Thesis. Graduate School of The Ohio State University. Jeremiah Lant, B.S. Graduate Program in Geodetic Science. The Ohio State University

Transcription

1 A Hydraulic Modeling Framework for Producing Urban Flood Maps for Zanesville, Ohio Thesis Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University By Jeremiah Lant, B.S. Graduate Program in Geodetic Science The Ohio State University 2011 Dissertation Committee: Doug Alsdorf, Advisor Mike Durand C.K. Shum

2 Copyright by Jeremiah Lant 2011

3 Abstract This project examines the flooding dynamics of the Muskingum River near the city of Zanesville, Ohio. Simulating various peak flood events using a hydrodynamic model will provide Muskingum County engineers with valuable information regarding inundated areas, extent, and effect on local communities for different flood events. The impact of various Muskingum River flood events, including the 100 year flood, on the urban environment in Zanesville, Ohio is studied. The project provides a useful hydraulic modeling framework that produces urban flooding maps for the city of Zanesville. These flood maps show how water surface elevations and water depths vary spatially and temporally, and provide a more detailed picture of how flood waves move in urban environments. A hydrodynamic model called LISFLOOD-FP is used to simulate river flow and flooding. LISFLOOD-FP is a finite-difference flood inundation model that can accurately model 1D or 2D channel flow along with 2D floodplain flow. LISFLOOD-FP is a well-established hydrodynamic model that has been proven to properly simulate flood inundation for fluvial, coastal, and urban events. Modeling efforts demonstrate that better knowledge of discharge on the Muskingum River provides a valuable insight into how floods travel through the floodplain. The Federal Emergency Management Agency, FEMA, conducts flood insurance studies to identify a community s flood risk. The flood risk study is based upon statistical data for river flow, storm tides, hydrologic and hydraulic analyses, and rainfall and topographic surveys. The FEMA maps only provide a one-time snapshot of a flood, and do not describe the full extent of the flood event including the spatial and temporal variability ii

4 of various flood events. Questions, such as the changes in flood inundation extent with time for the city of Zanesville, cannot be fully explored using the FEMA maps. It has been shown that accurate mapping of urban flooding events must take hydraulic connectivity and mass conservation into account (Bates, et al., 2005). In other words, extending potential flood elevations along lines of equal elevation given a river elevation, the so-called Planar GIS method, may be inadequate for characterizing urban flooding. An alternative approach involves the simulation of hydraulic processes, which would control flooding and inundation patterns in downtown Zanesville given the FEMA 100 year Muskingum River main stem water surface elevation. Such an approach provides the framework, not only for producing dynamic maps of different frequency flood events for the city of Zanesville, but also evaluating the impacts of adding and/or removing structures or changing land use on urban flooding. Model simulations of the 100 year flood defined by FEMA have been created as part of this research. In addition, a 1D HEC-RAS model was built to compare 100 year flood profiles with the 2D LISFLOOD-FP model. The LISFLOOD-FP model agrees with the FEMA flooding map in spatial extent, and in river height predictions to within 1 foot. A 1-D HEC-RAS model also agrees with both the FEMA and LISFLOOD-FP model predictions. Sensitivity tests indicate that a flood wave with a flow rate of 100,000 cubic feet per second would be required to inundate downtown Zanesville. The study of urban flooding on the Muskingum River also represents an opportunity to more fully understand the performance of the upcoming Surface Water Ocean Topography Mission, SWOT, over modest sized rivers in an urban environment. The iii

5 SWOT satellite will measure water surface elevations and inundated areas for fresh water bodies around the world. The importance of flood events, like the 100 year flood, to river communities lies in the understanding and awareness of how peak floods occur and travel, especially in urban environments. From this, steps towards flood control and damage prevention can be made. The fully parameterized Muskingum River model will allow us to (1) produce urban flooding maps for Zanesville, (2) build a framework for sensitivity studies on impacts from urbanization and climate change, and (3) predict inundation throughout Zanesville, a city of 25,000 people. Key Words: hydrodynamic model; LISFLOOD-FP; Muskingum River; urban flooding; FEMA iv

6 Dedication Dedicated to the students at Ohio State University v

7 Ackowledgements Thank you to my advisor Prof. Doug Alsdorf, Prof. Mike Durand, and Dr. Kostas Andreadis for your wisdom and guidance. Your knowledge, insight, and expertise have made this research possible. I have learned so much from each of you. vi

8 Vita 2006 B.S. Physics, University of Kentucky 2009 to Present...Graduate Research Associate Department of Earth Science, The Ohio State University Major Field: Geodetic Science Fields of Study vii

9 Table of Contents Abstract... ii Dedication... v Acknowledgments... vi Vita... vii List of Tables... ix List of Figures... x Chapter 1: Introduction... 1 Chapter 2: Methods... 6 Chapter 3: Research Area and Model Simulations Chapter 4: Results and Discussion Chapter 5: Conclusion References viii

