Citation
A comparison of flood damage estimates using data developed by standard methods and GIS spatial analysis

Material Information

Title:
A comparison of flood damage estimates using data developed by standard methods and GIS spatial analysis
Creator:
Wilkening, Craig Robert
Publication Date:
Language:
English
Physical Description:
v, 103 leaves : illustrations ; 29 cm

Subjects

Subjects / Keywords:
Flood damage prevention -- United States ( lcsh )
Geographic information systems ( lcsh )
Flood damage prevention ( fast )
Geographic information systems ( fast )
United States ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 100-103).
General Note:
Submitted in partial fulfillment of the requirements for the degree, Master of Science, Civil Engineering.
General Note:
Department of Civil Engineering
Statement of Responsibility:
by Craig Robert Wilkening.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
36420015 ( OCLC )
ocm36420015
Classification:
LD1190.E53 1996m .W55 ( lcc )

Full Text
A COMPARISON OF FLOOD DAMAGE ESTIMATES USING DATA
DEVELOPED BY STANDARD METHODS AND GIS SPATIAL ANALYSIS
by
Craig Robert Wilkening
B.S., Colorado State University, 1982
A Thesis Submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Master of Science
Civil Engineering
1996


1996 by Craig Robert Wilkening
Ail Rights Reserved


This thesis for the Master of Science
degree by
Craig Robert Wilkening
has been approved by
Lynn E7~3otlnson
AA/j, /99L
Date


Wilkening, Craig Robert (M.S., Civil Engineering)
A Comparison of Flood Damage Estimates Using Data Developed by Standard
Methods and GIS Spatial Analysis
Thesis Directed by Associate Professor Lynn E. Johnson
ABSTRACT
The catastrophic flooding in the midwest in 1993 and in California in
1995 resulted in flood damage estimates in the hundreds of millions to billions
of dollars. Throughout the U.S., millions of structures are unprotected from
flooding because of past floodplain management policies, budget limitations,
and/or lack of adequate information on floodplain boundaries. As communities
along rivers and streams become more aware of potential flood hazards,
increased emphasis will be placed on the development of flood control
improvements and the effects of these improvements on the water surface
elevations of various frequency floods. Equally important will be the analysis of
the costs and benefits of these flood control improvements. Flood damage
analysis methods in the past have required tedious inventories of the finished
floor elevations of structures in a floodplain in order to determine the water
surface elevation-damage curves for each land use in a stream reach. In this
study, a Geographic Information System (GIS) was used to develop elevation
data for the structures in the floodplain along a reach of Clear Creek in Wheat
Ridge, Colorado. Using the surface modeling capabilities of a GIS, GRID and
TIN models of the study area were constructed from a U.S.G.S. Digital
IV


Elevation Model (DEM). Structure location and floodplain coverages were
superimposed on the GIS surface models and the structure elevation data was
determined using an on-screen query of the graphical and attribute databases
in the GIS. The elevation information was input into flood damage analysis
models and compared to a baseline condition to determine the effectiveness of
the GIS for data development. The results of the analysis show that the 30
meter (98.4 feet) resolution and 1 meter (3.3 feet) elevation increments of the
DEM are too coarse for the accurate data development required for flood
damage analysis modeling. The results also indicate that the data developed
from the GIS surface models can be used to quantify total flood damages and
provide information on the number of flooded structures within various flood
ranges. Research of other data sources with finer resolutions and/or
adjustments to the characteristics of the surface models is recommended for
future analysis.
This abstract accurately represents the content of the candidates thesis. I
recommend its publication.
Signed
nn E.Johnson


DEDICATION
Throughout the course of my graduate studies, both at the University of
Texas at Arlington and at the University of Colorado at Denver, one person has
been my source of strength, my inspiration, and has often had more confidence
in me than I had in myself. It is with the deepest and sincerest love and
devotion that I dedicate this thesis to my loving wife and best friend, Paula.
Without her patience, understanding, and encouragement, many of the
accomplishments I have had in my life, including this thesis, would not have
been possible.
I would also like to dedicate this thesis to my son Carey, who is also a
source of inspiration in my life. Although he is only three years old, it is my
hope that my academic achievements can be a source of encouragement for
his future learning.


ACKNOWLEDGMENTS
This thesis is the direct result of my interest in Water Resources and
Geographic Information Systems (GIS). My coursework at the University of
Texas at Arlington and at the University of Colorado at Denver has provided
me with many opportunities. Throughout my graduate studies and the
preparation of this thesis, I received assistance from many individuals, both
personally and professionally.
I would like to thank Dr. Max Spindler, Associate Professor of Civil
Engineering at the University of Texas at Arlington. As my first advisor, he
allowed me the opportunity to begin my graduate studies in Civil
Engineering. I am most grateful for his initial encouragement.
I would also like to express my appreciation to Mr. Troy Lynn Lovell
of Albert H. Halff Associates, Consulting Engineers and Scientists, Fort
Worth, Texas. The knowledge and experience I gained from him through
the first five years of my professional career were invaluable, and I will
always be thankful for his guidance.
Special thanks are also expressed to Mr. Mike Burnham and Ms.
Donna Lydon of the U.S. Army Corps of Engineers, Hydrologic Engineering
Center, Davis, California. The information and guidance they provided on


present and future flood damage modeling techniques were highly
appreciated and I am looking forward to assisting them with the beta testing
of the next generation of flood damage analysis modeling software.
I would like to express my appreciation to Mr. Greg Bryant of the
Denver Water Department. As my ARC/INFO instructor during the Spring
Semester of 1995, his professionalism and encouragement motivated me to
continue to learn about GIS techniques, applications, and concepts and I
am thankful for his help during the preparation of this thesis.
I would also like to thank Dr. William Hughes, Professor of Civil
Engineering, University of Colorado at Denver, and Dr. John Liou, Regional
Hydrologist, Federal Emergency Management Agency. As members of my
graduate committee, their comments and suggestions for the content of this
thesis were invaluable, and I am most grateful for their assistance.
I would like to extend my highest regards and personal thanks to Dr.
Lynn E. Johnson of the University of Colorado at Denver for serving as my
graduate advisor. His dedication to the development of a GIS program in
the Civil Engineering Department as well as his encouragement,
experience, and assistance will always be highly appreciated.


Finally, I humbly acknowledge that the presentation of this report and
any contributions it might make in the future are through the abilities
granted to me by our Creator.


CONTENTS
Figures................................................xii
Tables.................................................xiii
CHAPTER
1. INTRODUCTION.......................................1
2. LITERATURE REVIEW..................................4
3. THEORY AND OBJECTIVES.............................23
Flood Damage Analysis Modeling...............24
Structural Inventory of Damages..............26
Expected Annual Damages......................27
Geographic Information Systems (GIS).........30
GIS Purpose............................31
Model Integration......................34
Statistical Analysis...................35
Advances in Technology.................36
Spatial Data Accuracy..................36
GIS Analysis...........................37
Legal Issues...........................38
Political Issues.......................41
4. CASE STUDY........................................43
Clear Creek Hydrology and Hydraulics.........43
Standard Flood Damage Modeling...............47
Flood Damage Modeling Using GIS..............55
GRID Modeling..........................58
TIN Modeling...........................58
5. RESULTS ..........................................60
GRID Modeling................................61
TIN Modeling.................................67
6. CONCLUSIONS AND DISCUSSION........................73
GRID Modeling................................74
TIN Modeling.................................77


Coordinate Transformation issues.......81
Probability of FFE Differences.........83
7. RECOMMENDATIONS..............................89
GLOSSARY OF TERMS .....................................97
REFERENCES .................................100
XI


FIGURES
Figure
3.1 Summary of Stage-Damage-Frequency Relationships..........29
3.2 GIS and Related Technologies...............................32
3.3 GIS Structure..............................................33
3.4 GIS Analysis Functions.....................................40
4.1 Clear Creek Watershed......................................44
4.2 Study Area.................................................45
4.3 100-year Floodplain Delineation............................46
4.4 Frequency-Discharge Curve..................................49
4.5 DEM Plan View..............................................56
4.6 DEM Three Dimensional View.................................57
6.1 Comparison of GRID and TIN Data Structures.................79
6.2 Relative Frequency of FFE Differences......................87
6.3 Cumulative Relative Frequency of FFE Differences...........88
xii


TABLES
Table
1.1 Structural and Non-Structural Floodplain Improvement
Alternatives..............................................2
4.1 Computed Finished Floor Elevations for Floodplain
Structures................................................51
4.2 Corps of Engineers, Structural Stage-Damage Data..........54
4.3 Corps of Engineers, Contents Stage-Damage Data............54
5.1 Comparison of Finished Floor Elevations Determined by
Standard Methods and GRID Surface Modeling Techniques.....62
5.2 Summary of Computed Damages, Standard Methods
and GRID Data Model.......................................65
5.3 Summary of Number of Flooded Structures, Standard
Methods and GRID Data Model...............................65
5.4 Comparison of Finished Floor Elevations Determined by
Standard Methods and TIN Surface Modeling Techniques......69
5.5 Summary of Computed Damages, Standard Methods
and TIN Data Model........................................72
5.6 Summary of Number of Flooded Structures, Standard
Methods and TIN Data Model................................72
7.1 Summary of Data and Analysis Functions....................96
xiii


