Dam Breach Model of Lake Anza Dam Using Hec-Ras A

DAM BREACH MODEL

OF LAKE ANZA DAM USING HEC-RAS

A Project

Presented to the faculty of the Department of Civil Engineering

California State University, Sacramento

Submitted in partial satisfaction of the requirements for the degree of

MASTER OF SCIENCE

Civil Engineering

(Water Resources Engineering)

Robert Joseph Greenberg

SPRING 2018

DAM BREACH MODEL

OF LAKE ANZA DAM USING HEC-RAS

A Project

Robert Joseph Greenberg

Approved by:

______, Committee Chair Cristina Maria Poindexter, P.E., Ph.D.

______, Second Reader Scott Meyer

______Date

Student: Robert Joseph Greenberg

I certify that this student has met the requirements for format contained in the University format manual, and that this project is suitable for shelving in the Library and credit is to be awarded for the project.

______, Graduate Coordinator ______Saad Merayyan, Ph.D. Date

Department of Civil Engineering

iii

Abstract

DAM BREACH MODEL

OF LAKE ANZA DAM USING HEC-RAS

Robert Joseph Greenberg

For this project, HEC-RAS was used to create an inundation map for the failure of the Charles

Lee Tilden Dam in Berkeley, California. A two-dimensional HEC-RAS model was created to simulate the Lake Anza Reservoir, C.L. Tilden Dam, and surrounding topography. HEC-RAS used the Shallow Water Equations to calculate hydraulic characteristics along the most likely path of dam breach flow. The inundation map, which shows the areas which would be adversely affected by the breach of the dam, was created using the maximum depth of water at each point along the dam breach flow’s path. This method can effectively predict the areas which will be inundated by water released by a dam failure and allows evacuation personnel to make informed decisions about which populated areas to evacuate in the case of such an emergency.

______, Committee Chair Cristina Maria Poindexter, P.E., Ph.D.

______Date iv

ACKNOWLEDGEMENTS

I would like to express my deepest appreciation to all those who supported me in the completion of this report. I would like to express my deepest gratitude to my grandmother, Lu Charlotte, without whom, I wouldn’t have been able to attend graduate school. Furthermore, I would like to thank my parents for supporting me throughout this endeavor. A special thanks to Lara Schenck for the helping me not only with the report, but also supporting me through the entirety of my time in school. For assisting me in generating a usable terrain file, I’d like to thank fellow graduate student, Robert Sherrick. I’d also like to thank Luis Mercado, who assisted me with running the experiment and writing the report. Last, but not least, I’d like to thank Cristina Maria Poindexter for her support as my advisor through this project.

TABLE OF CONTENTS Page

Acknowledgements ...... v

List of Figures ...... vii

Chapter

1. BACKGROUND ...... 1

2. METHODS ...... 5

3. RESULTS ...... 15

4. CONCLUSIONS ...... 24

APPENDIX A. Inundation Maps at Different Time Intervals ...... 25

References ...... 28

LIST OF FIGURES Figures Page

1. Oroville Dam Spillway Failure, February 2017 .... .……………………………….1

2. Diagram of Jurisdictional Dam Criteria ...... ……………………………. 2

3. Lake Anza ...... ………….………………………………….3

4. C.L. Tilden Dam...... …………………………. 4

5. HEC-RAS Geometry Editor ...... ………………………………. 7

6. Storage Area Editor ...... ……………………………. 10

7. Connection Editor ...... ………….…………………………………. 11

8. Dam Breach Parameters ...... …………………………. 13

9. Trees growing on the dam face ...... ………………………………. 14

10. Maximum Depth Map ...... ……………………………. 15

11. Depth Vs. Time at Curran Trailhead ...... …………………………. 16

12. Depth Vs. Time at Little Farm ...... …………………………. 17

13. Extent of Flooding 2 hours after dam breach ...... ………………………………. 18

14. Maximum Extent of Flooding at Interstate 80 Overcrossing .………………………. 19

15. Interstate 80 Freeway ...... …………………………. 20

16. Interstate 80 Freeway at San Pablo Exit ...... …………………………. 20

17. Max Depth along Interstate 80 Freeway ...... ………………………………. 21

18. DSS Wise Results ...... ……………………………. 22

vii

1. BACKGROUND

Figure 1: Oroville Dam Spillway Failure, February 2017

During the winter of 2017, there was record rainfall throughout the State of California leading to proportionately large runoff. California’s reservoirs, which had been practically empty in previous years due to prolong drought were rapidly filling and the California Department of