10 List of Tables Table 1. Range of Friction Coefficients used by FEMA Table 2. Comparison of Water Surface Elevations ix

11 List of Figures Figure 1. Muskingum River Watershed... 2 Figure 2. Flood Control Structures... 3 Figure 3. FEMA Flood Map of Downtown Zanesville, Ohio... 5 Figure 4. Aerial Image of Research Area Figure 5. Sample Model Inputs Figure 6. Processed Digital Elevation Model of Research Area Figure 7. Channel and Floodplain Friction Coefficient Map Figure 8. Input Hydrographs for FEMA 100 Year Flood Simulation Figure 9. Triangular Irregular Network Map Figure 10. Water Surface Elevation Maps of FEMA 100 Year Flood Simulation Figure 11. Water Depth Maps of FEMA 100 Year Flood Simulation Figure 12. Flood Extent Comparison between FEMA and LISFLOOD-FP Figure 13. HEC-RAS Outputs of FEMA 100 Year Flood Simulation Figure 14. Comparison of Water Surface Profiles Figure 15. Multiple Linearly Scaled Hydrographs Figure 16. Water Depth Maps Figure 17. Water Depth Map of Major Flood Inundation in Zanesville, Ohio Figure 18. Input Hydrographs for Streamstats 100 Year Flood Simulation Figure 19. Water Depth Maps of Streamstats 100 Year Flood Simulation x

12 Chapter 1: Introduction Flooding continues to be a major natural disaster across the United States, capable of great damage and loss to property and life that can occur despite engineering efforts to control and diminish their damaging effects. Midwest regions, including Ohio, are subject to significant flood events due to its geographical location and topography. The modest sloped region is prone to floods caused by general heavy rainstorms, snowmelt, and hurricane aftermaths. Since January 1, 1964, Ohio has federally declared disasters costing $622.2 million, excluding insurance claims (Ohio Emergency Management Agency, 2009). According to the Ohio Department of Natural Resources, ODNR, flooding occurs every year in Ohio (ODNR, 2005). Unfortunately, large floods throughout the past century in Ohio, such as those in 1913, 1937, 1959, 1963, 1964, 1969, 1990, 1997, and 1998, have resulted in many lost lives as well as billions of dollars worth of property damage (ODNR, 2005). The Muskingum River is a tributary of the Ohio River, formed by the confluence of the Tuscarawas and Walhonding rivers in Coshocton County. The river follows a meandering course for approximately 111 miles until it meets the Ohio River in the city of Marietta, Ohio. The river flows past many cities, including Dresden, Zanesville, and McConnelsville. The Muskingum River Watershed is the largest watershed in Ohio, draining 8,036 square miles or approximately one fifth of the state (Figure 1). Along its main stem and connecting tributaries, the Muskingum River has a system of sixteen dams to provide water resources and protection against flood events to neighboring communities (Figure 2, U.S. Army Corps of Engineers, 2006). Much of the system of flood control reservoirs were 1

13 built in the early 1900s, and are operated and managed in a joint partnership between the U.S. Army Corps of Engineers and the Muskingum Conservatory District. Despite the system of dams, this study investigates whether a peak 100 year flood could inundate the city of Zanesville which is located along the banks of the Muskingum River. Most engineering designs are based upon a flood frequency of 100 years. However, these designs may need to be updated to support flood frequencies greater than 100 years because climate change is causing the global water cycle to intensify resulting in an increase of frequency of great floods (Milly et al., 2001). Changing climatic patterns call for more flood studies that are a physically based representation of hydraulic processes that govern and control flooding. The Federal Emergency Management Agency, FEMA, conducts flood insurance studies to identify the flood risk of the 100 year Figure 1. Muskingum River Watershed. (Musser, 2007) flood event on a community. The 100 year flood event is a disaster that has a chance of occurring once in a period of hundred years, or has a one percent chance of occurring in any given year. The flood risk study consists of statistical data for river flow, storm tides, hydrologic and hydraulic analyses, and rainfall and topographic surveys. For example, Figure 2

14 3 shows a flood risk map for downtown Zanesville, due to flood risk on both the Muskingum and Licking River. FEMA uses this data to create the flood hazard maps that outline a community s different flood risk areas. Since floodplain boundaries are affected by changing weather patterns, erosion, and new developments, FEMA is in the process of updating and modernizing the nation s Flood Insurance Rate Maps based upon the new factors influencing floodplain boundaries. These digital flood hazard maps provide an official depiction of flood hazards for each community and for properties located within it. Flood management uses these maps that are based on a flood frequency analysis. The maps provide water surface elevations for different frequency floods. However, the FEMA maps only provide a onetime snapshot of a flood event, and do not describe the full extent of the flood event including the spatial and temporal variability of various flood events. Specifically, for the city of Zanesville, Ohio, the implications of flooding in the 100 year Figure 2. Sixteen flood control reservoirs that protect the city of Zanesville (U.S. Army Corps, 2006). flood based on FEMA water surface elevations are not fully explored. In addition, questions as to what parts of the city of Zanesville will flood and for how long are not fully answered. A Planar GIS method could be applied that is based on one elevation or contour; however, this method ignores hydraulic connectivity and does not conserve mass (Bates et al. 2005). 3

