposes. 3.1 Modeling Considerations Summary: Details: To enhance a 2D Automated Engineering analysis in RAS5 for regulatory purposes, breaklines can be added at select locations within a 2D mesh. Offset breaklines can be used to approximate structure crossings within a 2D mesh Internal connections with culvert openings can be used to approximate structure crossings within a 2D mesh. Accredited levees should have breaklines positioned along the crest, while non-accredited levees should not. A RAS5 2D LSAE model can be upgraded significantly with the use of breaklines. These can be imported from shapefiles, or drawn and incorporated into a 2D mesh directly within RAS5. Breaklines can be used along, or offset from, transportation and other obstructions of a flooding source to model the ability of these features to pass runoff through embankments. Breaklines can be converted to internal connections, providing stage and discharge hydrographs at specific locations within a 2D mesh, as well as for approximating structure crossings within a 2D mesh. Breaklines and internal connections are very useful for describing features within a 2D mesh that require finer resolution than the nominal mesh cell size. Breaklines that are offset from road embankments, such that the edge or face of a mesh cell is not adjacent to the crest of these features, but rather straddles embankments so cell faces exist on either side of the crest, can be used to model structure crossings within a 2D mesh. As described below, this approach allows runoff to leak through the embankment. When cell edges are situated adjacent to the Page 3-1

26 crest of road embankments, runoff passing through obstructions is prevented, and instead retained behind these embankments as dictated by surrounding topography. The length and cell size of offset breaklines ought to be limited to reasonably model the actual conveyance provided in an event. That is, the amount of runoff allowed to pass through an embankment should be set to avoid significantly overor underestimating water surface elevations and volumes of runoff upstream of obstructions. If the sensitivity of this offset breakline approach has undesirable impacts on model results, an alternative approach should be used, such as approximating culvert structures via an internal connection. Breaklines should be placed along dams, and the cell size for such breaklines may need to be set finer than the nominal 2D mesh cell size. Breaklines along dams can also be converted to internal connections, for which weir geometry can be used to represent the dam. Furthermore, consideration of using the normal 2D equations or weir flow equations should also be made at these features. Generally, the weir flow equations will cause greater instability, and may require finer time steps than may be appropriate for the entire study. However, for internal connections along dams and other features that behave as weirs, the weir equations should at the very least be considered, if not used as a general rule. When using internal connections for other features, the preference is to maintain using the 2D equations, and defining/refining the cell spacing for the breakline in order to achieve reasonable results, specifically WSEL s. If merely using a breakline, the same applies: the breakline cell spacing should be defined such that reasonable results are achieved through the feature. It has been shown that varying these configurations can produce significantly different results, often unreasonable. If a feature that should impede flow lies within a grid cell instead of on the edge, it is possible for water to flow across that feature before it gets high enough to actually overtop the structure (Goodell, 2D Mesh Leaking Part 2 2D Area Breaklines, 2015). Breaklines defined for these features should have cell sizes fine enough to properly represent the feature, and can be user-specified at finer resolutions than the nominal cell size. Figure 6 demonstrates the effects of these refinements. Figure 8: 2D Mesh Refinement After performing a simulation, it should become more clear where significant trouble spots exist, creating artificial backwater or flow area. Care should be taken to refine the 2D mesh at these locations, Page 3-2

27 particularly along streams with SFHAs already defined or with flooding sources identified, such as within the NHD. A balance between refining the mesh to produce satisfactory results for trouble spots and maintaining reasonable effort is obvious; the point of diminishing returns should be relatively easy to identify after a first pass of refining significant trouble areas. Numerous structures in the 2D mesh will be hydraulically insignificant, and placing breaklines at these features should generally be avoided. RAS5 has the ability to include structures within a 2D mesh (via breaklines converted to Internal Connections), such as weirs with culverts or gates that can be used to represent crossings in the 2D mesh that are not represented in the terrain model. Bridges under pressure flow within a 2D mesh are not currently supported by RAS5. However, most bridge openings should generally be captured in a terrain model of 10 foot or finer grid cells. The dimensions of most culvert crossings within a 2D mesh will not often be readily available, or even known, however can be reasonably approximated using engineering judgement. It should be noted that RAS5 will not allow a culvert invert lower than the elevation of the 2D mesh cell(s) to which it is connected. Refining the cell size of the breakline used to generate this internal connection may be an option for resolving this error, as well as terrain model modifications (though not recommended) and simply raising a culvert invert. Figure 9: Internal Connections Levees can be problematic when using the procedures defined in this document, and attention should be given to input and results at and near levees and features that act as levees. However, additional considerations for levees should generally be limited to accredited levees for 2D LSAE using RAS5. Particular attention should generally only be given to accredited levees providing flood protection, and other levee features that produce unreasonable results without additional modeling considerations. Non-accredited levees are to essentially be modeled as a Natural Valley analysis, with no breaklines or internal connections used to define these features. Beyond these recommendations, engineering judgment should drive the assessment and modeling of levees and levee-like features. 3.2 Mapping Post-Processing Summary: Flood risk areas focused on established CNMS or Effective Mapping products Additional raw model outputs retained and categorized for informational purposes Seamless product covering entire project footprint Page 3-3