Water Resources (DWR) needed to release water as to not overtop the reservoirs. Lake Oroville, the second largest reservoir in California, was one such dam and its spillway was damaged, critically curtailing the amount of water that could be released through conventional means. This situation led to the eventual overtopping of the concrete weir at the emergency spillway and extensive scouring of the hillside below the spillway (see Figure 1). As this was unfolding, authorities in downstream communities evacuated over 180,000 people for fear that a catastrophic 2

damage failure would lead to massive loss of life (Independent Forensic Team, 2018). After the evacuation was lifted and several of the communities that had been evacuated filed lawsuits, there were doubts about the necessity of evacuating so large a population. It was revealed that so large an evacuation was made due to lack of information on the area that would be inundated by a catastrophic failure of the dam thus necessitating the authorities to take a conservative approach to the evacuation. As a result of this revelation, the California State Legislature amended the

California Water Code by adding section 6161 requiring all owners of jurisdictional dams to create inundation maps and make them available to the public.

Figure 2: Criteria for Jurisdictional Dams. (Department of Water Resources, State of California, n.d.)

Jurisdictional dams are defined as any dam that is higher than 6 feet impounding 50 or more acre feet of water or higher than 25 feet impounding more than 15 acre feet of water as shown in (see

Figure 2) (Department of Water Resources, State of California, n.d.). 3

Figure 3: Lake Anza

To explore the methods used by engineers to generate inundation maps, I chose to model the failure of the Charles Lee Tilden Dam which impounds Lake Anza (see Figure 3), a small lake in the San Francisco Bay Area’s Tilden Park. The lake was a favorite swimming hole for me as a child and its listing on the Department of Water Resources Jurisdictional Dams list with a hazard rating of “high” caught my interest (Department of Water Resources, State of California, 2017).

The dam (see Figure 4) is placed high in the Berkeley-Oakland Hills along Wildcat Creek which runs north from the saddle created by Vulmer and Grizzly Peak till it exits Wildcat Canyon just east of Interstate 80 before flowing into San Pablo Bay to the northwest. The majority of Wildcat 4

Creek is within Tilden Regional Park, part of the East Bay Regional Parks District and is accessible via fire trails and paved roads in certain locations.

Figure 4: C.L. Tilden Dam

2. METHODS

The data required to use the model at a rudimentary level is topography and information on the dam and its reservoir. The topography can be found in a number of ways; it can either be manually entered or generated using a terrain file from an outside source. Manually entering data involves drawing the river reaches then generating a series of cross sections to establish a hydraulic profile, but doing so is tedious and prone to error. An electronic terrain file, generated either through Light Detection and Ranging LIDAR, satellite mapping, or surveys is quicker and generally more accurate. For this experiment, an electronic terrain Digital Elevation Model

(DEM) file downloaded from the United States Geologic Survey (USGS) website was used. This terrain file did not include the topography below the surface of Lake Anza (the bathymetry). The

USGS DEM file also came with a projection file allowing for spatial projection in mapping software. Sacramento State graduate student Robert Sherrick, who is currently completing his culminating project on interpolating bathymetry, was able to incorporate Lake Anza into his project as a case study. In doing so, he generated a terrain file more pertinent to the purpose of this project than the DEM originally downloaded from the USGS website. In his edit of the

DEM, Robert Sherrick was able to modify the terrain around Lake Anza to reflect an interpolated bathymetry based on the surrounding topography. This had the desired effect of circumventing the limitations of the USGS DEM, which has a tendency to render aquatic topography invisible, but also effectively removing 80 years of sedimentation from the terrain. Additionally, because the USGS DEM file regarded the Tilden Dam embankment as part of the surrounding terrain,

Robert Sherrick altered the terrain such that the dam was removed and a dam could be placed there in HEC-RAS.