15 A hydraulic model of the inundation of downtown Zanesville given the FEMA 100 year main stem water surface elevation is required. The main project objective was to examine the impact of various peak Muskingum River flood events on the urban environment by providing a framework that produces urban flooding maps for the city of Zanesville, Ohio. The project includes a strong emphasis on the 100 year Muskingum River flood event. Simulations of these flood events is completed using a hydrodynamic model that provides Muskingum County engineers and water managers with information regarding inundated areas, extent, and effect on local communities. Flood maps from this study have added a new dimension to the static FEMA maps. The maps from this study add a dynamic picture of water surface elevations and water depths based upon any flood frequency. These maps show how water surface elevations and water depths change spatially and temporally, and provide a more detailed picture of how flood waves move in urban environments. This study has provided flood maps that conserve mass and account for connectivity in the city of Zanesville, Ohio. The maps show what specific areas in the urban environment of Zanesville, Ohio will flood and for how long. In addition, a framework for future studies regarding sensitivity to climate change, urbanization, new structures, and land cover changes has been created. Ultimately, the model framework built could probe the sensitivity of the FEMA 100 year water surface elevations to climatic change and urbanization. A follow-on study could use this framework to explore the sensitivity of Zanesville flooding patterns to climate change and urbanization. 4

16 Figure 3. FEMA map of downtown Zanesville, Ohio 5

17 Chapter 2: Methods Flood inundation on a floodplain is controlled by the floodplain topography and friction. Such flow is spatially complex, especially in the urban environment, with varying patterns of water velocity and depth that are two dimensional in space and dynamic in time. The creation of flood maps of water surface elevation and depth that provide a dynamic picture of flood inundation in the urban environment require a two dimensional hydrodynamic model. The LISFLOOD-FP hydrodynamic model is a two dimensional storage cell hydrodynamic model based on a finite difference scheme that can accurately simulate floodplain inundation in urban environments (Bates and De Roo, 2000; Trigg et al., 2009; Bates et al., 2010). The purpose of the LISFLOOD-FP code is to help improve understanding of flood hydraulics, flood inundation prediction, and flood risk assessment. Flood events that occur in urban environments contribute most to overall flood risk since most property and life exist in these city environments. Currently, there has been a need for computationally efficient urban flood models for which flood risk can be analyzed. The LISFLOOD-FP code has been shown to be computationally efficient and to yield good predictions of maximum inundation extent for fluvial flooding problems (e.g. Bates et al., 2005, Bates and De Roo, 2000; Horritt and Bates, 2001a, b, 2002). A computationally efficient model is one that is capable to simulating dynamic flood inundation in a simple process. One of the simplest processes has been shown to be a one-dimensional kinematic wave approximation for channel flow solved using an explicit finite difference scheme and a two-dimensional diffusion wave representation of floodplain flow (Bates and De Roo, 2000). 6

18 LISFLOOD-FP uses this process and has been shown to outperform more complex and simpler flow and flooding processes (Bates and De Roo, 2000). The LISFLOOD-FP hydrodynamic model is based on a raster grid with the capability of being used with a 1D/2D coupling between the channel and adjacent floodplain or with a complete 2D treatment of the channel and overlying floodplain. Since flood inundation is highly dependent upon floodplain topography, LISFLOOD-FP is designed to integrate high-resolution raster DEMs which continue to become increasing available from sources of terrain data collected from remote sensing techniques. The model uses a volume-filling process based on hydraulic principles with mass conservation and hydraulic connectivity. LISFLOOD-FP handles channel flow using a one dimensional or a two dimensional approach that is capable of capturing a flood wave propagating downstream and responding to free surface slope. These approaches are described in terms of 1D or 2D continuity and momentum equations. In 1D, the continuity and momentum equations are as follows: [ ] Q is the volumetric flow rate in the channel, x is the distance between cross sections, A is the cross sectional area of the flow, t is the time, q is the flow into the channel from the floodplain or tributary channels, S 0 is the channel bed slope, n is Manning s friction 7

