Академический Документы
Профессиональный Документы
Культура Документы
A Project
MASTER OF SCIENCE
in
Civil Engineering
by
FALL
2017
© 2017
ii
A COMPARISON OF 1D AND 2D HEC-RAS MODELS OF THE NAPA RIVER
A Project
by
Approved by:
____________________________
Date
iii
Student: Allison Jane Bratton
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
iv
Abstract
of
by
Floods are the leading cause of loss of property and loss of life due to natural disasters in
the United States. Flood inundation mapping is a critical tool for developing emergency
action plans that can help lessen flood related losses. Flood inundation maps are
developed using hydraulic models to provide information necessary for predicting the
impacts of a given flood event based on inundation extents and maximum water surface
elevations at critical locations. With this information, local agencies can develop
emergency action plans to make informed decisions based on forecasts and real time
stream gage data. Recent advances in hydraulic modeling provide the ability to develop
flood inundation maps using one-dimensional (1D) and two-dimensional (2D) models.
For this study 1D and 2D hydraulic models were developed using the United States Army
(HEC-RAS) version 5.0.3 to compare and contrast the two types of models and to
determine which type of model was better suited for flood inundation mapping. The
models were developed for a 4-mile reach of the Napa River, through the City of St.
v
Helena, California. The following data sources were used to develop the 1D and 2D
model files: terrain data was created using LiDAR data of the Napa River collected in
2014, land cover values and model boundary conditions were determined from the Napa
River Federal Insurance Study published by the Federal Emergency Management Agency
(FEMA) and the flow data was developed from the United States Geological Survey’s
(USGS) rating curve for gage 11456000 ‘Napa R Nr St Helena CA’. The 1D and 2D
models were each calibrated to gage heights within a 5% difference from the measured
data to ensure acceptable accuracy of the flood inundation extents. The 1D and 2D
models were also calibrated to within a 0.33% difference from each other to provide an
accurate comparison.
The results of the 1D and 2D models were compared based on the following: calibration,
run time, velocity, water surface elevation and inundation extents. Calibration results
showed that the 1D model required Manning’s n-values 25-34% higher than the 2D
model to achieve the same calibration results. This is because the hydraulic roughness
factor in the 1D model is used to account for most hydraulic losses, while the 2D model
accounts for some of those losses within the computation. The run time of the 1D model
was 96% faster than the 2D model due to the significantly larger number of computation
nodes within the 2D model grid. The 1D model calculated lower velocities in the
overbank and higher velocities in the channel when compared to the 2D model at
maximum velocity. The most compelling reason for this difference is because the
velocities are averaged over the entire overbank section of each 1D cross section while
vi
the velocities are calculated for each grid cell of the 2D model. The grid cells of the 2D
model are significantly smaller than the overbank area of the 1D cross section, allowing
for more accurate and detailed calculations in the 2D model. The water surface elevations
of the 1D model were lower than those of the 2D model by more than 2 feet in some
areas. The differences were mostly attributed to the calculation of expansion and
contraction losses within the models. The methods used in the 1D model typically lead to
elevations. The final point of comparison was the inundation extents which are
determined based on the water surface elevations of the model. The differences varied,
but overall the inundation extents of the 1D model was less than the inundation extent the
1D model.
Based on all model comparisons it was evident that the 2D model was better suited for
flood inundation mapping as it pertained to this study. The results provided by the 2D
model were more detailed and therefore more accurate in the overbank areas of the
model, which is an indication of floodplain extent potential. The 2D results for water
surface elevation and inundation extents were also more conservative than the 1D model.
The differences are largely attributed to the greater number of computation points in the
2D model which allow for more detailed results as well as the computation methods that
more accurately represent hydraulic losses within the model. The 2D model also accounts
vii
for lateral flow as well as longitudinal flow allowing for a more accurate representation
_______________________
Date
viii
ACKNOWLEDGEMENTS
- Ryan Greif for his mentorship and support through the development of this
project as well as Mead & Hunt for providing me the opportunity to learn and
- Napa County for allowing me to use the 2D Napa River model, which I developed
for them under the guidance of Ryan Grief. Their support made this study
possible.
- My advisor and committee chair, Dr. Merayyan for his feedback and guidance
- Finally, my friends and family, especially my boyfriend, whose love, support and
ix
TABLE OF CONTENTS
Page
Acknowledgements ...................................................................................................... ix
Chapter
1. INTRODUCTION ...................................................................................................1
2. BACKGROUND .....................................................................................................3
2D Computation Methods.........................................................................13
Calibration........................................................................................................26
Calibration........................................................................................................29
Velocity ............................................................................................................35
References ....................................................................................................................69
xi
LIST OF TABLES
Tables Page
1. Project study area and modeled Napa River Reach Through St. Helena................ 4
1D models ............................................................................................................. 28
5. Rating curve comparison of critical calibration points for measured and modeled
8. Right overbank subsections for 1D cross section transposed with the 2D grid .... 38
10. Maximum water surface elevation comparison at cross section 11526 ................ 43
xii
LIST OF FIGURES
Figures Page
1. Project study area and modeled Napa River Reach Through St. Helena................ 4
1D models ............................................................................................................. 28
5. Rating curve comparison of critical calibration points for measured and modeled
8. Right overbank subsections for 1D cross section transposed with the 2D grid .... 38
10. Maximum water surface elevation comparison at cross section 11526 ................ 43
xiii
1
CHAPTER 1
INTRODUCTION
Flood inundation maps are developed by hydraulic models and used for predicting
important information related to flood events such as inundation extents and water
property and reducing loss of life potential. Hydraulic models are a representation of the
hydraulic processes that take place during a flood event (Cook, 2008). The quality of a
flood inundation map is dependent on the model’s ability to accurately represent the
hydraulic processes over the river’s floodplain. Over the last thirty years, developments
in hydraulic modeling have allowed for more accurate predictions of the potential
There are various types of software available for development of 1D and 2D models, but
computations for a network of natural and constructed channels, floodplain areas and
levee protected areas (USACE Hydrologic Engineering Center, 2016). There are
models when compared to 2D models are that they require less data, they have shorter
computation times, they compute channel flow more efficiently and the file sizes are
more manageable. Whereas 2D model advantages include faster model set-up, better
2
computational accounting of hydraulic losses and output that is easier to interpret and
project is dependent on multiple factors, such as study area, available data, project
schedule and budget, regulatory requirements and the modelers experience level.