Once the modified DEM was acquired, the file needed to be modified slightly for HEC-RAS to convert it into a terrain file. In most geospatial models, there are areas for which there is no data 6

which can be presented in a number of ways. Often, these “No-Data Values” are denoted via an extreme outlier, a number so far outside of any expected value that the user would immediately recognize it as a placeholder for lack of data (e.g. -9999). Additionally, these extreme outliers are standardized as to be recognized by the computer as a no-data value. The DEM file used in this project denoted it’s no-data values with the text, “ndn” thus necessitating using a geographic information system program (GIS) to convert the no-data values into a number so HEC-RAS could read it. This was accomplished through a process referred to as “filling” which replaces text no-data values with numerical no-data values. After no-data values are converted to a form that is recognizable by HEC-RAS, the file can be imported into HEC-RAS through the RAS

Mapper, a mapping application that was introduced with the newest iteration of HEC-RAS, version 5.0.3. When importing the digital elevation model file to RAS Mapper, HEC-RAS must convert the file into its own native format, the terrain file. A projection file must be selected to set the datum of the map; in the case of this project, North American Datum 1983 (NAD83) was used. Additionally, it is important to be certain that the units displayed by the terrain file are the correct unit and datum. This initially proved to be a challenge for this project because the digital elevation model was in NAD83, but the units were in degrees rather than feet. It became necessary to manually (meaning rewriting the code by hand) change the units in the digital elevation model’s projection from “degrees” to “Foot_US”. Additionally, a multiplier that corresponded to the new measurement unit had to be changed; by default, the measurement unit is in meters, thus the multiplier is 1. To convert to US feet, a dimensional analysis operation to convert out of meters needed to be performed like so:

1 푚푒푡푒푟 = 0.30480060960121924 3.28084 푓푒푒푡 7

Once the terrain is in place in RAS Mapper, it can be used to view the topography of the area of interest in HEC-RAS; the color, contours, shading, and other display traits of the terrain can be changed. Once the terrain’s appearance has been appropriately modified, the RAS Mapper application can be put aside in favor of the Geometry Editor.

Figure 5: HEC-RAS Geometry Editor

The Geometry Editor, a longtime mainstay of the HEC-RAS program, as its name suggests, is used to generate and edit the geometry parameters for hydrologic models run by the program (see

Figure 5). In a one-dimensional HEC-RAS model, the horizontal dimensions of a river are represented as “reaches”, one dimensional lines showing where a river flows on the map. The third “Z” dimensions, representing depth, is established using “cross sections” where the dimensions of a given cross section along a reach are drawn to represent the topography surrounding the respective river reach. Flow calculations in HEC-RAS unsteady, one- dimensional models are performed using the Saint-Venant equations while two-dimensional models are calculated using the Shallow Water equations. 8

For two-dimensional flow, HEC-RAS performs calculations uses the Shallow Water Equations shown below:

퐶표푛푠푒푟푣푎푡𝑖표푛 표푓 푀푎푠푠:

휕퐴 휕푄 휕푄 + + = 0 휕푡 휕푥 휕푦

휕퐴 = 퐶ℎ푎푛푔푒 𝑖푛 푢푛𝑖푡 푓푙표푤 표푣푒푟 푡𝑖푚푒 휕푡 휕푄 푤ℎ푒푟푒: = 퐶ℎ푎푛푔푒 𝑖푛 푓푙표푤 푝푒푟 푢푛𝑖푡 푙푒푛푔푡ℎ 휕푥 휕푄 = 퐶ℎ푎푛푔푒 𝑖푛 푓푙표푤 푝푒푟 푢푛𝑖푡 푙푒푛푔푡ℎ {휕푥

퐶표푛푠푒푟푣푎푡𝑖표푛 표푓 푀표푚푒푛푡푢푚 (푥 − 푑𝑖푟푒푐푡𝑖표푛):

휕푢̅ 휕푢̅ 휕푢̅ 휕ℎ 휕2푢̅ 휕2푢̅ 휕2푢̅ 휕푧 + 푢̅ + 푣̅ = −푔 + 휐 ( + + ) + 푐 푢̅ − 푔 휕푡 휕푥 휕푦 휕푥 휕푥2 휕푦2 휕푧2 푓 휕푥