19 coefficient for the channel, P is the wetted perimeter of the flow, and h is the flow depth. For the 1D coupled 2D version, channel flow is handled using a 1D diffusive wave solved using an explicit finite difference procedure, and floodplain flow handled using a 2D diffusive wave. For the full 2D version, both channel flow and floodplain flow are handled using a 2D diffusive wave. Given two cross-sections Δx distance apart, equations (1) and (2) can be solved numerically using an explicit finite difference procedure to yield the discharge Q and the cross-sectional area A, from which water depth h can be found. The term in brackets in equation (2) is the diffusion wave term. This term forces the channel flow to respond to the bed slope and the free surface slope. LISFLOOD-FP allows for this term to be switched on or off in the model to allow for kinematic and diffusion wave approximations to be tested. LISFLOOD-FP assumes a shallow rectangular channel where the wetted perimeter can be approximated by the channel width. In the 1D/2D coupled mode the river channel is discretized as a single vector along its centerline which is separate from the overlying floodplain raster grid. At each point along the channel vector, channel parameters including width, Manning s n value, and bed elevation can be specified. From the combination of the floodplain DEM and channel vector, the bed slope and the bankfull depth can be calculated. The model linearly interpolates between each point along the vector, and the interpolated channel vector is used to identify cells in the overlying floodplain grid that have a channel beneath them. When the diffusive wave approach is used, a down reach slope is allowed to be positive if the raster Digital Elevation Model, DEM, grid has a varying down reach gradient. In full 2D mode, the river channel is cut into the floodplain DEM. The channel parameters required include bed elevation, channel slope, and Manning s n value. 8

20 Once bankfull depth is exceeded in a given channel cell, water is transferred onto the adjacent floodplain areas of the DEM. Floodplain flows are also described in terms of continuity and momentum equations discretized over a grid of square cells, which allow the model to represent two dimensional flows on the floodplain. When water is on the floodplain, each grid cell is treated as a storage volume, and so the change in cell volume over time is equal to the fluxes into and out of it during the time step. LISFLOOD-FP assumes that the flow between two cells is a function of free surface height difference between those cells (Bates et al., 2005). The following equation describes the process: Where V is the cell volume, t is the time, and Q up, Q down, Q right, Q left are the flow rates from upstream, downstream, right adjacent cell, and left adjacent cell. The flow rates between each cell can then be calculated using Manning s uniform flow formula. So the flow rate between two adjacent cells i and j, where i is the upstream cell, is: Where Q is the flow rate between cells i and j, A ij is the cross-sectional area at the interface of the two cells, R ij is the hydraulic radius at the interface of the two cells, S ij is the water surface slope between the two cells, and n is Manning s friction coefficient on the floodplain. 9

21 In addition to the kinematic and diffusive formulations, LISFLOOD-FP has been updated to include adaptive time stepping with an inertial formulation of the shallow water equations (Bates et al. 2010). This functionality was added to LISFLOOD-FP in order to improve model stability and model run time for high resolution (1 10 meter) DEM datasets. For the urban environment, the inertial formulation performs as well as the diffusive formulation, but with significant reduction in model run time (Bates et al. 2010). The minimum stable time step scales with 1/ x with the inertial equation instead of (1/ x) 2 with the diffusive equation allowing for stable time steps that are 1-3 orders of magnitude larger. Equation 5 shows the inertial formulation with the added term to the diffusive model in brackets. [ ] The model framework built in this study, uses the adaptive time stepping with the inertial formulation. 10

22 Chapter 3: Research Area and Model Simulations The research area consists of a reach of the Muskingum River beginning on the north end of the city of Zanesville, Ohio and ending in the south end of the city which includes one major tributary, the Licking River (Figure 4). The Licking River joins the Muskingum River in Zanesville. The research area is approximately 6 square miles, or 16 square kilometers, sitting within Muskingum County. The research area consists of one gage south of Zanesville s famous Y Bridge that spans the confluence of Muskingum and Licking rivers. There is an upstream gage Figure 4. Aerial image of research domain. approximately 16 miles north in Dresden, Ohio and a gage approximately 6 miles northwest on the tributary. A small eight foot dam lies just north of the Y Bridge, and a canal runs adjacent to the Muskingum River for approximately three fourths of a mile. 11

23 Datasets and Inputs The main inputs required for LISFLOOD-FP to simulate a flood event include a high resolution DEM, time-variable discharge, and channel geometry. Figure 5 shows some of the available datasets and inputs used to configure the LISFLOOD-FP hydrodynamic model for an urban flooding scenario. A 5 meter DEM derived from a 2.5 foot LIDAR dataset from the Ohio Geographically Referenced Information Program, OGRIP, was used (OGRIP, ogrip.oit.ohio.gov, 2010). Muskingum River cross-sections are from 1934 Army Corps of Engineer maps (Reed, 2009). These maps were acquired from the Ohio Department of Natural Resources. Average bed elevations at each data cross-section were computed within the river channels. Boundary conditions consisting of the 100 year flood discharge, water surface elevations, and flood profiles were taken from the FEMA Flood Insurance Study for the city of Zanesville. 12

24 Figure 5. The 2.5 ft bare earth DEM from OGRIP (top), and a Muskingum River cross section map from ODNR laid on top of the DEM (bottom). 13