This study will construct 1D and 2D hydraulic models of the Napa River through the City
of St. Helena, California using HEC-RAS version 5.0.3. The models will be calibrated
and compared based on the following results: calibration, run time, velocity, water
surface elevation and inundation extents. The most significant points of comparison will
be used to determine whether the 1D or 2D model is better suited for flood inundation
mapping.
Although this report discusses the applications of flood inundation mapping using the
City of St. Helena as a case study to demonstrate the suitability of 1D and 2D models in
floodplain mapping. The main focus of this project is to compare the models and make
studies, not provide an assessment of flood inundation risk for the City of St. Helena.
3
CHAPTER 2
BACKGROUND
Study Area
The drainage basin of the Napa River is approximately 50 miles long (north to south) and
10 miles wide covering a total area of 426 square miles. Numerous tributaries feed into
the Napa River from the adjacent foothills as it winds its way through Napa County and
ends in tidal sloughs approximately 9 miles north of the San Pablo Bay. For the purpose
of this study a 4-mile reach of the Napa River is modeled through the City of St. Helena
(St. Helena) extending from Pratt Avenue upstream to Zinfandel Avenue downstream
The Napa River flows along the eastern edge of St. Helena in the northern region of the
Napa Valley. Due to the river’s proximity, St. Helena has a required flood response when
the river reaches a specified stage. The stage information is provided by a flow gage
(USGS 11456000 Napa R NR ST HELENA CA) at the Pope Street Bridge; operated by
the United States Geological Survey in collaboration with The National Weather Service
(NWS). The gage station provides year-round forecasting and flood stage monitoring.
Figure 1 shows the project study areas and modeled river reach through St. Helena.
The Napa River through St. Helena has a maximum recorded flow of 18,300 cubic feet
per second (cfs) based on the available period of record of 75 years (USGS, 2017). The
average bed slope through the study area is 0.26%. The reach is bounded by Howell
4
Mountain to the northeast, which rises quickly from the valley floor, and a relatively
narrow floodplain to the southwest that expands downstream through the Napa Valley.
The region is at risk of slow rise flooding during peak rainfall events producing heavy
Figure 1 Project study area and modeled Napa River Reach Through St. Helena
5
events, specifically the extents of inundation and the maximum water surface elevations
at specific locations, such as roads and emergency response facilities. The information
gathered from this process is extremely important for development of emergency action
plans. Floods are the leading cause of loss due to natural disaster in the United States. Of
all Federally declared natural disaster, 75% of them are related to floods. Average flood
losses in the United States include nearly $8 billion and over 90 fatalities per year. Over
time the average annual number of fatalities has declined which is in large part due to
mapping. On the other hand, economic losses continue to rise as increased urbanization
causes development to expand into floodplains, further supporting the need for flood
multiple federal agencies collaborate by sharing data to provide current inundation maps
in an effort to assist local agencies in assessing flood risk. The major federal agencies
include: The National Weather Service (NWS), the United States Geological Survey
(USGS), The United States Bureau of Reclamation (USBR), The United States Army
NWS provides the Inundation Mapping Interface, a web based tool that combines all
collaborative inundation data, indicating when roadways, buildings, airports and other
6
critical facilities are likely to see the impacts of floodwaters during different storm
scenarios. With this information, local agencies can develop emergency action plans and
make informed operational decisions during a flood event based on NWS flood forecasts
As one of the collaborative partners to NWS, USGS works with communities to identify
appropriate stream sections where flooding is likely to occur and gather necessary data to
build models and produce inundation maps ensuring scientific integrity. Scientific
integrity is achieved by carefully calibrating the hydraulic model used to produce the
coefficients to match the modeled water surface elevations to within one half foot of
previously recorder flood events (Colorado Water Conservation Board, 2006). Reliable
flood data from an established stream gage is crucial to producing a calibrated model that
surface elevation data the hydraulic model produces a gridded surface defining the
Introduction to HEC-RAS
There are multiple software options when it comes to hydraulic model development, but
HEC-RAS is preferred among most engineers. The United States Army Corps of
7
Engineers released HEC-RAS version 1.0 in July of 1995. Since the original release there
have been several major updates leading up to the current version, 5.0.3. Software
version 5.0 introduced the capabilities for 2D hydraulic modeling. The following sections
1D vs 2D HEC-RAS Models
1D and 2D unsteady HEC-RAS models are both solved using the fully dynamic St.
Venant equations of conservation of mass and momentum. For 1D models, the equation
is solved along a singular dimension, which is defined by the modeler. This means the
modeler must determine the flow direction during model set-up which can be a time
consuming iterative process. 2D models simply require the definition of a mesh over the
model area and the hydraulic computations are completed in two dimensions. The flow
direction is determined by the model (West Consultants Inc., 2017). The assumptions for
of: surveyed station-elevation data, reach lengths, Manning’s n-values, main channel
only requires mesh definition. However, the 2D model requires a digital elevation model
(DEM) for the defined mesh area. The model uses the DEM to determine where the water
will go, so its quality/resolution are critical for model accuracy (West Consultants Inc.,
2017). There are various sources for DEM data and it is becoming easier to find, but in
9
many areas DEMs aren’t available and data the collection required to produce a high-
The applications of 1D verses 2D models vary. 1D models are typically best for
channelized flow like canals or mountain streams while 2D models are much better for
areas with dynamic flow patterns, such as alluvial fans or estuaries (Colorado Water
Conservation Board, 2006). Ultimately the type of model used will be dependent on the
objectives of the project and the amount of detail required (West Consultants Inc., 2017).
net rate of flow into the control volume must be equal to the rate of change within the
control volume. The continuity equation is shown in Equation 2.7 with the simplified
final form of the equation presented in Equation 2.8 USACE Hydrologic Engineering
Center, 2016).