CHAPTER 1
INTRODUCTION
During the Summer of 1993, catastrophic flooding occurred in the
Mississippi and Missouri river basins throughout the Midwestern U.S. In the
Spring of 1995, many areas throughout California also experienced severe
flooding. These flood events were the direct result of excessive rainfall
occurrences and resulted in flood damage estimates ranging from hundreds
of millions to billions of dollars. Although these are only two examples of
extreme flooding in recent years, millions of structures throughout the
United States are still unprotected from flooding because of past floodplain
management policies, budget limitations, apathy, and/or lack of adequate
information on floodplain boundaries.
People living in communities along rivers and streams are becoming
more aware of potential flood hazards. Despite the lessons of the past,
homes, schools, businesses, and numerous other types of structures are
still being constructed in flood prone areas, and people affected by flooding
will continue to pressure federal, state, and local governments to improve
the drainage conditions of these areas. Increased emphasis has been
placed on the development of structural and non-structural flood control
improvements and the effects of these improvements on the water surface
elevations of various frequency floods.
As flood control measures are developed, the reduction in flood
damages resulting from the improvements must also be analyzed. Flood
l


damage analysis provides the engineer and/or floodplain manager with the
quantitative information needed to assess the social cost of flooding and to
provide data for formulating, evaluating, and implementing floodplain
management alternatives. A flood damage analysis also provides data on
the long term economic costs and benefits of proposed improvements.
Structural and non-structural measures, such as those shown in Table 1.1,
can all be effectively analyzed.
TABLE 1*1 Structural and Non-Structural Floodplain Improvement Alternatives
Structural Alternatives Non-Structural Alternatives
Detention Reservoirs Flood Proofing
Levees Structure Relocation
Channelization Floodplain Management
Diversions Land-Use Zoning
Lift Stations and Gravity Outfalls Purchase of Flood Prone
Structures
No Action
Flood damage analyses of existing developments in flood prone
areas provide information on critical problem regions and allow insurance
agencies to calculate flood insurance premiums. Damages computed in the
aftermath of major flood events provide data to relief agencies for the
effective allocation of funding and other emergency assistance. Damages
2


computed for future developments can provide local agencies with the data
necessary to make wise land use and floodplain management decisions.
Since the late 1960s, aerial photography and satellite imagery have
provided extensive information on floodplain limits and flood damaged
areas. Before the development of this remote sensing technology,
floodplain and damage data had to be developed by tedious and time
consuming field survey methods. These field surveys generally required
weeks or months to complete, and the conclusions developed from the
analysis of the compiled data often came after the damaged areas were
reconstructed. As remote sensing technology became more sophisticated
and widely used, it was found that this digital imagery provided engineers
with real-time data that could be quickly analyzed as the flooding occurred.
In the mid 1980s, the development of Geographic Information
System (GIS) technology further enhanced the capabilities of decision
makers. A GIS database consisting of existing digital terrain data and aerial
imagery from remote sensing systems, provides a sophisticated, accurate,
immediate, and cost effective tool for the analysis of flood hydraulics,
improvement alternatives, and emergency contingency plans. As computer
hardware and software continues to develop and become more
sophisticated, the modeling capabilities and data available to engineers,
floodplain managers, and other decision makers will certainly reduce the
time and enhance the level of response to future flood disasters.
3


CHAPTER 2
LITERATURE REVIEW
The use of GIS in Water Resources Engineering and Floodplain
Management generally began in the late 1960s. During this period, GIS
systems consisted primarily of aerial photographs and digital satellite data
(remote sensing) and the use of this data to interpret the spatial extent of
floods on major river systems and watersheds. The earliest documentation
of the use of remote sensing systems in water resources was described by
Robinove (1967). This study addressed the potential of the use of remote
sensing because of the inadequate coverage of traditional resource surveys
and the need for updated basic information for development planning.
Another benefit was the fact that large areas could be evaluated
economically and in a short period of time. Several examples describing
the use of remote sensing are presented. These examples include space
photography, infrared aerial and satellite photographs of flow patterns, and
side-looking radar images of topographic features. The study concluded
that the use of remote sensing systems was limited but had important
applications. Hydrologists would be required to increase their abilities in
the use and interpretation of this data and an accelerated research program
would need to be developed to assess the benefits from more sophisticated
aerial sensors.
In the early 1970s, the practical use of remote sensing systems was
described in a report by Ruff, Keys, and Skinner (1974) on the Clarks Fork
4


of the Yellowstone River Watershed. In this analysis, the study of the
floodplain was not an objective. However, it did show that the use of color
and thermal infrared (IR) photography, ground truthing, and high-level and
low-level aerial photographs could be used as a practical and cost effective
method for analyzing the sediment problems on the Clarks Fork of the
Yellowstone River. Color IR photography effectively detected the changes
in the concentrations of suspended particles and could be used to
qualitatively assess the suspended material concentrations between
tributary inflows and the main channel. The thermal IR images provided an
indication of the relative temperature differences in different portions of the
stream channels. The mixing zones could also be clearly identified from the
photographs. This study identified 106 existing and potential sediment
sources to the Clarks Fork of the Yellowstone River and concluded that the
combination of thermal and color IR images provided a unique perspective
of the sediment transport process in the watershed.
The use of radar in Urban Hydrology was discussed by Austin and
Austin (1974). In this study, radar images were used to analyze rainfall
events that led to flooding in house basements between 1969 and 1972.
The rainfall events included slow moving frontal storms as well as storms
that were elongated in the direction of their travel. The study found that
radar records were ideal for analyzing the spatial and temporal
characteristics of flood producing storms and provided data that allowed
scientists to identify and classify the storms. The use of radar data was
found to be a useful tool in Urban Hydrology because the resolution of the
data was well matched to urban areas. The radar data was calibrated to
rainfall data from several rain gauges in the study area and was found to be
5


useful in terms of supplementing existing rain gauge data and in predicting
rainfall from the radar images.
A comprehensive study by Davis (1978) was used to test various
analytical methodologies for floodplain information and analysis. This study
contains some of the earliest documented use of GIS techniques for
watershed modeling, floodplain analysis, and quantification of flood
damages. The purpose of this study was to assess existing and future flood
hazards, damage potential, and environmental concerns that allowed local
governmental agencies to make sound decisions about future development
strategies. This study utilized a complex combination of U.S. Army Corps of
Engineers hydrologic and hydraulic computer models that accessed a
common grid cell database. These computer models are collectively known
as the HEC-SAM (Spatial Data Management and Comprehensive Analysis)
system. The information stored in the grid cell database included the soil
classification, hydrologic soil group, land slope, existing and future land
use, and the cell elevation. In one portion of the study, the parameters for
an HEC-1 model of the watershed were automatically generated from the
information stored in the grid cell database. Once the HEC-1 model was
completed, the discharges were input into an HEC-2 model of the stream
reach in order to determine the water surface elevations for various flood
events.
The discharges and water surface elevations from the HEC-1 and
HEC-2 models were then used to perform a flood damage analysis. The
analysis evaluated the potential damages of alternative land use patterns
and development scenarios. A unique elevation-damage function was
developed by automated analysis of the land use and topographic data
6


contained in each grid cell in the floodplain. The total damages for each
land use were computed based on the data at each grid cell. This study did
not quantify the damages at individual structures. It was determined that a
general damage analysis, using composite damage relationships, was
appropriate for the assessment of alternative land use patterns. The
composite damage relationships were created by averaging the structural
and content values of field data for each of the land use categories being
analyzed. The grid cell database and the composite damage relationships
were accessed by the U.S. Army Corps of Engineers DAMCAL program.
The DAMCAL program then calculates the elevation-damage functions for
the land use categories and damage reaches in the study area. Alternative
land use development scenarios could be easily evaluated once the
existing conditions and baseline data was generated.
The HEC-SAM system was also used to perform an environmental
and water quality assessment along with the floodplain and flood damage
analysis. The study concluded that despite the tedious data development
and the required attention to detail, the spatial data management
techniques used for the analyses could perform comprehensive
assessments quickly and systematically.
Sollers, Rango, and Henninger (1978) used multispectral aircraft and
satellite images to evaluate the potential benefits of remote sensing for
floodplain analysis. In this study, several test sites were chosen to evaluate
the applicability of digital aircraft and satellite data for use in floodplain
mapping. The results indicated that the digital aircraft data could not be
used to delineate a continuous floodplain boundary. It was found that the
study area had very complex topographic and land cover differences and
7


the small pixel size of the aircraft data (high resolution) resulted in an
overabundance of detail that shadowed the floodplain boundaries in several
areas.
Further research in more uniform study areas resulted in
considerably better results. Floodplain boundaries derived from satellite
data were found to be comparable to delineations based on engineering
modeling and computations. The lower resolution of the Landsat data was
more effective in identifying floodplain boundaries. The larger pixel size in
satellite data produced radiance values in flood prone areas that were
greater and easier to identity than those derived from aircraft data. The
study concluded that the use of aircraft and satellite digital data was
preferred over conventional photo interpretation methods. It was also
concluded that the remote sensing techniques would be better used as a
form of preliminary planning and as a means of verifying ongoing floodplain
studies. The multispectral data could provide detailed information on land
uses in flood prone areas, could be used as a base to assess potential
flood damages, and would provide a means for verifying community
compliance with the guidelines of the National Flood Insurance Program.
A unique use of Landsat satellite data was described in a study by
Green, Whitehouse, and Outhet (1983). The purpose of the study was to
use the digital images to analyze the streamlines of rivers at various flood
stages and determine if they could be used as indicators of floodways. The
streamlines are a direct result of the differences in sediment concentrations
and particle sizes found in the river at various flood stages. In the study, it
was found that the streamlines could be used to formulate channel
improvements. The convergence and divergence of the streamlines were
8


aerial indicators of acceleration and deceleration zones. This was found to
be useful in the design of structural flood protection measures because the
limits of these velocity zones could be verified through the correct visual
interpretation of satellite images. Once the velocity zones were determined,
flood protection measures could be designed to fit into the stream reach
instead of using designs that were based on limited velocity measurements
and theoretical computations that are sometimes difficult to verify.
The images used in the study showed that the streamlines were best
defined during or immediately after the flood peak and that areas of erosion
and deposition generated light and dark stream lines, respectively.
Stretching the contrasts in the Landsat images enhanced the light and dark
streamlines and simplified the identification of erosion and deposition areas.
The results of the study indicated that Landsat images are ideal for
locating and mapping floodways. Images of slow moving floods, typical of
the study area, showed light and dark streamlines indicating the occurrence
of erosion and deposition. When interpreted accurately, these images
could be used to pinpoint the locations of flood control and flood mitigation
measures that can ultimately be more feasible and economical.
Another example of floodplain delineation and damage assessments
was highlighted in a study by Imhoff, Vermillion, Story, Choudhury, Gafoor,
and Polcyn (1987). In this analysis, Shuttle Imaging Radar (SIR) and
Landsat multispectral data were used to map flood boundaries and assess
flood damages in Bangladesh. In the past, many developing and third world
countries have found this data to be very useful because of the absence of
aerial survey coverage and the difficulties associated with the thick and
extensive cloud cover that accompanies the seasonal monsoon cycles.
9