28 Details: Figure 10: Boundary Creation Workflow Rain-on-grid 2D modeling produces a product that calculates a depth at nearly every cell. While there may be depths associated with numerous raster cells, they may not all necessarily be deemed as a floodplain. These broad overview steps will help filter the raw results down to our proposed flood risk areas. The flood risk polygons will be based off of existing CNMS features and effective floodplains. Initial Automated Mapping Cleanup steps: 1. Convert Depth Raster to polygon feature class or shapefile a. Simplify polygons removes extraneous bends while preserving essential shape b. Smooth polygons smooths sharp angles in polygon outlines to improve aesthetic or cartographic quality; Suggested Smoothing Tolerance = 2.5x input raster cell size (i.e. 10ft raster = 25ft Smoothing Tolerance) 2. Create a scope layer that will be used to identify our extents of proposed flood risk areas a. CNMS Unknown, Unverified, and Unmapped b. Effective flood zones from NFHL or FEMA Q3 digital data Page 3-4

29 3. Use a minimal buffer to select raw model polygons that intersect our scope layer 4. All Additional flood polygons should be retained in a separate layer and should contain a categorization of Greater Than or Less Than 1 foot Manual Mapping Considerations: At this point, the floodplain mapping cleanup becomes much more of a manual process rather than automated. The 2D results can vary from study area to study area due to a number of reasons, and a one-size-fits-all approach cannot always be taken. For example, areas that are extremely flat, have low velocities, and have very shallow depths can create unique challenges. In these situations, if setting the initial depth threshold to 0.1ft can remove a significant amount of noise in a floodplain while retaining the CNMS identified features, then that step can be taken. When the selection of data for the proposed flood risk areas does not capture a portion of the floodplain, a manual selection and incorporating into the proposed flood risk dataset is needed. Typical GIS mapping cleanup processes will be undertaken to address these situations on a case-by-case basis. Aerial imagery will also be used as needed to help support these activities. Similar to other mapping projects, small islands within the mapped floodplain will be filled in as appropriate. Tie Ins: Since the modeling is not done at county wide level, the mapping of work areas will need to be reviewed for proper tie-ins. The floodplain data will ultimately be submitted as a county-wide layer, but should be a seamless feature that will connect across political boundaries. The use of the same county boundary shapefile for these submittal extents should be used. Floodplains from adjacent work areas will need to be checked as well for connectivity and to create the seamless layer. All CNMS Valid Zone AE studies will be identified and the proposed zone A flood risk areas will tie into those areas. The data to tie into will come from NFHL as the top priority, but if not available in that dataset, the FEMA Q3 data should be used. 3.3 CNMS Zone A Validation Summary: To validate the effective Zone A polygons using 2D outputs, the required datasets are: Ground Elevation Grid 1%(+) Water Surface Elevation Grid 1%(-) Water Surface Elevation Grid Effective Zone A Polygons o Use NFHL Data if available. If not, use Q3 or best available digital representation of effective Zone A boundaries The fundamental evaluation procedure is to create a confidence band by using the 1%(+) and 1%(-) values and determining if the effective boundary is mapped within it. Simply put, if the effective Zone A floodplain boundary falls in that range, it passes. If it does not fall within it, it fails. In keeping with the approach outlined in the Automated Engineering Guidance document (May 2016), an allowable horizontal offset tolerance is applied to help in these checks. Details: Create Test points Page 3-5

30 Create Test points along effective zone A boundary lines spaced at an interval of 250ft. (1/4 space on 1 =1000 FIRM panel) To account for an allowable horizontal offset, buffer the point by 75. Create the Confidence Band To create the minimum value of the band, extract the minimum elevation from the 1%(-) grid within each buffered test point and transfer that value to the Test Point To create the maximum value of the band, extract the maximum elevation from the 1%(+) grid within each buffered test point and transfer that value to the Test Point Assign Effective Zone A Boundary Elevation If we assume that the boundary is the same as where the 1% WSE meets the ground, we can use the value from the ground to equal our tested 1% WSE. Extract Ground values and populate that value into the test point. Perform Validation on the Test Points Determine if each test point ground elevation is within the confidence band that was created earlier. If point value is within the band, it passes. If it is not within the band, it fails. Figure 11: Profile Example of Validation Results Page 3-6

31 Report Validation results The spatial test points pass/fail rates can be aggregated and reported on a stream basis for reporting in CNMS. The spatial test points pass/fail rates can also be aggregated and reported on a HUC-12 basis for reporting measures (or perhaps HUC-10 or HUC-8 levels). Page 3-7

32 04 Deliverables County-based production and delivery Terrain DEM file for each County Engineering o HEC-RAS model files subdivided by work areas for each County RAS5 Geometry Flow Files Plan Files Simulation Files o HEC-HMS model files subdivided by work areas for each County o Gage Analysis Reports o Land Cover GIS data by County o 10% WSEL grid o 4% WSEL grid o 2% WSEL grid o 1% WSEL grid o 1%(+) WSEL grid o 1%(-) WSEL grid o 0.2% WSEL grid o 10% depth grid o 4% depth grid o 2% depth grid o 1% depth grid o 1%(+) depth grid o 1%(-) depth grid o 0.2% depth grid o 1% velocity grid o S_FLD_HAZ_AR o S_FLD_HAZ_LN o Changes Since Last FIRM (where available NFHL exists) Report Community proposed flood hazard change maps Page 4-1

33