10
𝜕𝐴 𝜕𝑄 ∆𝑥 𝜕𝑄 ∆𝑥
𝜌 𝜕𝑡 ∆𝑥 = 𝜌[(𝑄 − 𝜕𝑥 ) − (𝑄 + 𝜕𝑥 ) + 𝑄𝑙 ] Equation 2.7
2 2
Where:
𝜕𝐴
∆𝑥 = Rate of change in storage
𝜕𝑡
𝜕𝑄 ∆𝑥
(𝑄 − 𝜕𝑥 ) = Rate of inflow
2
𝜕𝑄 ∆𝑥
(𝑄 + 𝜕𝑥 ) = Rate of outflow
2
𝜌 = Fluid density
𝜕𝐴 𝜕𝑄
= ( 𝜕𝑥 ) − 𝑞𝑙 =0 Equation 2.8
𝜕𝑡
stating that the momentum flux (net rate of momentum flow into the control volume)
plus the sum of all external forces acting on the volume must be equal to the rate at
which the momentum accumulates. The equation is a vector equation. For 1D flow, it is
applied in the direction of flow (x-direction). The external forces considered are gravity,
pressure and boundary drag (friction) (USACE Hydrologic Engineering Center, 2016).
-
→
∑ 𝐹𝑥 = 𝑑𝑀 Equation 2.9
𝑑𝑡
The external force on the fluid in the control volume due to gravity is defined in Equation
2.10.
11
𝜕𝑧𝑜
𝐹𝑔 = −𝜌𝑔𝐴 ∆𝑥 Equation 2.10
𝜕𝑥
Where:
- The total external pressure force is the integral of the product of pressure and area over
the cross section, which is assumed to be hydrostatic (varying linearly with depth)
USACE Hydrologic Engineering Center, 2016). The net pressure force equation is
𝜕ℎ
𝐹𝑃𝑛 = −𝜌𝑔𝐴 𝜕𝑥 ∆𝑥 Equation 2.11
Where:
h = depth
The external drag force is due to the frictional forces between the channel and the fluid,
Where:
𝑆𝑓 = Friction slope
𝑄|𝑄|𝑛2
𝑆𝑓 = Equation 2.13
2.208𝑅 4/3 𝐴2
12
- Equation 2.13 defines the friction slop using the Manning equation instead of the Chezy
equation. It is the form used most predominantly in the United States and therefore used
The momentum flux entering the control volume is defined in Equation 2.14 and the
𝜕𝑄𝑉
−𝜌 ∆𝑥 Equation 2.14
𝜕𝑥
𝜕 𝜕𝑄
(𝜌𝑄∆𝑥) = 𝜌∆𝑥 Equation 2.15
𝜕𝑡 𝜕𝑡
The sum of the external forces (Equations 2.10-2.12) plus the momentum flux into the
(Equation 2.15). All equations combine to form the momentum equation (Equation 2.16).
𝜕𝑄 𝜕𝑄𝑉 𝜕ℎ 𝜕𝑧𝑜
𝜌∆𝑥 𝜕𝑡 = −𝜌 ∆𝑥 − 𝜌𝑔𝐴𝑆𝑓 ∆𝑥 − 𝜌𝑔𝐴 𝜕𝑥 ∆𝑥 − 𝜌𝑔𝐴 ∆𝑥 Equation 2.16
𝜕𝑥 𝜕𝑥
Equation 2.16 can be further simplified based on Equation 2.17 since the elevation of the
water surface z is equal to zo +h. The final form of the momentum equation is presented
in Equation 2.18.
𝜕𝑧 𝜕ℎ 𝜕𝑧𝑜
= 𝜕𝑥 + Equation 2.17
𝜕𝑥 𝜕𝑥
13
𝜕𝑄 𝜕𝑄𝑉 𝜕𝑧
+ + 𝑔𝐴 (𝜕𝑥 + 𝑆𝑓 ) = 0 Equation 2.18
𝜕𝑡 𝜕𝑥
2D Computation Methods
three directions. When considering channels and floodplains in 2D models, the equations
are further simplified. As with 1D models, 2D models utilize the principles of mass and
momentum conservation. All equations related to 2D analysis assume the bottom surface
elevation is z(x,y), the water depth is h(x,y,t), (USACE Hydrologic Engineering Center,
The unsteady differential form of the continuity equation is presented in Equation 2.20,
𝜕𝐻 𝜕(ℎ𝑢) 𝜕(ℎ𝑣)
+ + +𝑞 =0 Equation 2.20
𝜕𝑡 𝜕𝑥 𝜕𝑦
Where:
The integral form of the equation is defined in Equation 2.21. This form of the equation is
𝜕
𝜕𝑡
∭𝛺 𝛺 + ∬𝑆 𝑉 ∙ 𝑛𝑑𝑆 + 𝑄 = 0 Equation 2.21
Where:
S = side boundaries
bathymetry detail into the relatively coarse grid required to calculate the numeric model.
The grid cells contain hydraulic radius, volume and cross-sectional area information that
is pre-computed using the fine bathymetry thus preserving some of the information from
the high-resolution grid which can be accounted for in the numerical method of the
coarser grid through mass conservation. Since the free water surface is smoother than the
detailed bathymetry the coarse grid can effectively compute the variability in the free
2.22
Where:
k = cell face
The horizontal length scales in 2D models are much larger than the vertical length scale,
volume conservation implies that the vertical velocity is small. In this case the Navier
Stokes vertical momentum equation justifies pressure as nearly hydrostatic. Without non-
hydrostatic pressure, variable density and wind a vertically averaged form of the
represented in Equations 2.23 and 2.24. The left-hand sides of the equations contain the
acceleration terms and the right-hand sides the internal and external forces acting on the
fluid.
𝜕𝑢 𝜕𝑢 𝜕𝑢 𝜕𝐻 𝜕2 𝑢 𝜕2 𝑢
+ 𝑢 𝜕𝑥 + 𝑣 𝜕𝑦 = −𝑔 𝜕𝑥 + 𝑣𝑡 (𝜕𝑥 2 + 𝜕𝑦 2 ) − 𝑐𝑓 𝑢 + 𝑓𝑣 Equation 2.23
𝜕𝑡
𝜕𝑣 𝜕𝑣 𝜕𝑣 𝜕𝐻 𝜕2 𝑣 𝜕2 𝑣
+ 𝑢 𝜕𝑥 + 𝑣 𝜕𝑦 = −𝑔 𝜕𝑦 + 𝑣𝑡 (𝜕𝑥 2 + 𝜕𝑦 2 ) − 𝑐𝑓 𝑣 + 𝑓𝑢 Equation 2.24
𝜕𝑡
Where:
R = Hydraulic Radius
f = Coriolis parameter
17
CHAPTER 3
MODELING METHODS
This section discusses the modeling methods used to develop the 1D and 2D models. The
sub-sections are presented in the order of the workflow used to develop the models.