휕푢̅ = 푙표푐푎푙 푎푐푐푒푙푒푟푎푡𝑖표푛 휕푡 휕푢̅ 휕푢̅ 푢̅ + 푣̅ = 푐표푛푣푒푐푡𝑖푣푒 푎푐푐푒푙푒푟푎푡𝑖표푛 휕푥 휕푦 휕ℎ 푔 = 푓표푟푐푒 푝푒푟 푢푛𝑖푡 푚푎푠푠 푑푢푒 푡표 푝푟푒푠푠푢푟푒 푔푟푎푑𝑖푒푛푡 휕푥 푤ℎ푒푟푒: 휕2푢̅ 휕2푢̅ 휕2푢̅

휐 ( 2 + 2 + 2) = 푓표푟푐푒 푝푒푟 푢푛𝑖푡 푚푎푠푠 푑푢푒 푡표 푓푟𝑖푐푡𝑖표푛 휕푥 휕푦 휕푧 푛2푔|푢̅| 푛 = 푚푎푛푛𝑖푛푔푠 푐표푛푠푡푎푛푡 푐푓푢̅ = 푠푘𝑖푛 푓푟𝑖푐푡𝑖표푛 − 푤ℎ푒푟푒 푐푓 = 4 { ⁄3 푅ℎ = ℎ푦푑푟푎푢푙𝑖푐 푟푎푑𝑖푢푠 푅ℎ 휕푧 푔 = 푓표푟푐푒 푝푒푟 푢푛𝑖푡 푚푎푠푠 푑푢푒 푡표 푡ℎ푒 푏푒푑 푠푙표푝푒 { 휕푥

퐶표푛푠푒푟푣푎푡𝑖표푛 표푓 푀표푚푒푛푡푢푚 (푦 − 푑𝑖푟푒푐푡𝑖표푛):

휕푣̅ 휕푣̅ 휕푣̅ 휕ℎ 휕2푣̅ 휕2푣̅ 휕2푣̅ 휕푧 + 푢̅ + 푣̅ = −푔 + 휐 ( + + ) + 푐 푣̅ − 푔 휕푡 휕푥 휕푦 휕푦 휕푥2 휕푦2 휕푧2 푓 휕푦 9

Because of the large disparity between the length of a channel and its depth, Shallow Water equations assume a uniform depth along a channel and thus ignore the velocity in the vertical (Z) direction thus there is no term for z-velocities.

The difficulty in using a one-dimensional model for a dam breach simulation lays in the assumptions and labor of creating the geometry. Not only would it involve having to drawing cross sections large enough to account for all the possible areas that could be inundated by the breach, but also with the regularity along the reach as to yield worthwhile results. In a one- dimensional model, the dimensions of a cross section are interpolated between points along a reach where the cross section is known. This makes it necessary to draw many cross sections along a reach to maintain accuracy and avoid instability in flow calculations. Because of these reasons, it was decided to proceed using a two-dimensional flow model which uses the aforementioned terrain file rather than manually generated cross sections. To accomplish this, a

2D-Flow Area must be drawn on the map to establish which parts of the terrain will be included in flow calculations. At the very least, it’s necessary to delineate the area that will be inundated by the hypothetical dam breach. Because of the relatively small scale of this model and the terrain, it was possible to included practically the entire terrain in the 2D-Flow Area. This area extended from the Lake Anza dam all the way to the shore of San Pablo Bay where it was assumed any flows from the dam breach model would ultimately end up. In order to perform flow calculations across a relatively, large and dimensionally diverse flow area, a mesh of generally square and uniformly sized areas must be established in order to simplify the parameters involved. In this model, the mesh was specified as having 50 foot by 50 foot squares, but typically due to the diagonal boundaries of the outside of the model, some parts of the mesh at the margins of the 2D-Flow Area are non-square polygons. When drawing the 2D-Flow Area for this model, the upstream boundary was drawn so that it slightly overlapped the storage area, Lake 10

Anza. This action was taken to avoid having the simulation destabilized by aberrantly shaped mesh polygons which tend to be located near the margins of the 2D-Flow Area.