Because of these limitations, scientists, agricultural engineers, and
planners were unable to determine the interactions between their projects of
interest and the floods generated by monsoon rainfalls. The objective of
this study was to demonstrate the potential benefits of radar imaging
systems for flood boundary delineations, as an aid for agricultural and
infrastructure planning, and to quantify flood damages from monsoon
storms.
The use of Landsat images and aerial surveys in past studies was
subject to the absence of cloud cover. The use of Shuttle Imaging Radar
(SIR) was found to not only allow flood mapping in areas of heavy cloud
cover, but also beneath heavy vegetation canopies. In 1984, the Space
Shuttle Challenger was used to collect the imaging data in an area along
the Ganges River with a densely populated floodplain under heavy rice
cultivation. The basic approach of the study was to acquire Landsat and
SIR data during various flood events and compare the data to flood
measurements obtained in the field.
Coincidentally, a monsoon rainfall and flood occurred in September
of 1984 while the Space Shuttle Challenger was in orbit. The Landsat data
of this event was used to define the existing flooding. The SIR data was
taken approximately 5 days after the acquisition of the data from the
Landsat satellite. During this period, most of the flood waters had receded
to their natural boundary. A comparison of these two data sets were used
to identify the land cover classes and areas that were subject to flooding.
Data was also obtained from a stream gauge in the Ganges River
immediately outside of the study area. This stream data was compiled for
10


the conditions present in the Ganges River during the Landsat and
Challenger SIR overflights.
The results of the study indicated that the Landsat data was useful
for identifying flooded areas and delineating floodplain boundaries.
However, in terms of a flood damage assessment, the Landsat data could
not be used to spectrally separate the village land use class from
agricultural areas. Similar to other countries that have an economy based
on rice agriculture, the dwellings in this part of Bangladesh are actually
located throughout the agricultural fields. This pattern was significant
because infrastructure flood damages could not be distinguished from
agricultural damages. It was also found that the resolution limitations of the
Landsat data caused errors in the delineation of the floodplain where
agricultural areas were adjacent to flooded areas. The presence of silts
and clays in the flood waters also made the classification of surface
features more difficult.
For the SIR data, the land use classifications were found to be more
accurate than the Landsat data. The improvements in the delineation of the
flood boundaries and the assessment of damages were primarily due to the
smaller spatial resolution of the SIR data. The separate village and
agricultural land uses were well defined on the images. However, the
turbidity of the water had no effect on radar backscatter and as a result, the
SIR images could not be used to distinguish between flooded, ponded, and
flood-irrigated areas. Despite the problems in the land use classifications,
the results of the study indicated that the SIR data had an overall accuracy
of classification of 85% as compared to an accuracy of 77% for the Landsat
data. The study concluded that regardless of the classification problems
11


and the accuracy results, the data obtained from the Landsat and SIR
systems would certainly be more accurate than ground surveyed data.
Prior to the mid 1980s, the majority of the research utilized Remote
Sensing technology. Although Remote Sensing techniques are still widely
used, these methods have been generally considered part of an overall
Geographic Information System. Since the mid 1980s, the development of
hardware and software systems has become much more sophisticated.
During this period, the term Geographic Information System has become
much more predominant and related technologies such as Remote Sensing
Systems, Automated Mapping Systems, and Database Management
Systems have been generally grouped under the heading of a GIS.
The integration of GIS and HEC-2 for floodplain management was
highlighted in a study by DePodesta, Nimmrichter, and Scheckenberger
(1991). In this study, HEC-2 models were interfaced with the Geo/SQL
spatial database software. The computer link between HEC-2 and
Geo/SQL was intended to generate graphical displays of error messages or
other hydraulic problems, custom cross section and stream profile computer
plots, and interactive displays of photographs. The link was also designed
to transform information contained in the database into a HEC-2 input data
set as well as provide a method for interactive floodplain delineation and
interactive alteration of bridge and topographic data. Using this system,
the engineer would spend less time on mapping and data management and
more time on hydraulic analyses, formulation of channel improvement
alternatives, and identification of specific areas of flooding and erosion in
the channel. In the study, it was found that the most significant constraint in
the development of floodplain maps were the tasks that required interaction
12


with the topographic maps that were used for the study area. This resulted
in a loss of valuable time because engineers and technicians were required
to piece maps together, obtain properly scaled mapping from different
sources, analyze contour data, delineate floodplain boundaries, and
prepare maps for study reports. The study was intended to automate much
of the mapping and data management allowing the engineers and
technicians to gain a better understanding of the hydraulics of the stream
channel.
The use of GIS in identifying seasonal flooding is highlighted in a
study by Lipschultz and Glaser (1992). The study was done in the
aftermath of heavy rainfall in an area of west central Florida that received
17.5 inches of rain in a 12 hour period. Damage estimates were nearly 4
million dollars and local residents were questioning why flood control
improvements were not being made fast enough. In this study, the water
district engineers needed a way to visually analyze the effects of capital
improvement projects on homes and businesses in the floodplain. A GIS
was used to help with the tracking and planning of the capital improvement
projects planned for the study area.
A one square mile pilot area was chosen to test the effectiveness of
the GIS. Road data, floodplain and property boundaries, and structure
values were input into the GIS database constructed for the pilot study.
Using this data, the benefits of capital improvements could then be
assessed. During a 1977 flood, it was determined that 1,300 homes were in
the floodplain and were damaged to various extents. A test of the GIS
database showed that the number of flooded homes would decrease to 321
using drainage improvements targeted in the capital improvement program.
13


It was estimated that the savings in annual property damage premiums in
the flooded area would be as high as $55,000.
The ultimate goal of the water district conducting the study was to
link other maintenance, management, accounting, and database software
packages to the GIS system. These software links, combined with updated
topographic data, would allow water district engineers to make more
accurate assessments of the flooding problems within the district
boundaries. It would also allow engineering designs to be tested in the GIS
before they are constructed. This data would provide engineers with the
tools necessary to make better design decisions.
Another example of the use of GIS technology for floodplain analysis
was evident during the extreme flooding in the Midwest in the Summer of
1993. Mauney and Bottorff (1993), documented the GIS analysis
techniques used to assist the Federal Emergency Management Agency
(FEMA) with response and disaster relief efforts during the flooding. The
monetary losses to houses, businesses, and agriculture were estimated in
the billions of dollars. For the analysis of this flooding, FEMA assembled a
rapid situation assessment mapping team to inventory flooded structures
and transfer the information to FEMA and U.S. Army Corps of Engineers
GIS Specialists. Using Global Positioning System (GPS) technology and
GIS data analysis, the team was able to produce maps within a few days of
the data collection phase. These maps were used for disaster response,
recovery efforts, and risk mitigation. The team was able to demonstrate that
GIS technology could quickly compile and convert observations of disaster
conditions to paper maps and digital databases. The fast and efficient
14


development of this data could be used by numerous agencies for disaster
relief efforts.
The team was also able to record the boundaries of the floodplains
and locations of levee breaches by flying directly over the waterline and
logging the GPS coordinates. The mapping of this information was also
completed in a matter of days and transferred to local and federal officials.
Throughout August 1993, FEMA teams drove through each town
along the Mississippi River while the floodwaters receded. These teams
recorded data on every residential, commercial, and mixed-use structure
that was flooded greater than a depth of one foot. Photographs and GPS
coordinates were also collected for each damaged structure. The resulting
database that was compiled was used to produce maps for building
inspectors and FEMA insurance adjusters. These maps were used to
advise citizens and local authorities of available assistance and the
requirements for reconstruction of the structures damaged by the flooding.
The information contained in the GIS and the subsequent mapping was
critical for responding to a disaster of this magnitude. It was found that the
GIS was capable of providing information that would normally have to be
collected in the field over a period of weeks to months. The study also
showed that the GIS not only maximized resources, but also helped FEMA
respond much more quickly in a continuously changing situation.
The use of remote sensing and GIS technology for the flooding in the
Midwest was also described by Corbley (1993). During the first week of the
1993 flooding, the demand for imagery of the affected areas continued to
grow. GIS Specialists from EOSAT (Lanham, Md.) and ERDAS (Atlanta,
GA.) realized that the availability of Lansdat images had to keep pace with
15