Model development began with data gathering (elevation data), land cover delineation,
flow data set-up, 2D model set-up followed by 1D model set-up, plan file creation and
Elevation Data
The 1D and 2D model geometries are based on a Digital Elevation Model (DEM) of the
Napa River created using LiDAR data collected on June 19-20, 2014 (Napa County,
2014). The DEM was clipped to the model extents in order to decrease file size and
model computation time. The DEM grid cell size is 3 feet by 3 feet.
The major land use categories in St. Helena are: residential, commercial and mixed use,
business and industrial and community and natural resource. Of the total land in St.
Helena, 74.3% falls into the community and natural resources category, 48.3% of which
is agricultural land (City of St. Helena, 2015). The land use categories for the model
geometry were simplified into three categories: trees and dense brush, agriculture and
values used for land use classification are shown in Table 2. The values were determined
18
from the n-values used in the FEMA Flood Insurance Study (FEMA, 2016) for Napa
County dated August 3, 2016. Once the categories and n-values were established, the
model area was delineated using ArcGIS. The final n-values for each model differ from
the base values and are further discussed in the calibration section of the results.
Flow Data
The model flow file was developed using data from the USGS rating curve for gage
11456000 ‘Napa R Nr St Helena CA. Seven steady state flow rates were determined from
the rating curve based corresponding gage heights from 14-26 feet at 2 foot intervals. The
rating curve ends at 24 feet, so the flow rate for gage height 26 feet was extrapolated
from the curve, shown in Figure 2. The curve was extrapolated to include this maximum
flow rate in order to include an extreme case of flood inundation. Flow rates were
determined in this manner to simplify the calibration process for each model which will
be further discussed in the following calibration section. The HEC-RAS unsteady flow
30
25
20
Gage Height (feet)
15
10
0
0 5000 10000 15000 20000 25000
Flow (cfs)
Each flow rate was maintained for 8 hours of simulation time in order to allow the
models to equilibrate before ramping up the inflow to the next flow rate over another 8-
hour period. In order to simplify the model computations and eliminate the need for each
flow rate to ramp up from zero all flows were simulated in succession using a single
The boundary conditions for the models were developed based on the FIS report. Two
upstream boundary conditions were used, one on the Napa River at Pratt Avenue and one
on Sulphur Creek. The assumed energy grade slopes were determined by the average
20
slope of the 100-year water surface profiles upstream of the boundary condition location,
based on the FIS. The boundary conditions were used by the model to distribute the
inflow across the flow boundary line. The portion of total inflow assigned to the Sulphur
Creek inflow boundary for each simulation was assigned based on the ratio of 100-year
peak flow rates between Sulphur Creek at the confluence with the Napa River and the
Napa River at Zinfandel Lane as listed in the FIS. The downstream model boundary at
Zinfandel is a normal depth-type condition. The slope was calculated from the average
slope of the 100-year water surface profile extending downstream from Zinfandel
Avenue as depicted on the flood profiles in the FIS (Greif, 2016). All FIS report data
2D Model Set-up
The 2D model was created before the 1D model and is therefore discussed first. Many of
the model parameters of the 2D model were used to create the 1D model.
The 2D model geometry was created in the HEC-RAS geometry editor. First the DEM
was imported into RAS mapper to create the model terrain. Next, the perimeter of the 2D
grid area was drawn using the 500-year floodplain of the FIS as a guide, extending
laterally far enough to encompass the inundation areas along the Napa River, with the
included in the model was limited to 1,380 feet and only included to represent the
21
inundation area around its confluence with the Napa River. The grid area along the right
bank of the Napa River upstream of the Sulphur Creek confluence was drawn along the
Vineyard Valley Mobile Home Park Floodwall which was designed to withstand
maximum flood conditions. The finished grid was then associated with the previously
Once the 2D perimeter was established, the boundary conditions were drawn in at the
previously discussed locations and the computational mesh was created with an initial
grid spacing of 50 feet by 50 feet. The mesh was further refined using breaklines, which
force the computational cell faces of the mesh to align along high points in the terrain that
Engineering Center, 2016). Figure 3 shows the 2D geometry set-up: the perimeter is
represented by the bold black line and the breaklines are represented by the pink lines.
Mesh Max Cell Area Min Cell Area Average Cell Area
(# of cells) (sq ft) (sq ft) (sq ft)
24,725 21,775.4 979.7 2,522.09
1D Model Set-up
The geometry set-up for a 1D model is more involved than a 2D model as it requires
thoughtful planning of cross sections and flow paths while the 2D model simply requires
a grid boundary around the potential inundation extent. 1D model set-up was completed
procedures, tools and utilities for processing geospatial data (USACE Hydrologic
Engineering Center, 2009). Geometric data was prepared in ArcGIS and imported into
HEC-RAS. The 2D model geometry and resulting flow paths were used to guide the 1D
model set-up in order to create an equivalent model that could be used for comparison.
The previously created DEM was imported into GeoRAS and used as a base to create the
series of layers required for developing the HEC-RAS import file. The pertinent line
layers included the Stream Centerline (reach), Main Channel Banks (bank points), Cross
Section Cut Lines (cross sections), Flow Path Centerlines (flow lines), Land Cover and
Ineffective Areas.
The Stream Centerline layer was created first as is recommended since it establishes the
river reaches that tether the remaining geometry layers (USACE Hydrologic Engineering
Center, 2009). Three reaches were drawn from upstream to downstream, listed in order:
Napa River-St. Helena, Sulphur Creek and Napa River-St. Helena_ds. The confluence of
Sulphur Creek creates a connection that requires the main reach of the Napa River be
split into two sections, upstream and downstream of the confluence. Next the Main
Channel Banks layer was created to establish the left and right bank stations along each
reach.
