IMPROVED MODELING OF THE GREAT PEE DEE RIVER: DOCUMENTATION IN SUPPORT OF FEMA APPEAL. Horry County, South Carolina

Transcription

1 IMPROVED MODELING OF THE GREAT PEE DEE RIVER: DOCUMENTATION IN SUPPORT OF FEMA APPEAL Horry County, South Carolina July 15, 2016

2 CONTENTS 1 Introduction Hydrology HEC-RAS Model Cross Section Topography Separation of Individual Channels Updated Water Surface Elevations Conclusions and Recommendations References TABLES Table 1. HEC-HMS hydrologic parameters and computed peak flows... 5 Table 2. Comparison of computed peak discharges at locations indicated in Figure Table 3. Comparison of computed water-surface elevations at locations indicated in Figure FIGURES Figure proposed FEMA flood zones... 2 Figure 2. Horry County sub-basin boundaries... 6 Figure 3. HEC-HMS model domain... 7 Figure 4. Example of unrealistic floodplain representation due to missing topographic data... 8 Figure 5. Example of improved cross-section geometry (XS )... 9 Figure 6. Example of improved cross-section geometry (XS )... 9 Figure 7. Comparison of water-surface profiles using original and updated cross sections Figure 8. Original FEMA HEC-RAS model single-reach configuration Figure 9. Topography of cross section near Bucksport Figure 10. Improved multi-reach HEC-RAS model configuration Figure 11. Computed water-surface elevation comparison locations Figure 12. Comparison of 100-year flood inundated areas arcadis.com i

3 1 INTRODUCTION In September 2015, the Federal Emergency Management Agency (FEMA) updated the Flood Insurance Rate Map (FIRM) for Horry County, South Carolina. To examine the accuracy and applicability of the preliminary FIRM, representatives from Horry County requested that Arcadis assist with an evaluation of the preliminary results. Subsequently, Arcadis identified simplifications to the procedures and models FEMA used for the technical analysis that caused over-estimation of 100-year flood stages within Horry County in comparison to results of a more detailed analysis. According to correspondence from Maria Cox of the South Carolina Department of Natural Resources (SCDNR), the dramatic increase of the base flood elevation (BFE) on the Waccamaw River and the Atlantic Intracoastal Waterway in the new FIRM is due to backwater impacts from the Great Pee Dee River, which had not been calculated as part of the previous FIS for Horry County. The areas that were previously outside the AE zone (100-year floodplain determined by detailed methods) but added to the AE zone in FEMA s new mapping are shown in orange in Figure 1. Note that the new area includes the entire town of Bucksport. Figure proposed FEMA flood zones Note: Graphic provided by Horry County Stormwater (T. Garigen) arcadis.com 2