the increasing demand. Data needed to be provided to relief agencies in
the shortest period of time. Within 48 hours of satellite image acquisition,
the data was being used by FEMA, the U.S. Army Corps of Engineers, and
other federal and local agencies to assess damages to crops, bridges,
roads, and buildings and to determine where the relief efforts should be
focused.
During the flooding, the primary requirement by FEMA was to
determine the type of land use and how much of it was under water. FEMA
used the GIS software to compare the flood limits shown in the new images
to archived images. From this data, classification maps of the flooded
areas were created and provided the detail needed to determine the types
of areas that were flooded and whether highways and railroads were
passable. This information was found to be crucial in designing a
comprehensive relief effort. Teams from the U.S. Army Corps of Engineers
were also on site throughout the flooded areas assessing damages to
levees and offering assistance to victims. Much of the work done by the
Corps of Engineers required decision making in the field. The remote
sensing images and the GIS technology were used to generate hard-copy
maps that allowed the field personnel to see what was going on around
them.
Long after the flood waters had receded, it was expected that the
Landsat imagery would continue to play a vital role in the cleanup and
rebuilding of the damaged areas. The U.S. Army Corps of Engineers and
other federal agencies were expected to evaluate the information gained
from satellite data in order to develop response and emergency planning
procedures to lessen the impacts of future disasters. The Landsat data was
16


also to be used by FEMA to test the accuracy of existing flood models. The
satellite data would be analyzed to see of the flood waters had behaved as
their computer models had predicted.
Studies by Juhl (1993) and Frazier (1995) outline the implementation
of a GIS system in Jefferson Parish, Louisiana. One of the goals of the GIS
system was to evaluate alternatives to minimize flood damages and to help
improve drainage within the parish. The parish integrated the GIS with their
supervisory control and data acquisition (SCADA) system, thus improving
the overall flood control efforts. The majority of the parish is below sea
level and floodwaters are pumped over the levees protecting the parish and
into tributaries of the Gulf of Mexico. In the 15 years prior to the study, the
parish experienced millions of dollars in flood damages because the volume
of runoff often exceeded the ability of the pump stations to handle the flows.
By integrating the GIS with the SCADA system, parish engineers and
managers were able to model flows in the runoff collection system. Using
rainfall and pump operation data, the model could also be used to predict
the effectiveness of the collection system. This data was invaluable in
formulating capital improvement options.
The GIS was also used to map the distribution of FEMA flood
insurance claims. The GIS data has shown that investments in drainage
infrastructure could save FEMA hundreds of millions of dollars in flood
damage claims. Using this data, the Corps of Engineers could identify the
areas of the parish that experienced the most flood damage. The parish
was able to prove that the costs of the damage claims alone were greater
than the costs of the original facilities that would have prevented the
damages from occurring in the first place. The GIS has been adopted by
17


many other departments in the parish and has provided engineers and
planners with a sophisticated tool to quickly and effectively manage a large
amount of information. It has also given the parish the ability to contend
with complex water, drainage, and stormwater issues.
The use of ARC/INFO GIS software for hydrologic modeling was
summarized in a study by Warwick and Haness (1994). GIS systems have
data management features that are extremely adaptable to environmental
analysis problems and the goal of this paper was to identify the role of
professional judgment in the use of GIS and ARC/INFO technology. This
paper also summarized some problems and inaccuracies that were found in
ARC/INFO applications. In general, the authors found that ARC/INFO was
quite effective in determining basin parameters that are tedious to compute
manually. They also felt that the computed outflow hydrographs from the
model developed by the GIS were probably no more accurate than those
calculated by traditional methods. However, they did conclude that the use
of GIS for large projects with complicated data sets is economically
justifiable and a GIS is ideally suited for continuous updating and quick
computations of alternative scenarios. The authors determined that in some
cases, the use of a GIS for hydrologic modeling may be excessive for the
analysis required, analogous to performing simple word processing on a
super computer. They also concluded that the growing sophistication of
spatial information and analysis tools will drive the development of more
sophisticated hydrologic analysis models.
A flood management system for Bangladesh is the subject of a study
by Pandyal and Syme (1994). The country of Bangladesh lies in a delta at
the confluence of three of the largest river systems in the world. The total
18


drainage area of the Ganges, Brahmaputra, and Meghna Rivers, above
Bangladesh, is approximately 560,000 square miles. More than half of the
country lies below an elevation of 40 feet and in a normal year, over 30% of
the total agricultural area is flooded. Over 50% of this same agricultural
area is subject to monsoon or tidal floods. Severe floods in 1987 and 1988
resulted in the loss of over 3000 lives. These floods also caused
widespread damage to crops, roads, and cities and were a major setback to
the entire Bangladesh economy. In the aftermath of these floods, the
government underwent a major reassessment of their flood control policy.
In 1989, the government launched a Flood Action Plan (FAP) that required
the study of major flood discharges, storages, modeling, and analysis.
However, the FAP was not able to detail areas and depths of inundation.
As a result, a Flood Management Model (FMM) was developed to provide
more detailed information on floodplain limits and inundation. It was
determined that the most effective flood management strategy would be to
utilize computer models as well as GIS spatial analysis.
The use of floodplain modeling and GIS technology was successful
in developing maps of flood depths and inundation limits and was effective
in allowing decision makers to compute the costs of different flood events
and flood protection alternatives. The GIS was used to develop this
information very quickly. The study concluded that the integration of the
flood analysis models and the GIS accelerated the study of existing flooding
problems and proposed flood control alternatives.
Another application of the use of GIS and satellite imagery, for the
flooding on the Mississippi River in 1993, is illustrated by Speed (1994).
The U.S. Army Corps of Engineers (COE) and the Federal Emergency
19


Management Agency (FEMA) were the federal agencies responsible for
damage assessments and disaster relief during the 1993 flooding. The
study outlines how the COE and FEMA worked with other government
agencies to evaluate the extent of the flooding and formulate rebuilding
efforts. Identifying areas threatened by floods and coordinating the
evacuation of residents in affected areas required a sophisticated
combination of GIS software and remote sensing technologies. During the
earliest stages of the flooding, the COE and FEMA had no single or reliable
source of information to assess the severity of the floods or to coordinate
emergency response strategies. In the aftermath of Hurricane Andrew in
1992, the relief efforts by FEMA were heavily criticized. The COE and
FEMA determined that information needed to be in the hands of decision
makers as quickly as possible. Using satellite imagery and aerial
photography, flood maps of the entire basin from the Canadian border to
the Gulf of Mexico were generated within 24 hours.
The Soil Conservation Service (SCS) was also in need of rapid data
generation and analysis. Traditional methods used to assess acreages of
damaged farmlands, forests, and urban areas, were too costly and time
consuming. The solution adopted by the SCS was the use of Landsat
satellite imagery to compute the limits of the flooding and the extents of
damaged areas. During this crisis, many of the GIS industry vendors
donated staff and equipment to assemble a database that could be used by
the government agencies responding to the floods. The database
generated by these efforts allowed the SCS, COE, and FEMA to quantify
damages to crops, roads, bridges, and other transportation facilities. It was
also used to update flood hazard maps, locate electrical lines and sewers,
20


plan emergency response efforts, locate water treatment plant and
superfund sites, and calculate population densities.
This study concluded that the knowledge gained from the use of GIS
technology for the Mississippi River flooding will give federal, state, and
local agencies the potential to respond much quicker and more effectively in
future disaster situations. The study also found that advancements in
higher resolution satellite imagery, GIS software, and environmental models
will also enhance the effectiveness of government agency responses to
future disasters.
The implementation of a GIS system for another severe flood in 1993
was the subject of a paper by Bocco, Sanchez, and Riemann (1995). The
system was established to assess severe flooding in Tiajuana, Mexico in
January, 1993. Rapid industrialization in Tiajuana over the past several
years has resulted in highly accelerated urban growth without adequate
planning efforts. Flash flood areas, characterized by valley bottoms with
steep slopes, are often inhabited by low income families. Drainage facilities
in other areas of the city are often inadequate or nonexistent and the
forested slopes in many areas were completely cleared of vegetation
without any quality control procedures.
In January of 1993, several major rainfall events generated severe
flooding that killed over 40 people and caused millions of dollars in urban
damages. During these flood events, the GIS was established to use as a
tool for damage assessments and urban planning efforts. The databases
were used to store information on topography, geology, drainage,
population density, housing characteristics, and income. Using the data
stored in the GIS, the damaged area was computed to be approximately
21


1,960 hectares. Approximately 57% of this area was damaged from erosion
and mass movements, while deposition of sediments caused damages to
approximately 20% of the study area. The damage in the remaining 23% of
the study area was attributed to standard channel and overbank flooding.
The majority of the damage was due to the acceleration of the erosion and
deposition processes caused by inadequate urban planning.
The results of the study indicated that different social groups suffered
various types and levels of damage. Low income neighborhoods were
affected by erosion while middle class areas experienced damages from
deposition and a lack of adequate drainage infrastructure. The
identification of the differences in flood damages allowed local officials to
assess different contingency strategies for each social group. The GIS
efforts led to the development of a detailed public works program in
Tiajuana. With this program, the effects of major flood events could be
analyzed much more effectively. It also gave the decision makers the
proper tools to formulate and design sensible and effective structural and
non-structural flood control strategies.
22


CHAPTER 3
THEORY AND OBJECTIVES
The purpose of this study is to perform a flood damage analysis for a
chosen stream reach using two different analysis methods and to compare
the flood-damage information generated by each method. The first method
used was the development of a flood damage model by traditional
techniques. The most significant variable in any flood damage analysis is
the finished floor elevation (FFE) of the structure affected by flooding. For a
typical flood damage reach, the FFE of each structure in the delineated
100-year or 500-year floodplain must be determined. In small damage
reaches, this data can be obtained from a detailed construction or property
survey. However, for extensive flooded areas, detailed field surveys can be
costly and time consuming. In such cases, the FFE data is calculated from
available topographic mapping. For most major urban areas, topographic
maps with a one-foot or two-foot contour interval have been developed
using advanced photogrammetric techniques. These maps also contain the
outlines of all structures present in the mapping area when the aerial
photographs used to generate the maps were obtained.
Using the available aerial mapping, the engineer or technician
locates the structure in question and computes the FFE by interpolating the
elevation between the major and minor map contours. Again, for extensive
areas of projected or historical flooding, this process can be extremely
tedious, costly, and time consuming and is subject to human error. The
23