24
Using the established reaches and bank stations as a guide, the cross sections were drawn
from the left overbank to the right overbank, looking downstream. The cross sections
were drawn at an average interval of 200 feet to represent the flow carrying capacity of
the channel and adjacent floodplain. The left and right extents of the cross sections were
drawn perpendicular to the anticipated flow lines extended across the floodplain to match
the extents of the previously developed 2D grid perimeter. Adjustments were made to all
Flow Path Centerlines were drawn through the centroid of flow of the left overbank,
channel and right overbank of each reach which when imported into HEC-RAS provided
the downstream reach lengths between cross sections. This layer is optional, but if
omitted the reach lengths for the entire cross section are considered uniform and
Relying only on the Stream Centerline layer decreases the accuracy of overbank flow
Additional optional layers included in the GeoRAS set-up were the Land Cover layer and
the Ineffective Areas layer. Creation of the land cover layer was previously described in
section 3.2 and used in the GeoRAS set-up to apply the base Manning’s n values to each
cross section. The ineffective areas were drawn to define areas of inundation that are not
actively conveying water (USACE Hydrologic Engineering Center, 2016). The 2D model
results were used as a guide to determine the ineffective areas throughout the model. In
25
GeoRAS the ineffective areas are only defined in the horizontal plane, the elevation
assigned is not based on the terrain. Since ineffective areas typically change as the water
surface elevation changes, increased water surface elevations result in less ineffective
area, the ineffective area elevations were manually adjusted once they were imported into
HEC-RAS.
The final step in the model set-up required the creation of plan files, which determine the
overall calculation settings for each model run. The plan file is used to set the geometry
and flow files, the simulation time window and the computation settings. All of which
were discussed in previous sections except for the computation settings, which determine
the computational interval, the mapping output interval, the hydrograph output interval
and the detailed output interval. The computation interval is a critical component of
HEC-RAS model set-up as an incorrect time step can lead to model instability issues.
This is because the unsteady flow equations are derivatives calculated with respect to
distance and time. If hydraulic properties are changing quickly and the computation
interval is too large the model becomes unstable. In contrast if the computational interval
is too small the leading edge of the flood wave will steepen, which could cause
For the purpose of these models the computational interval was determined using the
courant number, the equation 3.1. The courant number is a function of the velocity and
the cell size for 2D models and cross-section distance for 1D models.
𝑉𝑆 ∆𝑇
𝐶= Equation 3.1
∆𝑋
For the 2D model, two plans were created. One was for model calibration and the other
was for the final model results, both used the full momentum equation. It is
recommended for the Courant number to be less than or equal to 1 with a maximum of 3
for the full momentum equation (West Consultants Inc., 2017). All 2D model plans were
Two plan files were created for the 1D model, one for the calibration and one for the final
model. Both models also utilized a computational interval of 30 seconds since the cross-
section spacing was comparable to the mesh size and there were no instability issues
Calibration
values with acceptable accuracy (West Consultants Inc., 2017). For the purpose of this
project, the goal of calibration was two-fold. First was to produce results that could be
used to accurately assess flood inundation risk by modeling water surface elevations that
27
matched measured water surface elevations. Second was to create modeled results that
could be accurately compared to each other by calibrating the 1D model to the previously
The USGS gage at Pope Street was used as the point of calibration because it provided
the necessary measured values. As previously discussed in the flow section, the flow data
was set-up to support efficient calibration. Since each of the seven flow rates corresponds
to an even interval of the USGS rating curve the models were adjusted to the measured
stage at each flow rate. Calibration was achieved by changing the models’ Manning’s n-
values until the water surface elevations at the Pope Street gage location closely matched
the gage height of the rating curve. Based on the Manning’s equation (Equation 3.2)
1.468 2⁄
𝐾= 𝐴𝑅 3 Equation 3.2
𝑛
modeled n-values are inversely related to conveyance and directly proportional to the
resulted in an increase in the calculated water surface elevations and vice versa. Figure 4
shows the modeled water surface elevations of the calibrated vs. uncalibrated 1D model
at Pope Street. The figure supports the aforementioned correlation because the calibrated
n-values for the 1D model were increased from the base values and the resulting water
28
surface also increased. Final n-values for both models differ from the base values and are
220
215
210
205
200
195
190
0 500 1000 1500 2000 2500
Station (feet)
Figure 4 Calculated water surface elevation comparison of calibrated and un-calibrated 1D models
29
CHAPTER 4
This chapter will compare the results of the HEC-RAS 5.0.3 1D and 2D models of the
Napa River through the City of St. Helena. The model results are compared based on
calibration, run time, velocity, water surface elevation profiles and inundation extents.
Within each category, the significant model differences are determined, the causes of
those differences are explained and recommendations are made as to which model is
better suited for flood inundation mapping. For the purpose of this report only the
significant points of comparison are discussed, the additional results for water surface
Calibration
Upon final calibration, the Manning’s n-values for the 1D model were 25-34% greater
than the n-values of the 2D model, depending on the land use type. Table 4 shows the
resulting n-values for each model. The final column of the table shows the calculated
Table 5 presents the results of the calibration based on comparison of the measured and
modeled gage heights at the Pope Street gage location. All resulting gage heights for the
1D and 2D models are less than 5.5% different from the measured values. This is
mapping tools.