Figure 6: Storage Area Editor

After establishing the 2D-Flow Area, a storage area must be established to represent the reservoir subject to the dam breach in the model. This is accomplished by using the storage area tooltip to draw a polygon around the area of the reservoir on the terrain file. Once the geometry is established, the volume of the reservoir can be calculated by either using a simple calculation in which a surface area and depth are specified or through a more complex Volume-Elevation curve 11

that relates the volume to the surrounding terrain (see Figure 6). Because the elevation contours that corresponded to the top of the reservoir were not entirely contained within the area covered by the terrain file, the former of the two volumetric estimation methods was chosen, effectively rendering the topography of the reservoir irrelevant. For this model, a surface acreage of 4 acres was chosen with a depth of 60 feet. This allowed the volume to get as close to the accepted value of 268 acre feet as possible while not placing the elevation below that of the corresponding contour at the foot of the dam.

Figure 7: Connection Editor 12

To model the dam itself, a “Surface Area-2D Area Connection” needed to be added to the map in the Geometry Editor (see Figure 7). The connection can either be a weir, a culvert (gated or not gated), or linear routing, implying that the connection between the storage area and the 2D-Flow

Area doesn’t involve a structure. For this model, a broad crested weir with no culvert was selected because even though the Lake Anza Dam does technically have a spillway, HEC-RAS cannot breach structures with culverts.

The weir was specified to have a crest length of 400 feet, a crest elevation of 790 feet above mean sea level, the default weir coefficient Cd of 3, and a weir width of 150 feet. Additionally, the maximum head water elevation was set to 790 feet above mean sea level in the Parameters for

Hydraulic Property Tables. Connections can be set up to specify which flow areas are bounded by the dam; in this model, the upstream connection is the Lake Anza storage area and the downstream connection is to the Wildcat Canyon 2D-Flow Area. The dam breach parameters selected were based on the values given through the “Parameter Calculator” which allows the user to specify inputs based on the dam dimensions and then HEC-RAS calculates certain parameters based on five different dam breach equations. The available equations are:

1. MacDonald

2. Froehlich (1995)

3. Froehlich (2008)

4. Von Thun & Gillete

5. Xu & Zhang 13

Figure 8: Dam Breach Parameters

For this model, the Von Thun and Gillete equation was selected because the equations, which were based on the work of Froehlich and MacDonald, were positively reviewed in a literature review prepared in 1998 by the Dam Safety Office (Wahl, 1998). Once the dam breach parameters were in place, the model’s geometry was complete (see Figure 8). HEC-RAS can simulate both overtopping and piping failures, though due to the preponderance of trees growing on the dam face (see Figure 9) and the relatively small size of the aquifer upstream of the reservoir, piping failure was chosen for this model. 14

Figure 9: Trees growing on the dam face

Unsteady Flow parameters need to be set up for two-dimensional flow models to establish initial and boundary conditions. Boundary conditions can be added to the model such as groundwater inflow into the reservoir, lateral flow into the reservoir, or precipitation into both the reservoir and the 2D-Flow Area. The only initial condition is the starting elevation of 783 feet because this model is what is referred to as a “Sunny Day” failure (no precipitation) and groundwater or lateral inflows would likely have a negligible effect on the dam breach. Once the Unsteady Flow parameters are set up, the model can be run which involves choosing a simulation time window

(i.e. the amount of time that passes during the simulation) and the time intervals for the calculations and the outputs. When modeling a dam failure, it’s imperative to the stability of the model to set a proper computation interval. Since the events of a dam breach model happen relatively quickly compared to other events, a computation interval of seconds rather than minutes is necessary. A time interval that’s too long yields an unstable model for the same reason that a radar that tracks aircraft too infrequently can’t effectively track a plane’s course. Once the time window and computation/mapping intervals are set, the model can be run by pressing the compute button. 15

3. RESULTS

Figure 10: Maximum Depth

The results can be viewed in RAS-Mapper as either depth, velocity displays the depth of the flow at timesteps that correspond to the mapping intervals set prior to running the model. In addition to being able to see the depth or velocity of the water released by the dam breach over time, the user can elect to see the maximum extent of the flow (see Figure 10) at all points along its path over the simulation’s chosen duration. The failure wave’s progress at different time intervals is shown in Appendix A. Though the resultant map doesn’t show flow depths that correspond to any one time during the simulation, it does show the maximum extent of the flow within the 2D-Flow