25 1D coupled 2D Mode versus Full 2DMode LISFLOOD-FP has the capability of being run in 1D coupled 2D mode or in full 2D mode. In 1D coupled 2D mode, the flow within the channel is treated as a 1D diffusive wave and flow spilling into the floodplain is treated as a 2D diffusive wave. LISFLOOD-FP was run in both modes for the city of Zanesville, with the full 2D mode performing more properly than the 1D coupled 2D mode. The full 2D model worked best for the Zanesville area because the 2D model physics handled the high resolution DEM dataset better. In the DEM, there are a large number of raster cells that make up the channel in the high resolution DEM. In 1D coupled 2D mode, LISFLOOD-FP treats raster cells within the actual channel as floodplain cells instead of channel cells because LISFLOOD-FP creates a 1D channel vector along the length of the river reach consisting of a line of raster cells. With the high resolution DEM having tens of cells within the channel, the 1D channel vector in the 1D coupled 2D model improperly treats cells adjacent to the 1D vector cells as floodplain cells. This causes LISFLOOD-FP to treat most of the channel as a floodplain which results in incorrect water depths and flooding extent. The full 2D model was used to produce the urban flooding maps. DEM processing On the original DEM, raster cells within the channel consisted of water surface elevation values. For the full 2D model simulation, the raster cells with water surface elevation values were lowered to the corresponding average bed elevations from the ODNR maps using ESRI ArcMap software. To alter the original raster cell values, polygons were first created within each river channel, each given a respective average bed elevation value. 14

26 The polygons were converted into rasters, and the channel raster cell values were changed using raster algebra techniques given the river channels a rectangular cross-section. For an urban flooding scenario, building heights for the city of Zanesville, acquired from the Muskingum County engineers office, were added onto the high resolution bare earth DEM using the same techniques to change the channel cell values. The original 2.5 ft resolution DEM was resampled to 5 meters. For an urban flooding scenario, the DEM resolution should be approximately the same size as the street widths and building lengths (Fewtrell, et al., 2008). At 5 meters, the DEM resolution is high enough to capture flood wave moving down city streets and between city buildings. Figure 6 shows the final processed DEM. Channel Slope Calculation The channel slope within the research domain was calculated from the original 2.5 ft DEM before preparations for the 2D simulation were made. The main stem flow path from the top to the bottom of the DEM domain was measured to be 5,486 meters, and the high and low elevations were measured to be meters and meters respectively. This results in a slope of , approximately 3 ft/mi. 15

27 Figure 6. The final processed DEM. 16

28 Channel and Floodplain Friction Values LISFLOOD-FP has the capability of having a spatially variable friction value. Using the hydraulic roughness coefficients, Manning n values categorized by Ven Te Chow (Chow, 1959), the Muskingum main stem and Licking tributary channel friction coefficient value was chosen to be 0.04 and the respective adjacent floodplain friction coefficient value to be The Muskingum and Licking River can be described as natural, clean, and winding with some pools and shoals. According to Chow, this type of river channel should have a normal friction value of The floodplains adjacent to the Muskingum and Licking River can be described as having light brush and trees. According to Chow, this type of floodplain should have a normal friction value of The FEMA flood insurance study for Zanesville, Ohio has a range of Manning n values for the Muskingum main stem and Licking tributary along with a range of friction values for the floodplain (FEMA, 2010). However, the FEMA flood insurance study does not write where along the channel or where in the floodplain these friction coefficient values are used. Table 2 shows the range of Manning n values used by FEMA: Table 1. Manning n value ranges used by FEMA. 17

29 The friction coefficients chosen for the river channels and floodplain from Chow do coincide with the range of values used by FEMA in the FIS. Figure 7 shows the Manning s spatially variable friction map. Boundary Conditions From the Federal Insurance Study (FIS) conducted for the city of Zanesville, FEMA defined the 100 year flood discharge along the Muskingum River near the city of Zanesville, Ohio to be 68,000 cubic feet per second, and the 100 year flood discharge along the Licking River below the Dillon Dam to be 7,200 cubic feet per second. To show the spatial and temporal evolution of a possible 100 year flood for Zanesville, hydrographs from the USGS for the Muskingum River and Licking River with peaks nearest the FEMA defined 100 Figure 7. Variable friction coefficient map. Areas in white define the river channel friction coefficient of 0.04, and areas in black define the floodplain friction coefficient of year flood discharge were found and used (USGS Gage Numbers: , ). These hydrographs consisted of a twelve day flood event. Each hydrograph was linearly scaled to match the respective FEMA peak. In LISFLOOD-FP, the flux of this discharge is 18

30 distributed across each of the river channel cells at the head of the Muskingum River main stem and Licking River tributary. The downstream boundary condition consisted of a free surface slope. Figure 8 shows the input hydrographs for the Muskingum River and Licking River. Before simulations of the FEMA 100 year were conducted, a triangulated irregular network (TIN) surface was generated from the DEM to get an intuitive sense of where major flood waters would travel based on the floodplain topography. Figure 9 shows the generated TIN surface. From Figure 9, the darkest colored areas give a sense of where major flood waters on the floodplain could be. 19