The final column of Table 5 shows the calculated percent difference between the 1D and
2D modeled gage heights. The differences range from 0.33-0.01%. The highlighted
values in Table 5, gage heights 20-24, are the points of critical calibration because the
flooding extents at these gage heights become much more significant and therefore model
accuracy is most important. Figure 5 shows these points graphically compared with the
USGS rating curve. The difference between these points ranges from 0.01-0.03% which
is considered negligible. This means the 1D and 2D models are well suited for an
accurate comparison. This also means the previously mentioned percent increase of
31
Table 5 Comparison of measured and modeled gage heights at Pope Street gage
2D 1D
% % %
HEC-RAS HEC-RAS
USGS Gage Difference Difference Difference
Gage Gage
Height (ft) of USGS of USGS of
Height Height
and 2D and 1D 1D and 2D
(ft) (ft)
14 14.75 5.4% 14.68 4.9% 0.33%
16 16.02 0.1% 15.96 -0.2% -0.02%
18 17.52 -2.7% 17.52 -2.7% -0.15%
20 19.87 -0.6% 19.88 -0.6% -0.03%
22 22.18 0.8% 22.04 0.2% 0.01%
24 23.97 -0.1% 23.93 -0.3% -0.01%
26 24.72 -4.9% 24.71 -5.0% -0.20%
25
24
Gage Height (feet)
23
22
21
20
19
10000 11000 12000 13000 14000 15000 16000 17000 18000 19000 20000
Flow (cfs)
Figure 5 Rating curve comparison of critical calibration points for measured and modeled gage
heights
32
methods used to calculate loss in 1D and 2D models. The calculation processes of 1D and
2D models are closely related, but not exactly the same. The hydraulic roughness
parameter for the 1D model is used to account for the following: friction losses due to the
bed material of the channel or floodplain, form losses due to turbulence in a channel or
floodplain caused by channel geometry, drag losses caused by obstructions in the channel
or floodplain, form losses between cross sections due to variations in geometry and losses
due to channel bends. In the 2D model computation most of these losses are included in
the calculations (West Consultants Inc., 2017). Table 6 summarizes the forms of the
The bed friction (cf), viscosity/turbulence (vt) and Coriolis effect (fv) terms in the 2D
conservation of momentum equation account for some of the previously discussed losses
that have to be included in the hydraulic roughness parameter of the 1D model (USACE
Table 6 Conservation of mass and momentum equation comparison between 1D and 2D models
𝝏𝑯 𝝏(𝒉𝒖) 𝝏(𝒉𝒗) 𝜕𝑢 𝜕𝑢 𝜕𝑢
+ + +𝒒=𝟎 + 𝑢 𝜕𝑥 + 𝑣 𝜕𝑦 =
𝝏𝒕 𝝏𝒙 𝝏𝒚 𝜕𝑡
𝜕𝐻 𝜕2 𝑢 𝜕2 𝑢
−𝑔 𝜕𝑥 + 𝑣𝑡 (𝜕𝑥 2 + 𝜕𝑦 2 ) − 𝑐𝑓 𝑢 + 𝑓𝑣
𝜕𝑣 𝜕𝑣 𝜕𝑣
+ 𝑢 𝜕𝑥 + 𝑣 𝜕𝑦 =
𝜕𝑡
𝜕𝐻 𝜕2 𝑣 𝜕2 𝑣
−𝑔 𝜕𝑦 + 𝑣𝑡 (𝜕𝑥 2 + 𝜕𝑦 2 ) − 𝑐𝑓 𝑣 + 𝑓𝑢
Although the final Manning’s n-values vary significantly between the 1D and 2D models,
the difference does not necessarily make one model better suited for flood inundation
mapping than the other. The critical component of calibration is the quality of the
observed/measured data and the models’ ability to accurately represent those values. As
long as the appropriate steps are taken to calibrate the model to the observed data within
results. However, if using a 1D model the calibration process will be considerably faster,
due to less calibration model runs, if the base Manning’s n-values are increased slightly
Run Time
As was expected the run time for the 1D model was much faster than the 2D model.
Table 7 summarizes the model run times based on each computational task with the total
run time summarized in the final column. The 2D model took twenty-two minutes and
three seconds, while the 1D model only took forty-nine seconds. The largest difference in
run time (include percent difference between the run time of 1D and 2D) between the two
models is the computation of unsteady flow, which is the most significant task. The run
time for larger and more complicated projects is expected to be longer for 2D model
compared to 1D model, which makes this comparison a more significant factor during
flood events.
The significant increase in run time for the 2D model is explained by the inherently more
complex nature of the 2D model computations. There are multiple model factors that are
1D) and transverse variations in water surface elevation (neglected in 1D) (Colorado
35
Water Resources Board, 2006). Although the calculation of these factors increases the
model run time it also provides a more accurate and detailed model.
When considered on run time alone, it is clear that the 1D model is more favorable for a
flood inundation mapping project because a project that takes less time to complete is
typically a more budget friendly project. A faster model also allows for quicker
calibration. However, the computation factors that cause a 2D model to run slower are
factors that provide a more detailed model result; especially in areas where flow is
expected to spread out into the floodplain, as it would in a model for flood inundation
mapping. The difference in run time between 1D and 2D models raises important
questions. On one hand, the 1D model is likely better for the budget/scope of the project,
while the 2D model is likely to predict the flood inundation extents more accurately. This
Velocity
The significant differences in the velocities of the 1D and 2D model are most evident
when the velocities are at their maximum values. This occurs in the final time step of the
model when the flow rate and resulting flood inundation extents are also at their
maximum. There are three locations in the models that present significant differences
between the 1D and 2D velocities; they are indicated with black ovals in Figures 6 and 7.
In these locations, the 1D velocities are higher than the 2D velocities through the main
36
channel while the 2D velocities are higher in the overbank areas. There are two potential
The first potential cause for this difference and likely the most significant is the way in
model are calculated based on cross sections that are sub-divided into three categories:
left overbank, right overbank and channel. The overbank sections are further subdivided
into sections based on the n-value break points (USACE Hydrologic Engineering Center,
2016). By averaging the velocities in this way, the velocity gradients are smoothed over,
which can lead to slower velocities in the overbank area of the model. The 2D model
38
velocity is calculated for each individual cell which results in a more detailed depiction of
Figure 8 shows three 1D cross sections with the 2D model grid transposed over the top.
The figure shows the right overbank subsections of the top cross section with their
corresponding lengths. The total length of the right overbank is approximately 5,750 feet.
Based on that length and an average 2D cell width of 50 feet there are approximately 115
cells in the 2D grid, each calculating its own velocity, while the 1D velocity is calculated
for the three subsections that are then averaged to provide the right overbank velocity.
With this example, it is easy to see how the velocity can be underestimated in the
overbank area of the 1D model due to averaging, especially when there is minimal
Subsection 1 Subsection 2
~3670 ft ~1988 ft
Subsection 3
~92 ft
Figure 8 Right overbank subsections for 1D cross section transposed with the 2D grid
The second cause for lower overbank velocities in the 1D model is a result of
inaccurately defined ineffective areas. The ineffective areas define parts of the cross
39
section that contain water that is not being actively conveyed. They are defined in areas
of the model where water was likely to pond, however in many of these areas the
effectiveness of the flow is dependent on the water surface elevation. For example, a low
spot along a cross section may be ineffective until the upstream water surface elevation is
high enough to create a flow path through that low spot, turning it into effective flow.
Due to the varying nature of ineffective flow areas, setting them up can be tedious and
time consuming. Upon further review of the ineffective areas in the model, it is evident
that the elevations of some ineffective areas are too high, causing a decrease in
conveyance which explains the lower velocity in some overbank areas of the 1D model.
The higher channel velocities in the 1D model are a result of the lower velocities in the
overbank area. The flow conveyed at these points is the same in the 1D and 2D models.