FFE data for each structure is input into the flood damage model. The
model also includes data on the value of the structure and its contents, the
structure type, and the location of the structure relative to the stream
stationing used in the hydraulic model of the stream reach.
The second method uses a digital elevation model (DEM) to
determine the FFEs of the structures in the designated floodplain. Using
the spatial analysis and database capabilities of GIS software, the FFE data
was determined by simple point and click queries of the on-screen
graphical display and the associated database attributes. With this method,
FFE data can be generated quickly by the GIS in a fraction of the time
required by the traditional mapping or surveying methods. The results from
these two techniques were compared to determine if the model developed
from the GIS spatial analysis is as accurate as the model developed by
traditional methods.
As stated earlier, data development by hand for flood damage
models is often tedious, time consuming, expensive, and subject to human
error. The objective of this study is to show that data development and
model input using GIS techniques is significantly faster, more efficient, and
as accurate as the manual methods.
Flood Damage Analysis Modeling
The calculations of flood damages for the stream in this study are
performed using the Flood Damage Analysis Package (FDA), developed by
the U.S. Army Corps of Engineers, Hydrologic Engineering Center (HEC) in
24


Davis, California. The FDA system can be used to assess existing
floodplains and identify problem areas, to appraise damages in the
aftermath of flood events to allocate relief funds and other emergency
assistance, and to estimate the reductions in flood damages as a result of
channel improvement alternatives and development scenarios. For the past
several years, the Corps of Engineers has been actively developing a wide
variety of hydrologic and hydraulic computer software programs to meet the
information needs of engineers and scientists involved in water resource
analysis. These programs have been used extensively around the world as
stand alone models and in conjunction with other programs. Over the past
few years, sophisticated data management and programming software has
provided an opportunity to link HEC computer programs to form an efficient
data management and analysis system.
The FDA system of software links hydrologic, hydraulic, and
economic computer programs that have been previously developed by the
Corps of Engineers. The FDA Package is comprised of HEC-1 (Flood
Hydrograph Package), HEC-2 (Water Surface Profiles), HEC-5 (Simulation
of Flood Control and Conservation Systems), SID (Structure Inventory for
Damage Analysis), SIDEDT (SID Edit Program), DAMCAL (Damage Reach
Stage-Damage Calculation), EAD (Expected Annual Damage Computation),
and FDA2PO (HEC-2 Post-processor Program).
When used collectively, the programs give the user the capability to
perform a complete flood damage analysis for existing floodplain conditions
and for a wide range of structural and non-structural floodplain improvement
alternatives. The structural and non-structural measures previously shown
in Table 1.1 can be effectively analyzed through the linking and use of the
25


appropriate programs. The FDA system also has the capability to accept
input from hydrologic and hydraulic programs developed by other agencies
and institutions.
Structural Inventory of Damages
The first step in the FDA modeling process is the development of the
Structural Inventory of Damage model (SID). In order to develop the SID
model for a given stream reach, the structures within a designated
floodplain must be inventoried. The finished floor elevation for each
structure is determined along with the value of the structure and the value of
its contents. These variables are input into the SID model. To complete
the model input, the water surface elevations for given flood events at the
structure are also added. Generally, the water surface elevations are
determined using the Corps of Engineers, HEC-2 Water Surface Profiles
Program. The discharges for the HEC-2 model are determined by the HEC-
1 Flood Hydrograph Package and linked to the HEC-2 model by the FDA or
are determined by another hydrologic model and directly input into the
HEC-2 data file. The SID model processes the inventories of the structures
in the floodplain and computes the elevation-damage relationships for each
structure within the stream reach.
26


Expected Annual Damages
The final step is to execute the Expected Annual Damage (EAD)
model for the specific conditions and floodplain management alternatives
being analyzed. Expected annual damages can be computed using a long,
historic record of damage data and computing the average values. In many
cases, this data may not be available. As an alternative, the FDA system
manipulates the data in a manner that calculates the damage potential from
specific flood events and weights specific damage values under the
probability that these events may be exceeded. This is often referred to as
the frequency method. The result is the computed EAD value for the reach
of stream being analyzed.
The frequency method is based on the idea that flood damages to
property, individual structures, or groups of structures can be calculated by
determining the dollar values of the damage for various magnitude floods
and computing the percent chance of exceedance of each of these floods.
In order to compute the damage that would be expected in any single year,
the damages corresponding to each magnitude of flooding are weighted by
the probability of each event being exceeded. The sum of these weighted
damages is the EAD. The most common ways to express these
relationships are to relate the stream stage and/or flow to damage and to
relate damage to the exceedance frequency.
The output from the SID model, containing the frequency and
damage information in terms of present-day dollars, is input into the EAD
model. The EAD model uses the SID output, a discount rate, and an
estimate of the length of analysis (design life of floodplain improvements) to
27


compute the expected annual damages for the existing floodplain condition
and for proposed design alternatives. The reductions in expected annual
damages (benefits) resulting from the implementation of a specific
alternative are compared to the annual costs of that alternative. These
benefit-cost ratios are used to indicate whether the alternatives to improve
the channel and overbank conditions in a floodplain can be economically
justified.
Once floodplain alternatives are developed for a given stream reach,
the process is repeated and the benefit-cost ratios are determined for other
proposed alternatives. Using these ratios, several floodplain alternatives
can be compared to determine the most cost-effective plan. Generally, it is
common practice to divide a river into reaches for an EAD analysis because
the flow, frequency, and damage relationships vary along a river channel.
Dividing the channel into reaches allows different alternatives to be
analyzed for different portions of the stream in question. In this approach,
the engineer or floodplain manager can determine the best option for each
reach of the river instead of being limited to a single alternative for the
entire stream. One of the primary reasons for computing the flood damages
is to determine the effectiveness of various alternative plans in reducing
damages in the existing floodplain. In general, the damages for the existing
floodplain condition are computed to use as a baseline for the formulation of
alternatives. If the damages are significant, measures to reduce the
damages are developed. These mitigation plans will consist of one or more
damage reduction measures. Various plans will alter the flood damage
relationships in many different ways. The various stage-damage-frequency
relationships, developed by the FDA system, are summarized in Figure 3.1.
28


FIGURE 3.1 Derived
Damage Relationships
May 1996


Geographic Information Systems (GIS)
In the mid to late 1960s, the continued development of remote
sensing techniques, aerial photography, and other image processing
systems generated an explosion of geographic data production and the
need for more sophisticated analysis. Geographic data was being
generated faster than it could be analyzed. The spatial relationships of
small amounts of data could easily be retrieved and studied. However, the
technology to handle large volumes of data simply did not exist until the
1970s when digital computers began to be developed. It was during this
period of time that the computer based GIS system began to provide the
power necessary to process and analyze large volumes of data.
A Geographic Information System (GIS) is defined as an organized
collection of computer hardware, software, geographic data, and personnel
designed to efficiently capture, store, update, manipulate, analyze, and
display all forms of geographically referenced information (Lyon and
McCarthy, 1995). The term GIS encompasses the geographic analysis
tools that are available to the user. There are many related technologies
that can be used as stand alone systems or can be used within the context
of a GIS. These related technologies are shown in Figure 3.2. A GIS is
typically made up of several different components, depending on the
capabilities of various hardware and software combinations. The basic
structure of a GIS system is shown in Figure 3.3. The figure shows that the
nucleus of a GIS system is the spatial and attribute database where the
characteristics of an area of interest are stored in digital form. The
30


components of the GIS system access the information in the central
database to perform hundreds of geographical analysis functions.
GIS Purpose
The main purpose of a GIS is to analyze data and information by
integrating data layers and displaying and manipulating the data in different
ways and perspectives. It is the spatial analysis capabilities that distinguish
a GIS system from a computer aided drafting (CAD) system. In a sense, it
gives the decision maker the ability to view and analyze the data in a
manner that cannot be done using conventional computer modeling or
graphic (CAD) displays. The data can be assembled and used in a
completely different context than is done with other computer modeling
systems.
The lowest level of effort for a GIS would be to use the system to
inventory information. Simple data layers such as land cover types, roads,
and streams can still be analyzed spatially, providing valuable data analysis
for anyone interested in planning and management. The variety and
quantities of this data can also be summarized statistically. Although these
computations are considered simple, low level functions, the inherent ability
of a GIS to perform spatial analyses makes the manipulation of the low level
data a valuable tool for decision makers.
31


Database Management
Systems.....
Remote Sensing 1111 I
Systems
Automated Mapping
Systems
o IZZI
liiiflnnMfliilh
Facilities Management
Systems
FIGURE 3.2 Related Technologies
May 1996


GIS FACILITY
OTHER
DATABASES
i
COMPUTER
MODELS
\
USERS
FIGURE 3.3 GIS Structure
May 1996


Model Integration
At the highest level, a GIS system provides data and analysis to
support engineering and environmental models. The true power of a GIS
system is in the ability of the software to perform a spatial analysis and
provide input data, through model integration, to the models being used for
a particular area of study. In fact, the integration of GIS systems and
environmental and engineering models is becoming one of the most widely
used applications for GIS software. Many of the existing models that are
favored by engineers and scientists have little or no database access
capability and the results from the models have no spatially distributed
determinations. As GIS systems and engineering models become more
sophisticated, the ability of the software to exchange spatial data might
become a necessity for efficient use of the system.
Many of the current engineering models use coefficients for various
parameters. Another advantage of a GIS system is to use the remote
sensing technology within a GIS (if available) to calibrate the coefficients
and further refine the values based on actual measurements of their
characteristics. For example, the Universal Soil Loss Equation (USLE) has
several coefficients that need to be estimated for the soil loss computations.
By calibrating the coefficients with known values or measured field data, the
variability between the model simulation and the natural system will
decrease when the spatial analysis techniques within a GIS are utilized.
34