31 Figure 8. Scaled input hydrographs with Muskingum River (top) and Licking River (bottom). 20

32 Figure 9. TIN surface of research domain. Darkest areas show an approximate first look of where a major flood event could occur. Figure 9. TIN surface of research domain. Darkest areas show an approximate first look of where a major flood event could occur. 21

33 Chapter 4: Results and Discussion FEMA 100 Year Flood Simulations FEMA maps provide a static picture of water surface elevations based upon a 100 year flood frequency (Figure 3). In contrast, the maps produced during this study provide value added to the FEMA maps by i) taking hydraulic connectivity into account, ii) ensuring that mass is conserved, and iii) providing a temporally dynamic picture of water depths in Zanesville in the event of a 100 year flood. These maps show how water surface elevations and water depths vary spatially and temporally, and provide a more detailed picture of how flood waves move in urban environments. The maps produced by the model show what specific areas in the urban environment of Zanesville, Ohio will flood and for how long they remain inundated. In addition to the reconstruction of major flood events, and the production of flooding hazard maps, a deliverable of this study, the framework is built for future flooding studies regarding the potential sensitivity of changes in flooding patterns in Zanesville to factors such as land cover/use changes, urbanization, construction of new structures, and climate change. The model setup is general enough that a Zanesville flooding scenario could be explored given any hydrograph on the Muskingum River. For example, a follow-on study could use this framework to explore the sensitivity of Zanesville flooding patterns to climate change and urbanization at some future time. Hydrologic models would be used to generate the river hydrograph given current levels of dam management. The effect of a given scenario on Zanesville flooding could then be quantified. Such a hydraulic 22

34 modeling framework should prove valuable to the City and County water management offices. Water surface elevation maps, Figure 10, of the dynamic FEMA 100 year flood simulation show how a flood wave would move through the research domain. Each map represents one day in the twelve day hydrograph used to simulate the FEMA 100 year flood. As the maps show, the flood wave on the Muskingum River is mainly contained within the channel throughout the 100 year flood event. Near the top of the domain, flood water on the Muskingum River does spill onto the floodplain. The low lying floodplain near the top of the domain is a small area of a larger park named Jaycee Riverside Park. The section of the park that would be inundated consists of a parking lot, a small building, and a grass field. The flood wave on the Table 2. Comparison of water surface elevations. Licking River does exceed the river banks approximately 1.25 miles upstream of the confluence with the Muskingum River. The inundated area mainly consists of low lying woodlands, several small buildings, and grass fields. Figure 11 shows a full extent water depth map of the flood peak along with a close up of the inundated area on the Licking River. The LISFLOOD-FP simulated flood extent shows good agreement with the FEMA flood extent boundary (Figure 12). Within the research domain, the FEMA has seven water surface elevation values on the Muskingum River and two water surface elevation values on 23

35 the Licking River. Table 1 shows the good agreement between LISFLOOD-FP water surface elevation values and the documented FEMA water surface elevations. The water surface elevation and water depth maps show that the FEMA defined 100 year flood does not flood the city of Zanesville. The large number of flood control reservoirs within the Muskingum River basin protect the city of Zanesville from FEMA defined 100 year flood. The FEMA defined 100 year flood discharge accounts for the flood control reservoirs. Without this full system of dams, the 100 year flood discharge would be much higher. The purpose of such a system of flood control reservoirs was to prevent excessive flood damage due to the 100 year flood. The LISFLOOD-FP results show that the flood control reservoirs do protect the city of Zanesville from excessive flood damage. The flooding due to the FEMA 100 year flood is in low lying areas without a large number of buildings. 24

36 Figure 10. Successive daily water surface elevation maps of the FEMA defined 100 year flood. The elevation range is from 214 meters (blue) to 204 meters (red). 25

37 Figure 11. Water depth maps of FEMA 100 year flood event. The bottom image shows flooding on the Licking River with wooded areas, few buildings, and grass fields being inundated. 26

38 Figure 12. LISFLOOD-FP flood extent shows good agreement with FEMA 100 year flood extent (red). Water surface elevation values are in meters. 27

39 HEC-RAS Simulation The U.S. Army Corps of Engineer s River Analysis System (HEC-RAS) is a hydraulic model used to simulate river flow (U.S. Army Corps, 2010). HEC-RAS performs onedimensional steady and unsteady flow calculations for a full network of natural and constructed channels. For steady flow, HEC-RAS solves the one-dimensional energy equation. For unsteady flow, HEC-RAS solves the full dynamic St. Venant equation using an implicit finite difference scheme. A HEC-RAS model of the research domain was created to compare water surface elevation profiles between the 2D LISFLOOD-FP and the 1D HEC-RAS. Due to its onedimensional nature, HEC-RAS is very accurate at computing water surface elevations within a river channel. However, when the water surface elevation within a channel exceeds the bank elevation, water on the floodplain is not controlled by the overlying topography, as treated by LISLOOD-FP. The water in the floodplain will have the same water surface elevation as the channel, which is the so-called planar GIS method. The geometry used for the HEC-RAS simulation is the same geometry used for the LISFLOOD-FP simulation. A steady flow analysis was performed as a first step using the FEMA defined 100 year peak discharge for the Licking and Muskingum Rivers. Figure 13 shows some sample HEC-RAS output. In comparing the 2D LISFLOOD-FP and 1D HEC-RAS results, the water surface elevations within the river channels are similar. The channel water surface elevations produced by LISFLOOD-FP are slightly higher on the Muskingum River and slightly lower on the Licking River compared to HEC-RAS channel water surface elevations. On the Licking River where flooding occurred approximately 1.25 miles above the confluence of the Muskingum River, the HEC-RAS water surface elevation in the floodplain is the same as the 28