So, if the overbank velocity and conveyance are underestimated in the 1D model then the
conveyance through the channel has to increase to satisfy the continuity of mass. The
higher velocities calculated in the channel of the 1D model create further implications for
other model variables. Due to the way water surface elevations are calculated in areas of
contraction and expansion a higher velocity can result in a lower water surface. This will
cross section averaging in 1D models, it is clear that the 2D model is better suited for
flood inundation mapping. The velocity results provided by the 2D model are more
40
detailed and therefor more accurate than the 1D model results. For the purposes of flood
inundation mapping, velocity is not a critical variable to report, but it is consequent to the
calculated water surface elevation which is critical to the accurate representation of flood
inundation extents. Even though, the velocity is not a critical variable to report in a flood
event, it can indicate the velocity of the flood wave which might have flood damage
strength implications.
The lower velocities in the 1D model that were caused by inaccurately defined ineffective
areas provides an additional argument for the use of a 2D model. The proper
engineering judgement during the set-up of model geometry. The 2D model geometry
set-up is faster and doesn’t require the same level of expertise since the areas of
ineffective flow are calculated inherently within the model. Therefore, the use of a 2D
model is more effective for the project scope and budget as well as the model results. In
addition, the time saved during model set-up could make up for the time lost due to
Modeled water surface elevations are the most important results produced during a flood
inundation mapping project, along with inundation extents, which will be discussed in the
following section. The water surface elevations produced by the model are used to predict
the maximum water depths at critical locations during a given flood event. The results
41
provided by this study show that on average, the maximum water surface elevations in
the 2D model are greater than the 1D model, with some exception. The maximum water
surface profile through the Napa River centerline in Figure 9 provides the evidence.
XS 11526
230
XS 7540
Zinfandel Ln.
220
Elevation (NAVD88)
210
200
190
180
170
160
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
Station (feet)
2D 1D Terrain
Table 8 presents the average difference in water surface elevation between the 1D and 2D
models at four model cross sections, from upstream to downstream: Pratt Avenue, Cross
Section 11526, Cross Section 7540 and Zinfandel Avenue calculated for each of the
The negative values in the table represent the exceptions, when the water surface
elevations of the 1D model are greater the 2D model. These instances occur at cross
sections 11526 and 7540, which are used to represent the section of the models
downstream from the Pope Street gage location used for calibration. The average
differences in these cases can be considered negligible for the purpose of comparing the
1D and 2D model results in this study because they are less than a tenth of a foot and
occur mostly at lower flow rates when flooding is minimal. However, it is important to
note that the profile view in Figure 9 shows the 1D water surface elevation to be
somewhat significantly higher than the 2D model at cross section 11526. This is true if
only the centerline profile is considered, however Figure 10 below shows the cross-
section view at this location where one can see that the 1D model’s water surface is
higher in the channel, but the 2D model is higher in the overbank. This is why the
average difference at this location is only -0.019 feet. The results appear this way because
the 1D model water surface elevation is calculated as one value for the entire cross
section, while the 2D model calculates the water surface elevation based on each
43
individual grid cell making the 2D model better suited for calculating flow in the
210
205
200
195
190
185
0 100 200 300 400 500 600 700 800 900
Station (feet)
2D - 7 1D - 7 Terrain
The average increase in the 2D model water surface elevations can be mostly explained
by expansion and contraction. The 2D model accounts for expansion and contraction
using the model terrain. The 1D model calculates headloss due to expansion and
Engineering Center,2016):
44
𝛼1 𝑉12 𝛼2 𝑉22
ℎ =𝐶| − | Equation 4.1
2𝑔 2𝑔
When there are high velocities in the channel of the 1D model and headloss due to
sections don’t capture all the nuanced detail in the terrain. Although it is possible to
headloss, it still won’t be as accurate as the losses determined by the grid computations of
the 2D model. When the headloss due to contraction and expansion is underestimated the
The evidence provided by the differences in water surface elevations between the 1D and
2D models demonstrates that the 2D model is better suited for the application of flood
inundation mapping. A 2D model will provide more accurate results necessary for
Inundation Extents
The differences in the inundation extents of the 1D and 2D models vary, as is evident in
Figure 11, which distinguishes the areas of significant difference between the 1D and 2D
model inundation extents with red ovals. These differences are due to the manner in
which the water surface elevations are calculated in the 1D and 2D models. Since the
model inundation extents are determined by the computed water surface elevations, the
The first significant difference in the inundation extents is seen in the downstream end of
the study area where the 1D model inundation extents are greater than the 2D model. The
second significant difference is shown in the middle part of the study area where the 2D
model inundation extents exceed those of the 1D model. The differences in both of these
cases are attributed to the calculation of a single water surface elevation at each cross
section in the 1D model. Previously explained by Figure 10, a single water surface
elevation is calculated at each cross section in the 1D model and that value is extended
46
along the cross section until it intersects with the terrain. This causes the inundation
model because the inundation extent is based on the water surface elevations calculated
in each cell. By calculating the water surface elevation in each cell, the 2D model is able
to more accurately determine overbank flow paths in the model which is further
supported by Figure 12. Figure 12 is a comparison of the water surface elevations for the
1D and 2D models at cross section 9352, which is shown in Figure 11. The comparison
shows the overbank flow paths determined by the 2D model calculation which result in
209
204
199
194
189
184
400 600 800 1000 1200 1400 1600
Station (feet)
2D - 5 1D - 5 Terrain
Comparison of the 1D and 2D models based on their inundation extents provides further
support that the 2D model is better suited for flood inundation mapping projects. The 2D
model provides more detailed and therefore more accurate representation of the overbank
areas of the model which is critical for producing accurate flood inundation maps.