Area and the maximum depth occurring at any given point over the simulation. The result is effectively the inundation map for the dam breach simulation. The depth or velocity versus time graph from points of interest can be chosen and displayed to show the velocity or depth as it changes over the course of the event. 16

Figure 11: Depth Vs. Time at Curran Trailhead

The depth graph at the Curran trailhead (Figure 11), only 800 feet downstream of the dam face shows the depth rapidly increasing over the course of roughly 30 minutes before rapidly decreasing to nearly zero. The zero-depth trend continues until just before the one hour mark before the depth increases to one foot and holds steady for the duration of the model. It’s unclear as to why the depth drops to zero, but it’s possible that it has to do with the end of the 0.62 hour duration of the dam failure, per the Von Thun and Gillette calculations, corresponding almost exactly to point when the zero depth period begins. The increase in flow that occurs later might be drainage from upstream areas that were flooded and are now draining. 17

Figure 12: Depth Vs. Time at Little Farm

Located about 1.15 miles downstream of the dam, the Little Farm and nearby Tilden Nature

Center would be hit by the failure wave roughly 30 minutes after the dam breach (see Figure 12).

Being the first area downstream of the dam with any appreciable number of people, timing would be critical to evacuate people before the water from the dam struck. In a 2007 study, (Jonkman,

Vrijling, & Vrouwenvelder, 2007), on the estimation of loss of life due to floods, the relationship between the flood’s arrival time, TA, and population at risk, NPAR, was estimated to be:

푁푢푚푏푒푟 표푓 푐푎푠푢푎푙푡𝑖푒푠 = 푁 = 0.5푁푃퐴푅 푤ℎ푒푛 푇퐴 < 0.25 ℎ표푢푟푠

0.6 푁푢푚푏푒푟 표푓 푐푎푠푢푎푙푡𝑖푒푠 = 푁 = 푁푃퐴푅 푤ℎ푒푛 0.25 < 푇퐴 < 1.5 ℎ표푢푟푠

푁푢푚푏푒푟 표푓 푐푎푠푢푎푙푡𝑖푒푠 = 푁 = 0.0002푁푃퐴푅 푤ℎ푒푛 푇퐴 > 1.5 ℎ표푢푟푠 18

Considering that the Little Farm and the Tilden Nature Center are two of the more popular attractions for school groups and families throughout the year, preventing casualties with between

25 and 30 minutes of warning time would be challenging. The number of casualties at Little

Farm, assuming a total number of visitors and staff combined of 100 people, is as follows:

퐹표푟 푎 푝표푝푢푙푎푡𝑖표푛 푎푡 푟𝑖푠푘, 푁푃퐴푅, 표푓 100 푝푒표푝푙푒:

0.6 0.6 푁푢푚푏푒푟 표푓 푐푎푠푢푎푙푡𝑖푒푠 = 푁 = 푁푃퐴푅 = (100 푝푒표푝푙푒) ≈ 16 푐푎푠푢푎푙푡𝑖푒푠

Hikers upstream of the Little Farm would suffer even greater casualties with estimates around

50% for anybody that would be reached by the failure wave in 15 minutes or less.

Figure 13: Extent of flooding after 2 hours; leading edge of wave is circled in red. 19

As shown in Figure 13, the failure wave would inundate areas where houses were built after roughly 2 hours, which according to the casualty estimation equations above, would allow ample time to evacuate the vast majority of the population in the inundation area.

Figure 14: Maximum extent of flooding at Interstate 80 Overcrossing

The next major challenge would mostly involve evacuating and closing down Interstate 80 which would be inundated by the failure wave after roughly 2 hour and 25 minutes (see Figure 14). For this model, it’s assumed that the debris carried by the failure wave would very quickly block the culvert impeding flow under the Interstate 80and lead to flooding of the freeway itself. This is partially due to the data used to create the digital elevation model including the I-80 freeway berm in the topography, but I thought it was a reasonable assumption to make considering the potential for debris to be carried by the dam breach flow. Additionally, the sound-wall that runs along the east side of I-80 was assumed to be breached in at least one place, allowing the dam breach flow to inundate the freeway. With the assumptions in place, the dam failure wave flows 20

onto the freeway, inundating the road surface with 6-12 inches of water moving less than 10 miles per hour.