40 water surface in the channel; meters. However, the LISFLOOD-FP water surface elevation in the floodplain is slightly lower; meters. Because LISFLOOD-FP has 2D capability, the floodplain flow follows the topography in the floodplain. In that same area, the floodplain elevation drops below the bank elevation which accounts for why the LISFLOOD-FP water surface elevation is slightly lower than the HEC-RAS water surface elevation in the floodplain. The planar GIS method used by HEC-RAS to flood the floodplain is why the water surface on the floodplain is increased in the HEC-RAS model. The difference in water surface elevations in the floodplain is 0.33 meters, or approximately 1 foot. Figure 14 shows a comparison of water surface profiles between FEMA, LISFLOOD-FP, and HEC-RAS. Figure 13. Sample outputs from HEC-RAS simulation of the peak FEMA 100 year discharge; multiple cross-section plot (upper left), cross-section plot of Muskingum River above confluence with Licking River (lower left), Muskingum River profile (upper right), and Licking River profile (lower right). 29

41 Figure 14. Comparison of water surface profiles. 30

42 What discharge would cause major flooding in Zanesville, Ohio? The sixteen flood control structures within Muskingum watershed basin have a strong influence on whether the city of Zanesville will flood. The FEMA defined 100 year flood discharge does not flood the large areas of the city of Zanesville. However, the question as to what discharge amount, despite the system of flood control reservoirs, will cause major flooding in the city of Zanesville remains. In order to answer such a question, the hydrographs for the Muskingum and Licking Rivers, used for the FEMA 100 year flood simulations, were increased successively by ten percent until the FEMA 100 year flood discharge was doubled (Figure 15). LISFLOOD-FP was used to create maps of the peak flood at each increased level. The maps were analyzed to find the discharge amount that would cause flood waters to inundate the center of Zanesville. Figure 16 shows the results in a series of water depth grid images. The water depth grid images are oriented in increasing peak flow order from left to right. Figure 17 illustrates that Zanesville has major inundation at approximately a 50% increase in the FEMA flood peak. A 50% increase in the FEMA flood peak is equivalent to a discharge near 102,000 cubic meters per second. To flood the heart of the city of Zanesville, the Muskingum River would have to have flow larger than the annual average flow on the Ohio River. 31

43 Figure 15. Multiple linearly scaled hydrographs of the FEMA defined

44 Figure 16. Water depth output of each increasing flood peaks. Each successive increase is oriented from left to right with the starting FEMA water depth being the upper left most map. The output of increasing the FEMA flood peak by 50%, outlined in red, shows when downtown Zanesville is first heavily inundated. 33

45 Figure 17. Water depth map of increasing FEMA flood peak by 50%. Water depths are in meters. The city of Zanesville is inundated with water depths ranging from 0.1 meters to 1.5 meters. 34

46 How would the flood impact be different if there were no flood control structures protecting Zanesville, Ohio? Streamstats is a web based geographical information system developed by the United States Geological Survey that allows users to obtain river flow statistics and drainage basin characteristics. River information from Streamstats is made available for many states including Ohio. Streamstats defines the 100 year flood discharge without the influence of the flood control structures. Streamstats uses basin characteristics and a log-pearson Type III distribution to define the 100 year flood on the Muskingum River and Licking River (Koltun, et al., 2006). Without flood protection structures, the 100 year flood discharge would be much greater. Streamstats states that the discharge on the Muskingum River and Licking River as 141,000 cfs and 63,700 cfs respectively. The Streamstats discharges are 2.6 times and 12.8 times larger than the FEMA discharges on the same rivers. With the model framework built, LISFLOOD-FP was used to simulate a 100 year flood event without the protection of 16 control structures using discharge amounts defined by Streamstats. Figure 18 shows the linearly scaled hydrographs used to simulate a Streamstats defined 100 year flood. Figure 19 shows a water depth map of the peak flow. The map shows the city of Zanesville severely inundated. These results emphasize how well the city of Zaneville, Ohio is protected from major flooding events. The large number of flood control structures plays a vital role in protecting Zanesville, Ohio from flooding. 35