4 According to Horry County representatives, the town of Bucksport has no historical record of flooding and its inclusion represents a significant change to the regional understanding of flood risk. In consideration of the assertion that this change to perceived flood risk is primarily due to revised hydrologic and hydraulic analysis, including backwater profiles computed using a Hydrologic Engineering Center- River Analysis System (HEC-RAS) model of the Pee Dee river system, Arcadis has analyzed FEMA s data and models to determine if the change to the AE zone is justified. Arcadis was provided with the FIS and supporting documentation from the Hydrology Report, Georgetown County, SC, authored by Christopher Moss of Taylor Engineering. Arcadis was also provided the HEC-RAS models used by FEMA from the Georgetown County FIS that were applied without modification to the Horry County FIS. The HEC-RAS model of the Great Pee Dee River was constructed in two parts; an upstream model (PeeDee_US) and a downstream model (PeeDee_DS). The PeeDee_DS model extends from the upstream end of Winyah Bay to the confluence of the Great Pee Dee and Waccamaw riverine systems. The PeeDee_US model continues from the confluence up to the political boundary between Horry and Georgetown Counties. The two HEC-RAS models cover only a small portion of the lower river system. After thorough review of the FIS and supporting documentation, Arcadis identified major simplifications and errors in both the hydrology and the hydraulic models. These simplifications satisfy the category for basis of appeal as Technically Incorrect Application of the Methodology as defined in 44 Code of Federal Regulations (CFR) 67.6, Basis of appeal from proposed flood elevation determinations. Arcadis performed the following improvements to FEMA s analysis and models: The regional regression equations used to approximate the 100-year discharge were replaced with a distributed hydrologic model that combined and routed computed runoff for 100-year synthetic storms on individual sub-basins. The updated rainfall-runoff analysis was accomplished using the HEC- Hydrologic Modeling System (HMS) model. The steady-state HEC-RAS model was improved by increasing the accuracy of the topographic representation for individual cross sections, by improving the geometric representation of the multiple rivers within the region, and using lateral peak sub-basin inflows computed by HEC-HMS. Given more time, the accuracy of the model could be further improved by distributed (unsteady) flow routing of subarea runoff hydrographs using HEC-RAS, in which case peak flows would attenuate while moving downstream and peak flood profiles would be further reduced as a result. The effects of improved rainfall-runoff modeling using HEC-HMS and improved steady-state HEC-RAS modeling on the computed 100-year water surface elevations are described in more detail below. 2 HYDROLOGY The hydrologic analysis described in Section 3.1 of the FIS is based upon a simple rainfall-runoff model that applied nonlinear U.S. Geological Survey (USGS) regional regression equations (the same relationships originally derived in 1984) defining peak discharge as a function of drainage area. Two relationships were applied: one for the Upper Coastal Plain (UCP), and another for the Lower Coastal Plain (LCP). The total drainage area for Reach 1 of the Pee Dee River Basin was used throughout the analysis. The resulting peak flow for the UCP is 174,028, which jumps by nearly 31,000 cubic feet per second (cfs) arcadis.com 3

5 to 204,896 cfs at the UCP-LCP boundary. Moreover, the peak flows are applied everywhere in the basin simultaneously. Arcadis developed an alternative approach to determination of flood flows in the river system by rainfallrunoff simulation, combining and routing of sub-basin runoff hydrographs using HEC-HMS. The alternative approach makes use of HEC-HMS to determine spatially and time-varying flows for the 100-year rainfall event occurring basin-wide. Eight sub-basins ranging in size between 114 and 8,817 square miles were delineated from the USGS topographical data obtained from USGS, shown in Figure 2. Flow hydrographs for each sub-basin were computed by HEC-HMS using the U.S. Department of Agriculture (USDA) Natural Resources Conservation Service (NRCS) Curve Number Method to determine rainfall losses and runoff relationships for each subbasin. The NRCS Curve Number Method requires input parameters such as sub-basin area, composite curve numbers (CNs), hydrograph type, design storm rainfall depth, basin lag times, and channel routing parameters. Additionally, soil data are required and were obtained from the USDA NRCS Web Soil Survey, and land-use classification information from the National land Cover Database (NLCD). Topographical data were derived from a USGS Digital Elevation Model (DEM). The HEC-HMS model of the sub-basins is shown on Figure 3. Because long-term flow and rainfall records are not available for the area of interest, the NRCS dimensionless unit hydrograph method was used to develop runoff hydrographs for the watershed. The peak flow rate and the time base of the unit hydrograph were scaled for each sub-basin using the sub-basin lag time. The sub-basin characteristics for CN, lag time, area and peak flows are shown in Table 1 below. The Muskingum-Cunge method was used for channel routing, which computes peak outflow based upon an estimated travel time through the routing reach as determined by stream conditions and velocities. The design storm event was used for rainfall inputs into the hydrological model. The 24-hour rainfall depth for the 100-year storm event was obtained from the National Oceanic and Atmospheric Administration (NOAA), distributed over time using the Soil Conservation Service (SCS) Type III Rainfall distribution. For this region, the 100-year 24-hour storm event produces 11.2 inches of rainfall. Peak discharges computed by HEC-HMS for each sub-basin were then laterally input to the improved HEC-RAS model subsequently described, in place of peak flows computed by regional regression equations and uniformly imposed in the FEMA HEC-RAS model. arcadis.com 4