48
CHAPTER 5
The results of the 1D/2D model comparison, discussed in chapter 4, demonstrate that the
2D model provides a more accurate representation of the Napa River floodplain and in
this case, is better suited for flood inundation mapping. Based on the five parameters used
for comparison: calibration, run time, velocity, water surface elevation and inundation
extents the 1D model provided some advantage over the 2D model, but ultimately the
detail provided by the 2D model results was more favorable. The Manning’s n-values of
the 1D model were on average 29% higher than the 2D model in order to produce the
same calibrated results. This is because the manning’s n-values of the 1D model are used
to account for a significant portion of the hydraulic losses within the model, while the 2D
model computation scheme accounts for those same losses allowing for a more accurate
The computation time for the 1D model was 21 minutes and 14 seconds faster than the
2D model, providing much quicker results. This is an advantage of the 1D model and an
important consideration when it comes to the schedule and budget of a project. A quicker
run time allows more time for sensitivity analysis and calibration of a model. However,
the time and level of expertise required to develop a quality 1D model can sometimes
outweigh the benefits of computational run time. This was made evident by the
inaccuracies in the set-up the ineffective flow areas of the 1D model that led to some
The 2D model provided a more detailed representation of the floodplain velocities. This
is because of the way the conveyance and velocity are calculated in each of the models.
In the floodplain of the 2D model, the velocities were calculated across each 50 foot grid
cell. In the 1D model, the velocities were computed along each cross section based on
were then averaged across the entire overbank region of each cross section which ranged
from approximately 1,100-5,600 feet in the right overbank of the model. This is
significantly larger than the 2D grid cell size A better representation of the floodplain by
more detailed velocity and conveyance calculations makes the 2D model a better tool for
The modeled water surface elevations and inundation extents are variables of the velocity
and conveyance of the model which means the 2D model also provides more detailed
water surface elevation and inundation results. In the 1D model the water surface
elevation is calculated for each cross section and then linearly interpolated between cross
sections to determine the inundation extents. As previously discussed for velocity, the
water surface elevation and inundation extents for the 2D model are calculated for each
gird cell. The more detailed results of the 2D model computed water surface elevations to
be more than 2 feet higher than the 1D model in some areas. A 2 foot decrease in water
surface elevation could significantly impact the inundation extents of a model which
The goal of hydraulic model development for flood inundation mapping is to produce
results that can be used to predict important information related to flood events and use
that information to develop emergency action plans that can be used to lessen the impacts
on the surrounding community. The accuracy of the maximum water surface elevations
and inundation extents provided by the model, coupled with real time USGS gage data
and NWS forecasts are critical to making informed operational decisions during a flood
event. For the purpose of this study the 2D model is best suited to inform this process.
The models produced by this study address a specific application of hydraulic modeling
for the development of flood inundation maps. The 1D and 2D models were developed
using high resolution terrain data and the resulting computations showed that the 2D
model provided more detailed results. The quality of the terrain data used for model
development is critical to the accuracy of 2D model results. In the absence of high quality
terrain data, a 1D model can provide better results. Beyond flood inundation mapping
there are other applications that can be modeled using 1D or 2D hydraulic models, such
as canals or levee breaches. Developing a model for a canal or any other application with
a well-defined channel is best for a 1D model. Modeling a levee breach where flow
spreads over a wide flat area without a clearly defined flow path is best represented by a
2D model (West Consultants Inc., 2017). Ultimately, the type of model used is dependent
on the objective of the study, the quality of the data and the experience of the modeler.
51
APPENDIX A
MODEL INPUT DATA
52
Napa Sulphur
Total River Creek
Discharge Discharge Discharge
Date Time (cfs) (cfs) (cfs)
12/1/2016 0:00 14 4730 4083 647
12/1/2016 8:00 4730 4083 647
12/1/2016 16:00 16 5930 5118 812
12/2/2016 0:00 5930 5118 812
12/2/2016 8:00 18 7610 6568 1042
12/2/2016 16:00 7610 6568 1042
12/3/2016 0:00 20 10600 9149 1451
12/3/2016 8:00 10600 9149 1451
12/3/2016 16:00 22 14200 12256 1944
FIS report peak discharge values for Napa River (FEMA, 2016)
55
FIS report peak discharge values for Sulphur Creek (FEMA, 2016)
56
FIS report Napa River 100 year flood profile , upstream end of model (FEMA, 2016)
57
FIS report Napa River 100 year flood profile, downstream end of model (FEMA, 2016)
58
FIS report Sulphur Creek 100 year flood profile (FEMA, 2016)
59
APPENDIX B
ADDITIONAL RESULTS
60
235
Water Surface Elevation (NAVD88)
230
225
220
215
210
205
200
0 100 200 300 400 500 600 700 800 900 1000
Station (feet)
2D - 1 1D - 1 2D - 4 1D - 4 2D - 7 1D - 7 Terrain
220
215
210
205
200
195
190
0 100 200 300 400 500 600 700 800 900
Station (feet)
200
Water Surface Elevation (NAVD88)
195
190
185
180
175
170
165
160
0 500 1000 1500 2000 2500 3000 3500 4000
Station (feet)
2D - 1 1D - 1 2D - 4 1D - 4 2D - 7 1D - 7 Terrain
62
63
64
65
66
67
68
REFERENCES
Brunner, G. W., P.E. (n.d.). Common Model Stability Problems When Performing an
City of St. Helena Draft General Plan(Rep.). (2015). St. Helena, CA: City of St. Helena.
Section 1 Detailed Method, Rep.). (2006). Denver, CO: Colorado Water Conservation
Board.
Conditions ed., Section 1 Alluvial Fan Hazards, Rep.). (2006). Denver, CO: Colorado
University.
Flood Insurance Study Napa County, California and Incorporated Areas(Rep. No.
Greif, R. (2016). Technical Memorandum: Flood Response Operations for Napa River in St.
National Weather Service Corporate Image Web Team. (2000, January 01). California
http://www.cnrfc.noaa.gov/graphicalRVF.php?id=SHEC1
70
The National Weather Service, & The National Oceanic and Atmospheric Administration.
U.S. Geological Survey Flood Inundation Mapping Science. (n.d.). Flood Inundation
https://water.usgs.gov/osw/flood_inundation/
Service Advanced Hydrologic Prediction Service. Retrieved November 12, 2017, from
http://water.weather.gov/ahps2/hydrograph.php?gage=shec1&wfo=mtr
USACE Hydrologic Engineering Center. (2009). HEC-GeoRAS GIS Tools for Support of
USACE Hydrologic Engineering Center. (2016). HEC-RAS River Analysis System User's
USGS 11456000 Napa R Nr St Helena CA. (n.d.). Retrieved November 12, 2017, from
https://waterdata.usgs.gov/ca/nwis/uv?site_no=11456000
West Consultants Inc. (2017). Two Dimensional Modeling Using HEC-RAS(3-Day Course,