Statistical Analysis
Another useful application of a GIS system is the ability to perform a
statistical analysis on the variables in the database used to describe an
area of interest. A statistical approach studies the tendencies of the
variables by testing a hypothesis and developing relationships between
them. Depending on the quality of the statistical relationships, these
models can be applied to variables at different points in time, or can be
applied to other variables in the database. In other words, the model can
be used to predict what might happen to a variable in other places or times.
For example, statistical analyses have been effective for the study of
suspended sediments, water temperature variations, and the evaluation of
nonpoint pollution sources. Tests of the sensitivity of the variables is a
good strategy as well. It is advantageous to understand the contributions of
each of the individual variables to ensure that their contributions are
appropriate. Using the spatial database power of a GIS system, numerous
tests of variable sensitivities can be made quickly and economically.
Recently, GIS systems have been used to verify the results from
engineering model simulations. Many deterministic and statistical models
have no capability to input variables directly from GIS or Remote Sensing
systems. However, the variables for these models can often be evaluated
or even measured directly from GIS coverages. These variables can
actually be verified independently using GIS and Remote Sensing
technology. Several independent checks on the input data and output
results for the GIS and the engineering modeling can be used as an
effective validation of the model used.
35


Advances in Technology
Technological advances in hardware, software, and data acquisition
have completely changed the nature of planning and land management.
The migration from large computer hardware systems to desktop personal
computers and workstations has expanded the user capabilities within the
framework of a GIS system. Engineers, planners, and managers now have
the ability to view and use geographic data in ways that could not be done
before. Linking maps with databases and environmental models makes it
possible for decision makers to access graphical and tabular attributes
simultaneously. This process provides the tools to analyze and simulate
the effects of policy and management decisions. A single user has the
ability to quickly search, display, manipulate and model large quantities of
geographic data. However, the most integral parts of a GIS system are the
decision makers who collect, analyze, and manipulate the data. A GIS
system is destined for failure if an investment in an adequate and well
trained staff is not made.
Spatial Data Accuracy
The use of GIS technology has turned many costly and time
consuming spatial analyses into practical and economic tools. Before the
development of GIS systems, high spatial accuracy in geographic data was
not required, and errors in this data were generally ignored or considered
acceptable. As GIS systems and environmental models become more
36


sophisticated, the analyses performed will rely more heavily on site-specific
data from numerous possible sources. Spatial and attribute errors, when
combined collectively, could limit the value of model predictions. Before
any spatial analyses are performed, the accuracy of the data must be
addressed. The foundation of a GIS system is the spatial and tabular data
and major investments of time and money are spent on data development.
Data quality affects every aspect of the GIS system and the accuracy of the
data needs to be addressed if GIS users are to apply the tools in an
appropriate, effective, and accurate manner.
GIS Analysis
The spatial analysis capability of a GIS system is the feature that
differentiates this technology from other information systems or computer
aided drafting (CAD) systems. The functions that access the graphical and
tabular attributes in the GIS database are used to model and answer
questions about the real world. However, the more attributes that a model
takes into account means a more complex and expensive analysis. A
model that is more complicated may or may not provide better answers. In
this case, the GIS user needs to review the questions that need to be
answered.
Once the questions are defined, the appropriate data acquisition and
analysis methods can be ascertained. The best use of a model is in a
situation where it is impossible or too expensive to collect information
directly. The power of a GIS model becomes apparent because it can be
37


manipulated and tested more conveniently and at a much lower cost than
the actual conditions in the real world that the model simulates. Alternative
scenarios and repeated analyses can be continuously tested to answer
questions about existing conditions or to predict the effects of future
activities. The GIS can also be used to predict what might occur at a
different location or at another point in time. The general structure of a GIS
analysis is shown in Figure 3.4.
Although GIS technology becomes a powerful decision making tool, it
should never become a substitute for human understanding and
experience. The user must keep in mind the goals and values of the
organization that the GIS system is a part of. If human value judgments are
not an integral part of the GIS system, the analysis becomes a simple
exercise and not a valuable tool. As a worst case, the results of the
analysis may be deceptive. The GIS user cannot rely too heavily on the
computer hardware and software and too little on judgment and experience.
Legal Issues
Society continues to place more demands on the ability to analyze
geographic information. At the same time, world resources are becoming
more scarce. Subdivisions, power plants, and waste disposal facilities are
projects that are associated with population growth. The construction and
use of these facilities are being investigated by a wider variety of public and
private regulatory agencies and are also becoming a frequent target of
public opposition. GIS systems provide a powerful source of analysis
38


capabilities for investigations related to these types of facilities. The
flexibility of analysis provides information that experts from diverse areas of
study can use to solve difficult problems.
GIS systems have developed much more rapidly than the institutions
they are designed to serve. Although the technological changes have taken
place, many of the managerial, legal, and social issues associated with the
implementation of GIS systems have only begun to be understood.
Political and institutional issues are often the major obstacles to the
introduction of a GIS system. It can dramatically change the ways that data
is viewed and used. Data ownership can also become a restraint to the
introduction of GIS technology. Departments within an organization often
become possessive of the data they have developed despite the use of the
data by the entire corporation. It is sometimes difficult to perceive the GIS
system as a resource for the entire organization.
One of the most serious issues accompanying the use of GIS
technology is the legal liability of spatial analysis and the distribution of
geographic data. Anyone given access to a GIS system has a considerable
amount of information available to them. Although it is a powerful analysis
tool, it could also be used to abuse and invade the privacy of individuals,
companies, and other organizations. Legal and ethical guidelines, analysis
standards, and quality control measures, all exercising authority over the
use of GIS technology, need to be continuously updated and enforced. The
use of GIS systems have the potential to compromise individual rights.
There have already been difficult trade-offs associated with the protection of
39


WHAT IS THE DATA?
Source: Aronoff (1991)
WHAT COULD THE DATA BE
IN ANOTHER PLACE, TIME,
OR UNDER DIFFERENT
CONDITIONS?
QUESTIONS
WHAT ARE THE PATTERNS
IN THE DATA?
PRESENTATION OF DATA
PREDICTING NEW
INFORMATION
STORAGE AND RETRIEVAL
FINDING PATTERNS IN THE DATA
CONSTRAINED QUERY
FIGURE 3.4 GIS Analysis Categories
May 1996


private individuals while dealing with the legal rights of agencies to access
informational databases. Data standards have been introduced and
accepted to protect data vendors and provide data users with an acceptable
source of information.
Political Issues
As GIS technology becomes an integral part of an organization, the
control of the GIS database can give bureaucrats, decision makers,
technical experts, and computer literate individuals more power at the
expense of people or organizations who do not have access to the data.
The data becomes more authoritative and politically neutral language
describing the data collection, analysis methods, and results, often creates
a sense of impartiality. Computer modeling and analysis requires a choice
of modeling techniques, data acquisition, and interpretation of results.
Many of these choices can be politically motivated because they will directly
affect the results that will be obtained from the analysis.
Computer based analyses can confuse and clarify and political
choices can be hidden within the complexity of the GIS technology. Policy
choices and assumptions need to be explicitly stated and study procedures
fully documented. In addition, the results need to be made available for
public and private review and clearly stated for those who will use the
information. GIS systems have improved the speed and accuracy with
which geographic data can be gathered, manipulated, edited, and analyzed.
The technology is also being continuously updated and used in new ways.
41


A successful GIS requires qualified users as well as managers and other
decision makers who understand the benefits of GIS technology and the
issues involved in GIS implementation. The final result of a GIS must not
jeopardize values and objectives. Instead, it should enhance them.
42


CHAPTER 4
CASE STUDY
The stream chosen for this study is a portion of Clear Creek, in
Wheat Ridge, Colorado. The headwaters of Clear Creek follow the western
boundary of Clear Creek County from Grays Peak and Loveland Pass to Mt.
Nystrom and Berthoud Pass, approximately 40 miles west of the study area.
The Clear Creek drainage basin is shown in Figure 4.1. For this study, a
portion of Clear Creek located between Interstate 70 and Kipling Avenue,
immediately west of Denver, was chosen for the detailed flood damage
analysis. In this reach of Clear Creek, the 100-year floodplain inundates 88
structures in the left and right floodplain overbanks (Denver UDFCD, 1977).
The study area is shown in Figure 4.2. The detailed 100-year floodplain
delineation and the locations of the structures in the floodplain are shown in
Figure 4.3.
Clear Creek Hydrology and Hydraulics
The Clear Creek Flood Hazard Area Delineation (FHAD), completed
in November 1979 (Denver UDFCD), contains the discharge and floodplain
information for the flood damage analysis. The FHAD report also shows the
locations of the structures in the 100-year floodplain. In addition to the
report, the HEC-2 Water Surface Profiles model, used for the FHAD study,
was obtained from the Denver Urban Drainage and Flood Control District.
43