6 Table 1. HEC-HMS hydrologic parameters and computed peak flows Sub-Basin Name Area (square miles) SCS Curve Number Lag time (minutes) Computed peak flow (cfs) W W W W W W W W arcadis.com 5

7 Figure 2. Horry County sub-basin boundaries arcadis.com 6

8 Figure 3. HEC-HMS model domain 3 HEC-RAS MODEL 3.1 Cross Section Topography There are several topographic inaccuracies within the original FEMA HEC-RAS cross sections. It appears that insufficient topographic data were used to generate some of the cross sections. Large extents of the modeling domain were linearly interpolated between sparse data points. As an example, Section 165 (river station ) from the PeeDee_US model shows a linear extrapolation over a distance of more than 20,000 feet (ft) (Figure 4). The cross sections are displayed looking downstream, thus the east is on the left and west on the right. The FEMA model was originally developed for Georgetown County, thus the primary interest for the model was the western side. The eastern side of the model is of noticeably lower resolution than the west. The lower-resolution representation of the eastern floodplain may result in significant increases in computed water surface elevation by unrealistically restricting floodplain storage and conveyance. arcadis.com 7

9 Figure 4. Example of unrealistic floodplain representation due to missing topographic data Arcadis acquired topographic data for Horry County and Marion County from a recent flow Light Detection and Ranging (LIDAR) survey. The topographic DEM for these counties has a resolution of approximately 3 meter (m). LIDAR data for Georgetown were not available, therefore, a 10-m DEM from USGS was used. These data sources were compared to topographic data in the two FEMA HEC-RAS models. Arcadis methodically evaluated all of the cross sections and improved the representation of the cross sections in many locations. It appears that the lower reaches of the Great Pee Dee River were well represented in the original FEMA model, while many of the upstream cross sections were poorly approximated. Examples of the improved topographic representation are shown on Figure 5 and Figure 6. The pink line in these two figures is the original topography and the solid black line represents the updated topography. Note that the right side of the cross section is very similar between original and new, while the left side of the cross section indicates a more realistic representation using the updated topographic data. In many of the upper cross sections, the topographic update results in larger conveyance capacity for a steady-state backwater model and both conveyance and flood storage capacity in unsteady-flow HEC-RAS model. For a specified discharge in a steady-state model, increased conveyance results in lower computed stages at these cross sections. Water surface profiles computed using the original and modified cross sections are compared in Figure 7. Note that the modified cross section reduced computed water surface elevation by as much as 5.72 ft at the upstream portion of the river, where flood plain conveyance was significantly underestimated in the original FEMA model. The impact of these changes diminishes to 1 foot or less for a distance downstream. Below river station 70,000, the difference is negligible, as expected, since the lower cross sections are nearly identical in both the original FEMA and updated model. arcadis.com 8

10 Figure 5. Example of improved cross-section geometry (XS ) Figure 6. Example of improved cross-section geometry (XS ) arcadis.com 9