47 Figure 18. Input hydrographs for the Streamstats defined 100 year flood simulation. Streamstats defines the 100 year flood based on basin characteristics without considering the influence of flood control structures. 36

48 Figure 19. Output water depth map of Streamstats defined 100 year flood simulation. 37

49 Chapter 5: Conclusion The objective of this research project was to create an urban flood study using a 2D hydrodynamic model, LISFLOOD-FP, for the city of Zanesville, Ohio. The study produced a model framework that yields dynamic urban flood maps of Zanesville. The framework was built around simulating a FEMA defined 100 year flood. Modeling efforts demonstrate proper flood profiles and water surface elevations when compared to FEMA. The flood maps show how a 100 year flood wave evolves over time. A comparison study was made between 1D HEC-RAS and 2D LISFLOOD-FP models. Results have shown that both models produce comparable water surface elevations within the river channel and on the floodplain. With the model framework built, simulations of other flood events can be completed. Since the heart of Zanesville is fairly protected from the 100 year flood, an investigation into the amount of discharge needed to reach downtown Zanesville was conducted. It was found that a massive flood wave, over 100,000 cfs, might be needed to inundate the entire downtown area of Zanesville. Although highly improbable, the dynamic mapping of this flood event provides a deeper understanding of how flood waters could move in the urban environment. A flood event case with data from USGS Streamstats was completed to analyze how a flood might impact Zanesville without the influence of any flood control structures. Without such flood control structures, Zanesville would be seriously inundated if a 100 year flood occurred. The model framework built in this study allows the sensitivity of climate change and urbanization on the FEMA 100 year water surface elevations and extent to be analyzed. A 38

50 follow-on study could, ideally, use this framework to further explore changing flooding patterns in an urban environment due to sedimentation, land use and land cover change, climate change, and urbanization. This study of urban flooding on the Muskingum River also represents an opportunity to more fully understand the performance of the upcoming Surface Water Ocean Topography Mission, SWOT, over modest sized rivers in an urban environment. The SWOT satellite will have the capability of measuring temporal and spatial changes in water surface elevations and inundated areas for fresh water bodies around the world. From these measurements, depth and discharge along much of the Muskingum River can be extracted and used in hydrodynamic models like LISFLOOD-FP. Better knowledge of discharge on the Muskingum River will provide a valuable insight into how floods travel through the floodplain and affect the urban environment. 39

51 References Bates, P. and De Roo, A.P.J. (2000). A simple raster-based model for flood inundation simulation. Journal of Hydrology. 236(2000): Bates, P., Dawson, J., Hall, J., Horrit, M., Nicholls, R., Wicks, J., and Hassan, M. (2005). Simplified two-dimensional numerical modeling of coastal flooding and example applications. Coastal Engineering. 52(2005): Bates, P., Horritt, M., Fewtrell, T. (2010). A simple inertial formulation of the shallow water equations for efficient two-dimensional flood inundation modeling. Journal of Hydrology. 387(2010): Edelen, G. W., F. H. Ruggles, W. P. Cross (1964). Floods at Zanesville, Ohio. USGS Hydrologic Atlas Report Number 46. Federal Emergency Management Association. Muskingum County, Ohio. Flood Insurance Study Number 39119CV000A Fewtrell, T.J., Bates, P., Horrit, M., and Hunter, N.M. (2008). Evaluating the effect of scale in flood inundation modeling in urban environments. Hydrological Processes. 22: Hunter, N., Horritt, M., Bates, P., Wilson, M., Werner, M. (2005). An adaptive time step solution for raster-based storage cell modeling of floodplain inundation. Advances in Water Resources. 28(2005): Koltun, G.F., Kula, S.P., Puskas, B.M. A Streamflow Statistics (StreamStats) Web Application for Ohio. USGS Scientific Investigation Report Milly, P.C.D., Wetherald R.T, Dunne, K.A., and Delworth, T.L. (2002). Increasing risk of great floods in a changing climate. Nature. 415(2002): Musser, K Map of the Muskingum River Waterdshed. Acquired from Newton, B., office of the Muskingum County Engineer s Office, November 2009, Personal communication. Ohio Emergency Management Agency. Presidential Disaster Declarations in Ohio Acquired from Ohio Geographically References Information Program. Acquired from

52 Ohio Department of Natural Resources. Floods and Flood Damage Prevention. Division of Water Fact Sheet. Fact Sheet Reed, J., office of the Department of Natural Resources, November 2009, Personal communication. Trigg, M., Wilson, M., Bates, P., Horritt, M., Alsdorf, D., Forsberg, B., Vega, M. Amazon Flood Wave Hydraulics. Journal of Hydrology. 374(2009): United States Army Corps of Engineers. Muskingum River Basin Systems Operations Study. Huntington District. Great Lakes and Ohio River Division. (2006):3. United States Army Corps of Engineers. HEC-RAS River Analysis System. Hydraulic Reference Manual. Version