Watershed Boundary
FIGURE 4.1 Clear Creek Watershed || May 1996


4^
QUADRANGLE LOCATION
Study Area Limit
FIGURE 4.2 Site Location Map
May 1996


FIGURE 4.3 100-Year Floodplain/Structures
May 1996


The 100-year flood is the emphasis of study in the Clear Creek
FHAD and is the only floodplain delineated in the report. In order to
perform a flood damage analysis of the study area, a full range of
discharges is required. For a complete flood damage analysis, the 2-year,
5-year, 10-year, 25-year, 50-year, 100-year, and 500-year discharges are
required. Although the 100-year flood is emphasized in the FHAD, the 10-
year, 50-year, and 500-year discharges, computed by the Omaha District of
the U.S. Army Corps of Engineers (COE), were also published in the report.
These discharges were plotted on log-probability paper and were used to
interpolate the 25-year discharge and extrapolate the 2-year and 5-year
discharges. The computed discharges are shown in Figure 4.4. The final
discharges were input into the HEC-2 model of Clear Creek in order to
compute the water surface elevations for the range of frequencies listed
above.
Standard Flood Damage Modeling
The first step in performing a flood damage analysis is to complete
an inventory of the structures located within the flood prone areas. As
Figure 4.3 shows, there are 84 structures in the left overbank and 4
structures in the right overbank of Clear Creek between Wadsworth
Boulevard and West 44th Avenue. The approximate dollar value of each
structure, the finished floor elevation (FFE), and the location of the
structure, relative to the stationing used in the HEC-2 model of the study
area, are the required input variables for the SID (Structure Inventory for
47


Damages) model. A flood damage model, using FFE data determined
manually from aerial topographic maps, was constructed and used as a
baseline condition.
In most cases, the values of the structures in the floodplain can be
obtained from real estate professionals. For this study, a dollar value of
$100,000 was assigned to each structure. In practice, the value of the
contents of the structure are generally considered to be 50% of the value of
the structure, unless other data is available. The finished floor elevations
for each structure were originally determined from the topographic mapping
used for the floodplain delineation in the FHAD report. The FFE data was
also taken from aerial topographic mapping developed in 1985 by the City
of Wheat Ridge. The contour interval for the mapping is 2 feet, and each
finished floor elevation was computed based on the location of the structure
relative to the contours. The finished floor elevations for the flood prone
structures in the study area are shown in Table 4.1. The finished floor
elevations of the structures are the most difficult variables to compute for a
flood damage analysis. In a large scale study, the determination of these
variables can be tedious, time consuming, and subject to human error.
The structure data and HEC-2 water surface elevations are input into
the SID model. Another major portion of the SID input is the depth and
percent damage data for the structures and their contents. This data
provides the model with information on the percentage of damage to the
total value of the structure as the water level increases. The U.S. Army
Corps of Engineers has developed a standard set of depth and percent
damage information for several land uses based on research and historic
flood damage data. The COE depth and percent damage functions for
single family residential structures and contents are shown in Tables 4.2
48


Clear Creek FHAD Frequency Discharges
FIGURE 4.4 Frequency-Discharge Curve
May 1996


and 4.3, respectively.
The SID program processes the inventory of the structures and
incorporates the results of the HEC-2 model to develop the elevation-
damage functions for the flooded structures. The output from the SID model
contains the stage-damage data for individual structures and summarizes
the total flood damages by structure category and stream reach. The output
also contains a detailed listing of the number of structures in the floodplains
for each discharge frequency analyzed as well as the total dollar value of
the damages in the study area. The total damages for each incremental
stage increase and the damages for each return period are also calculated
by the SID model.
50


TABLE 4.1 Structure Finished Floor Elevations and Corresponding HEC-2 Station Location
STRUCTURE FFE City Topo Mapping (feet) HEC-2 Station
001 454+00
002 454+30
003 5314.3 457+50
004 5316.5 458+70
005 5325.0 477+90
006 5324.0 478+60
007 5324.3 479+20
008 5324.3 479+70
009 5324.4 480+10
010 5324.4 480+60
011 5324.6 481+80
012 5325.4 482+70
013 5326.3 483+30
014 5326.6 483+80
015 5327.1 484+40
016 5327.1 485+00
017 5328.0 485+50
018 5328.3 486+20
019 5328.5 486+70
020 5328.7 487+50
021 5328.9 488+10
022 5329.0 489+10
023 5329.3 489+80
024 5329.5 490+50
025 5329.5 491+10
026 5329.5 491+70
027 5329.5 492+20
028 5329.5 493+00
029 5330.3 493+80
030 5332.6 493+50
51


*n


m
o


TABLE 4.2
U.S. Army Corps of Engineers
Siege vs. Damage Curves
Single Family Residential (Structures)
Stage (feet) % Damage
-0.1 0
0.0 10
1.0 21
2.0 27
3.0 32
4.0 37
5.0 43
6.0 46
7.0 50
8.0 54
TABLE 4.3 U.S Army Corps of Engineers Stage vs. Damage Curves Single Family Residential (Contents)
Stage (feet) % Damage
-0.1 0
0.0 8
1.0 42
2.0 60
3.0 71
4.0 77
5.0 82
6.0 85
7.0 86
8.0 87
54


Flood Damage Modeling Using GIS
In order to develop the input data using GIS techniques, a Digital
Elevation Model (DEM) of the study area was required. A DEM for the 7.5
minute Arvada, Colorado Quadrangle (USGS, 1994), at a scale of 1:24,000,
was obtained and input into the GIS system. The structure locations and
the 100-year floodplain limits were digitized from the FHAD maps and also
input into the GIS database. The DEM and the digitized structure and
floodplain information were registered to the Universal Transverse Mercator
(UTM) Coordinate System. A plan view of the DEM obtained from the
USGS is shown in Figure 4.5. A three dimensional view of the DEM data,
showing the topographic detail in the study area, is shown in Figure 4.6.
After the DEM, floodplain, and structure information were input into the GIS,
the structure and floodplain data was overlaid on top of the DEM. From this
point, a simple on-screen database query was performed to determine the
finished floor elevations of each of the floodplain structures.
The ARC/INFO GIS software, developed by the Environmental
Systems Research Institute (ESRI), was used for the modeling in this study.
ARC/INFO is a vector-based GIS system and consists of a sophisticated
combination of geographic data processing modules designed to capture,
edit, manage, manipulate, and display georeferenced data. It provides the
tools that allow the user to access, visualize, and query spatial and tabular
attributes for enhanced analysis capabilities and decision making.
55


FIGURE 4.5 DEM/GRID Plan View
May 1996




GRID Modeling
The GRID module in ARC/INFO was the first GIS technique used in
this study. It is a combined raster-based (cell) and relational attribute
geoprocessing tool that can be used for simple and complex grid-cell
analyses. The GRID module is the most suitable for applications that are
based on a locational view of the world. It is also well suited for
representing and analyzing continuous data, overlay analysis, and distance
computations. It divides data into uniform cells with each cell representing
an actual portion of geographic space. Each cell contains information
relating to the theme of the GRID coverage. In the case of a DEM, each cell
contains the average elevation of the ground surface represented. The raw
DEM data is input into the GRID module. It then processes the raw data
into the cell-based format with each cell representing a square area, 30
meters (98.4 feet) on each side.
TIN Modeling
The TIN (Triangulated Irregular Network) module in ARC/INFO was
also used for the GIS analysis in this study. The TIN concept is a data
structure that allows for the efficient generation of surface models for the
analysis and display of topography as well as other types of surfaces. It is
designed to capture and compile digital elevation data and develop a
surface model that can be used for applications such as slope and aspect
calculations, contouring, watershed modeling, and three-dimensional
58


display. The model stores topographic surfaces as topological networks of
irregularly shaped triangles.
A TIN model of the study area was also developed from the DEM
data obtained for the earlier GRID modeling and analysis. The DEM data
structure consists of data points at an elevation increment of one meter
(3.28 feet). The finished floor elevations of the structures in the floodplain,
when computed from GRID data, have a potential elevation difference of +\-
0.5 meters (1.65 feet). The contouring capability in the TIN data model
interpolates the elevations between the one meter increments of the original
DEM data.
The most difficult data development for the SID module of the FDA
package involves the determination of the finished floor elevation for each
of the structures in the floodplain. The finished floor elevations, determined
from the standard methods and the GRID and TIN modules, were input into
the flood damage models. In this analysis, it was assumed that the finished
floor elevation is approximately equal to the corresponding grid cell
elevation or interpolated TIN surface at each structure location
59


CHAPTER 5
RESULTS
Two surface modeling techniques within the context of a GIS spatial
analysis were used to develop the input data for a flood damage model of
the study area. A 1:24,000 scale DEM of the Arvada, Colorado 7.5 minute
quadrangle, was input into ARC/INFO and converted into a GRID model
and into a TIN model. The 100-year floodplain boundary, structure
locations, and major streets in the study area were digitized using AutoCAD
software and translated to GIS coverages using standard ARC/INFO
conversion commands. The digitized structure and floodplain coverage was
registered to the UTM Coordinate System and overlaid onto the ARC/INFO
GRID and TIN models developed from the DEM.
The databases for both of the surface models were accessed using a
graphical on-screen query at the approximate center point of each structure.
This process was repeated for all of the structures within the floodplain
limits. The finished floor elevations computed from each GIS method, as
well as other structure variables, were input into the models comprising the
Flood Damage Analysis Package (U.S. Army Corps of Engineers, 1988). In
addition to the GIS techniques, a baseline model was constructed using
finished floor data determined from the most recent aerial topographic maps
of the study area (City of Wheat Ridge, Colorado, 1985). The finished floor
elevations were calculated by visually interpolating between the elevation
60


contours surrounding a given structure. This is the common method that is
used to develop input data for flood damage models.
In engineering practice, the determination of the finished floor
elevations from aerial mapping is laborious and time consuming, especially
for large scale study areas. The objective of this investigation is to
determine the suitability of GIS surface models for developing input data for
flood damage models. This study includes a comparison of the finished
floor elevations and the results of flood damage modeling for existing
channel conditions using data developed by the manual and GIS methods.
A comparison of the finished floor elevations determined by the manual
topographic map interpolation and the GRID surface model is shown in
Table 5.1.
GRID Modeling
Using these GIS spatial analysis techniques, the finished floor
elevations for each structure can be determined in a fraction of the time that
is required to manually interpolate the elevations from available mapping.
However, the finished floor elevations and subsequent flood damage
analysis using the GRID surface model yields poor results when compared
to the baseline conditions developed by standard methods. A summary of
the computed damages for each frequency using the standard methods and
the GRID data is shown in Table 5.2. The total number of flooded
structures computed by each method is shown in Table 5.3.
61