11 Figure 7. Comparison of water-surface profiles using original and updated cross sections 3.2 Separation of Individual Channels The geometric construction of the original FEMA HEC-RAS model contains geometric simplifications of the Great Pee Dee and Waccamaw riverine systems that impact its ability to correctly represent flows in the multiple river system. The original HEC-RAS model includes cross sections that extend over a width of 120,000 ft and contain multiple channels within a single cross section. Figure 8 provides a plan view of the combined PeeDee_US and PeeDee_DS model domain. The white lines indicate the individual cross sections. Note that many of the cross sections span the Little Pee Dee River, the Great Pee Dee River, the Waccamaw River, and additional tributaries. Several of these rivers are treated as overbank areas and do not have correct Manning s n values. More importantly, HEC-RAS computes a single stage for an entire cross section. The consequence of this modeling simplification is that the model treats all of these distinct rivers as a single entity over which the assigned discharge (flow) is evenly distributed. For instance, note the cross section in Figure 8 that passes through Bucksport. Because of the geometric simplification, the FEMA model forces the stage in the Waccamaw River on the east side of Bucksport to be identical to the stage in both the Little Pee Dee and the Great Pee Dee River on the west side of Bucksport. However, Figure 9 shows the topography for this section, for which topographic separations between channels would cause each channel to convey flow from its own drainage area rather than simply share the combined flow of the Great Pee Dee River drainage basin. Multiple channels exist in all of the cross sections in the FEMA model, as shown above in Figures 5 and 6. This simplification does not accurately compute backwater profiles in individual rivers. The correct approach is to model each channel independently with its own flow arcadis.com 10

12 rate, geometry, and Manning s n values, identify if and where overbank flows may occur, and then determine if backwater from the main channel or flow from tributaries pose the greater flood risk. Arcadis developed a multiple-reach river model using the FEMA HEC-RAS models with updated topography. The model specifies seven individual river reaches and allows for geometry, Manning s n values, and flowrate to be varied as appropriate to each watershed. The multiple-reach HEC-RAS model domain is shown in Figure 10. The cross sections are colored according to the individual model reach in which they are located. The seven reaches correspond to the inflows derived for the sub-basins analyzed with the HEC-HMS hydrologic model described above. Figure 8. Original FEMA HEC-RAS model single-reach configuration arcadis.com 11

13 Figure 9. Topography of cross section near Bucksport arcadis.com 12

14 Figure 10. Improved multi-reach HEC-RAS model configuration arcadis.com 13

15 4 UPDATED WATER SURFACE ELEVATIONS The improved hydrology and hydraulic models produced significant reallocations of flood flows in comparison to the original FEMA study, and significantly reduced computed water surface elevations (WSELs) throughout the combined river system as a result. First, the flow in the mid to lower Great Pee Dee increases because the flow in the new model is not artificially distributed across the Great Pee Dee and Waccamaw rivers. Instead, the great Pee Dee River correctly contains the flow from the upper water shed which it drains. Second, the backwater up the Waccamaw River near Bucksport is decreased and the hydraulic gradient reduced due to a more realistic peak discharge entering the upper Waccamaw. To quantify changes in discharge and WSEL throughout the region of interest, a series of geographical comparison locations are defined and mapped in Figure 11. The improved discharge rates applied have much greater spatial variability than the original approximation. Specific values at comparison locations are listed in Table 2. Updated WSELs corresponding to flood flows computed using the updated hydrologic and hydraulic models model are listed in Table 3. Note in particular that several of the areas of concern, Bucksport (location #7) and the AIWW (location #8) are shown to have WSEL reductions of 43% and 41% respectively. These WSEL reductions significantly reduce the 100-year floodplain in comparison to the FEMA preliminary maps, as shown in Figure 12. Note in Figure 12 that several areas of concern, including Bucksport and Socastee, have significantly less area within the 100- year floodplain. arcadis.com 14

16 Figure 11. Computed water-surface elevation comparison locations arcadis.com 15

17 Table 2. Comparison of computed peak discharges at locations indicated in Figure 11 Location number Original station New station Original peak discharge (cfs) New peak discharge (cfs) 1 Pee Dee DS Lower Waccamaw Pee Dee DS Lower Waccamaw Pee Dee DS Middle Waccamaw Pee Dee DS Middle Waccamaw Pee Dee DS Middle Waccamaw Pee Dee US Middle Waccamaw Pee Dee US Middle Waccamaw Pee Dee US Socastee Pee Dee US Upstream Waccamaw Pee Dee DS Great Pee Dee Downstream Pee Dee DS Great Pee Dee Downstream Pee Dee DS Great Pee Dee Downstream Pee Dee US Great Pee Dee Downstream Pee Dee US Great Pee Dee Downstream Pee Dee US Great Pee Dee Downstream Pee Dee US Little Pee Dee Pee Dee DS Lower Waccamaw arcadis.com 16