The freeway is placed in a depressed trench with shallow sides forming a channel which the flow travels south towards Berkeley as shown in Figure 15.

Figure 15: Interstate 80 Freeway

Upon reaching the edge of the hills, roughly at the San Pablo Ave Exit, the freeway goes from being in a depressed trench to an elevated structure (see Figure 16).

Figure 16: Interstate 80 Freeway at the San Pablo Exit 21

It is at this point that the flood water spills off the edge of the freeway flooding adjacent neighborhoods (see Figure 17).

Figure 17: Max Depth along Interstate 80 Freeway

As a means of comparison, Professor Cristina Maria Poindexter ran a model of the dam failure using a different program called DSS Wise-Lite, a program specifically designed for creating inundation maps. In addition to using DSS Wise-Lite, a “bare earth” digital elevation model was used, meaning that any manmade structures are removed from the terrain. This is the largest fundamental difference between the two models and is self-evident in the results. 22

Figure 18: DSS Wise Results

Due to the removal of the Interstate 80 overpass and other manmade objects that Wildcat

Creek passes under, the flows from the dam failure pass relatively unimpeded toward San Pablo

Bay, albeit still inundating areas directly around the creek (see Figure 18). The effect of this difference is that the assumption used in the other model, that the I-80 culvert becomes clogged with debris is null when using a bare earth DEM. It should be noted in the DSS-WISE Lite model that Wildcat Creek overflows its bank and flows into the nearby San Pablo Creek, which is less than 400 feet away in certain places. Additionally, DSS-WISE Lite used a mesh made up of 23

30 foot by 30 foot square cells and the formation time of the dam breach was assumed to be instantaneous in contrast to HEC-RAS assuming a 0.62 hours dam breach formation time.

4. CONCLUSIONS

The disparity between the results of the two models, HEC-RAS and DSS-WISE Lite, illustrates the challenges in creating an accurate inundation map for a given failure scenario. Are the assumptions more valid for one model than the other? What if the Interstate 80 bridge were to fail and collapse during the dam failure either due to an earthquake that caused the failure, the impact of the dam breach flow, or some combination of both? It’s for these reasons that planners tend to be conservative in their evacuation plans. While DSS Wise is more streamlined and robust, HEC-RAS is more versatile and diverse in terms of its applications. These two models showed what both HEC-RAS and DSS-Wise are capable of, but to effectively generate an inundation map that would account for enough eventualities as to minimize loss of life, more scenarios need would need to be considered and simulated. Lateral inflows from tributaries to the storage area and the 2D-Flow area, precipitation, and land cover’s effect on Manning’s n were neglected in this model, but could be included in order to improve the model. Though this model was relatively simple, HEC-RAS is a powerful tool capable of simulating a range of scenarios.

The possibilities for different eventualities during a dam breach warrants further study of a Lake

Anza dam breach in the future.

APPENDIX A: INUNDATION MAPS AT DIFFERENT TIME INTERVALS

Figure A1: Flood Extent after 15 minutes

Figure A2: Flood Extent after 30 minutes 26

Figure A3: Flood Extent after 60 minutes

Figure A4: Flood Extent after 90 minutes 27

Figure A5: Flood Extent after 120 minutes

References 1. Department of Water Resources, State of California. (2017, September). Dams Within Jurisdiction of the State of California. California, United States of America. 2. Department of Water Resources, State of California. (n.d.). Jurisdictional Sized Dams. Retrieved from California Department of Water Resources: https://www.water.ca.gov/Programs/All-Programs/Division-of-Safety-of- Dams/Jurisdictional-Sized-Dams 3. Independent Forensic Team. (2018, January 5). Independent Forensic Team Report Oroville Dam Spillway Incident. California, United States of America. 4. Jonkman, S. N., Vrijling, J. K., & Vrouwenvelder, A. (2007, August 12). Methods for the estimation of loss of life due to floods: a literature review and a proposal for a new method. The Netherlands. 5. Wahl, T. L. (1998, July). Prediction of Embankment Dam Breach Parameters, A Literature Review and Needs Assessment. United States of America.