18 Table 3. Comparison of computed water-surface elevations at locations indicated in Figure 11 Location number Original station New station Original WSEL (ft) New WSEL (ft) 1 Pee Dee DS Lower Waccamaw Pee Dee DS Lower Waccamaw Pee Dee DS Middle Waccamaw Pee Dee DS Middle Waccamaw Pee Dee DS Middle Waccamaw Pee Dee US Middle Waccamaw Pee Dee US Middle Waccamaw Pee Dee US Socastee Pee Dee US Upstream Waccamaw Pee Dee DS Great Pee Dee Downstream Pee Dee DS Great Pee Dee Downstream Pee Dee DS Great Pee Dee Downstream Pee Dee US Great Pee Dee Downstream Pee Dee US Great Pee Dee Downstream Pee Dee US Great Pee Dee Downstream Pee Dee US Little Pee Dee Pee Dee DS Lower Waccamaw arcadis.com 17

19 Figure 12. Comparison of 100-year flood inundated areas arcadis.com 18

20 5 CONCLUSIONS AND RECOMMENDATIONS Arcadis identified several simplifications in the original FEMA study as they applied to data and models used for the Horry County FIS. The revised analysis documented above addresses some of these simplifications and in the process demonstrates that the FEMA study significantly overestimates 100-year flood stages in portions of Horry County, and that the FIRM needs to be revised accordingly. The improved hydrologic and hydraulic models developed by Arcadis are provided with this report. The updated hydrology and newly computed water surface elevations were compared to the estimates released by FEMA and confirm that the FEMA preliminary FIRM significantly over-estimates flood risk. In addition to model improvements documented in this report, conversion of the HEC-RAS backwater model to an unsteady-flow routing model, utilizing boundary and lateral runoff hydrographs generated by HEC- HMS, offers significant potential for further reductions in 100-year flood flows and flood stages throughout Horry County. Because the original analysis is significantly incorrect in several ways, it is recommended that FEMA update the preliminary FIRM map with new information presented in this report and additional information recommended for development. Horry County and Arcadis are committed to working with FEMA to generate a corrected FIRM as soon as possible. The additional analysis described above (conversion to unsteadyflow hydraulic modeling) can be supplied within the 30-day comment period. arcadis.com 19

21 6 REFERENCES Arcement, G.J. and V.R. Schneider, Guide for Selecting Manning's Roughness Coefficients for Natural Channels and Flood Plains. U.S. Geological Survey Water-Supply Paper 2339, U.S. Geological Survey, Denver, Colorado, Barnes, H.H., Roughness Characteristics of Natural Channels. U.S. Geological Survey Water-Supply Paper 1849, U.S. Geological Survey, Washington D.C., Chow, V.T. Open Channel Hydraulics. McGraw-Hill Book Company, New York, Federal Emergency Management Agency (FEMA), HAZUS: Hazard loss estimation methodology, 2005, gov/hazus/index.shtm. Henderson, F.M., Open Channel Flow, Macmillan Publishing Company, New York, National Land Cover Database (NLCD), Multi-Resolution Land Characteristics (MRLC) Consortium, National Oceanic and Atmospheric Administration (NOAA) Atlas 14, National Weather Service Hydrometeorological Design Studies Center, Precipitation Frequency Data Server (PFDS), U.S. Army Corps of Engineers (USACE) Hydrologic Engineering Center Hydrologic Modeling System (HEC-HMS), Version 4.1., July 2015 USDA Technical Release 55, Urban Hydrology for Small Watersheds, June arcadis.com 20