TABLED Structure Finished Floor Elevations Standard Methods and <3lS GRID Modeling
STRUCTURE FFE City Topo Mapping (feet) FFE ARC/INFO GRID Model (feet) Absolute Value of Elevation Difference (feet)
001
002
003 5314.3 5316.9 2.6
004 5316.5 5320.2 3.7
005 5325.0 5323.4 1.6
006 5324.0 5323.4 0.6
007 5324.3 5323.4 0.9
008 5324.3 5323.4 0.9
009 5324.4 5326.7 2.3
010 5324.4 5326.7 2.3
011 5324.6 5326.7 2.1
012 5325.4 5330.0 4.6
013 5326.3 5330.0 3.7
014 5326.6 5323.4 3.2
015 5327.1 5326.7 0.4
016 5327.1 5323.4 3.7
017 5328.0 5323.4 4.6
018 5328.3 5326.7 1.6
019 5328.5 5326.7 1.8
020 5328.7 5323.4 5.3
021 5328.9 5330.0 1.1
022 5329.0 5323.4 5.6
023 5329.3 5323.4 5.9
024 5329.5 5326.7 2.8
025 5329.5 5326.7 2.8
026 5329.5 5326.7 2.8
027 5329.5 5333.3 3.8
028 5329.5 5330.0 0.5
029 5330.3 5333.3 3.0
030 5332.6 5333.3 0.7
62


VO


TABLE 5*t (Continued) Structure Brushed Boor Elevations Standard Methods and BIS GRID Modeling
STRUCTURE FFE City Topo FFE-ARC/INFO Absolute Value
Mapping GRID Model of Elevation
(feet) (feet) Difference (feet)
061 5327.7 5323.4 4.3
062 5328.4 5326.7 1.7
063 5328.7 5323.4 5.3
064 5329.0 5323.4 5.6
065 5329.3 5323.4 5.9
066 5329.5 5326.7 2.8
067 5329.7 5326.7 3.0
068 5330.2 5326.7 3.5
069 5330.2 5326.7 3.5
070 5330.2 5326.7 3.5
071 5330.2 5326.7 3.5
072 5330.1 5326.7 3.4
073 5330.1 5326.7 3.4
074 5330.0 5326.7 3.3
075 5329.7 5326.7 3.0
076 5328.5 5326.7 1.8
077 5329.7 5326.7 3.0
078 5330.5 5326.7 3.8
079 5330.5 5326.7 3.8
080 5330.5 5326.7 3.8
081 5330.5 5330.0 0.5
082 5330.5 5336.6 6.1
083 5330.5 5336.6 6.1
084 5333.2 5333.3 0.1
085 5332.2 5330.0 2.2
086 5332.2 5333.3 1.1
087 5332.2 5333.3 1.1
088 5333.4 5333.3 0.1
089 5333.0 5336.6 3.6
090 5328.8 5339.8 11.0
64


Flood Damage Computations Standard Methods and GIS GRID Analysis
Frequency Structural Damages Standard Methods Structural Damages GIS GRID Analysis
0-2 Year $0 $0
2-5 Year $0 $900,000
5-10 Year $0 $750,000
10-25 Year $450,000 $3,900,000
25-50 Year $1,650,000 $1,650,000
50-100 Year $6,900,000 $2,250,000
100-500 Year $4,200,000 $2,850,000
>500 Year $0 $900,000
Total $13,200,000 $13,200,000
Table 5.3
Number of Flooded Structures
Standard Methods and GIS GRID Analysis
Frequency Flooded Structures Standard Methods Flooded Structures GIS GRID Analysis
0-2 Year 0 0
2-5 Year 0 6
5-10 Year 0 5
10-25 Year 3 26
25-50 Year 11 11
50-100 Year 46 15
100-500 Year 28 19
>500 Year 0 6
Total 88 88
65


A comparison of the tables shows a wide variation in the finished
floor elevations and the computed flood damages for the GRID surface
model and the standard method analysis. The GRID model results are
generally higher for the flood events less than a 25 year frequency and
lower for any flood frequencies greater than the 25 year event. The
elevation differences between the two modeling techniques (GRID and
manual method) appear to be randomly scattered with no significant
patterns in the elevation differences compared to the location of the flooded
structures.
Table 5.1 shows that the GRID model predicted only 17% of the
finished floor elevations within 0.3 meters (1.0 feet) of the data determined
by the manual methods. The GRID model predicted 41% of the structures
within 0.6 meters (2.0 feet) of the finished floor elevations interpolated from
the aerial topographic maps.
Overall, the GRID model did not perform well in predicting the
finished floor elevations of the structures. In fact, as Table 5.1 shows, the
elevation differences in some cases were as high as 1.8 to 3.4 meters (6 to
11 feet). These differences are reflected in Tables 5.2 and 5.3, showing the
computed flood damages and the number of flooded structures between the
input data developed by the GRID model and the baseline model. With the
exception of the 25-50 year flood range, the damages and number of
structures using the data from the GRID model were either well above or
well below the baseline estimates.
66


TIN Modeling
Using the TIN analysis techniques to determine the finished floor
elevations for each structure yielded somewhat better results than the GRID
model. However, a comparison of the tables for the TIN modeling and the
standard methods shows another wide variation in the finished floor
elevations and the computed flood damages. The finished floor elevations
determined by the TIN model are shown in Table 5.4. The damages
computed for each frequency using the standard methods and the TIN data
are shown in Table 5.5. The total number of flooded structures computed
by each method is shown in Table 5.6. The TIN model finished floor
elevations and flood damage estimates were higher than the existing
conditions for frequencies less than the 50-year event, and lower for return
periods greater than or equal to the 100-year event. The elevation
differences between the two modeling techniques (TIN and manual method)
again appear to be randomly scattered with no significant trends in the
elevation differences when compared to the location of the flooded
structures.
Table 5.4 shows that the TIN model predicted approximately 32% of
the finished floor elevations within 0.3 meters (1.0 feet) of the data
computed by the manual methods. The TIN model was also able to predict
approximately 57% of the structures within 0.6 meters (2.0 feet) of the
finished floor elevations interpolated from the City of Wheat Ridge aerial
mapping.
Generally, the TIN model performed better than the GRID model but
still results in a wide variation in the comparison of the finished floor
67


elevations, damages, and number of structures in the floodplain. Similar to
the GRID data, some of the elevation differences range from 1.8 to 3.4
meters (6 to 11 feet). These differences are evident in Tables 5.5 and 5.6,
respectively. Although the TIN model development shows better results,
the damages and number of structures predicted from the TIN surface
model data are still generally above or below the baseline estimates.
68


TABLE 5.4 Structure Finished Boor Elevations Standard Methods and BIS TIM Modeling
STRUCTURE FFE City Topo Mapping (feet) FFE-ARC/INFO TIN Model (feet) Absolute Value of Elevation Difference (feet)
001
002
003 5314.3 5317.2 2.9
004 5316.5 5318.5 2.0
005 5325.0 5321.1 3.9
006 5324.0 5323.4 0.6
007 5324.3 5323.4 0.9
008 5324.3 5324.8 0.5
009 5324.4 5326.4 2.0
010 5324.4 5324.1 0.3
011 5324.6 5325.4 0.8
012 5325.4 5328.0 2.6
013 5326.3 5328.0 1.7
014 5326.6 5325.1 1.5
015 5327.1 5324.8 2.3
016 5327.1 5324.8 2.3
017 5328.0 5325.4 2.6
018 5328.3 5326.4 1.9
019 5328.5 5326.1 2.4
020 5328.7 5325.7 3.0
021 5328.9 5325.7 3.2
022 5329.0 5322.5 6.5
023 5329.3 5323.8 5.5
024 5329.5 5326.4 3.1
025 5329.5 5328.0 1.5
026 5329.5 5328.7 0.8
027 5329.5 5329.7 0.2
028 5329.5 5331.3 1.8
029 5330.3 5331.3 1.0
030 5332.6 5331.6 1.0
69


TABLE 5,4 (Continued) Structure finished Floor Elevations Standard Methods and GtS TIN Modeling
STRUCTURE FFE City Topo Mapping (feet) FFE ARC/INFO TIN Model (feet) Absolute Value of Elevation Difference (feet)
031 5332.6 5332.0 0.6
032 5333.0 5333.6 0.6
033 5326.5 5325.4 1.1
034 5326.2 5323.8 2.4
035 5326.2 5325.7 0.5
036 5326.2 5326.7 0.5
037 5326.5 5326.7 0.2
038 5326.5 5326.7 0.2
039 5326.5 5326.7 0.2
040 5328.2 5325.4 2.8
041 5328.2 5326.1 2.1
042 5328.2 5326.1 2.1
043 5328.2 5326.7 1.5
044 5328.0 5326.7 1.3
045 5328.0 5323.4 4.6
046 5328.2 5323.4 4.8
047 5328.2 5324.8 3.4
048 5328.2 5326.1 2.1
049 5328.2 5326.7 1.5
050 5328.2 5326.7 1.5
051 5328.3 5326.7 1.6
052 5328.4 5327.0 1.4
053 5328.6 5329.0 0.4
054 5328.2 5327.4 0.8
055 5328.2 5326.4 1.8
056 5328.2 5327.4 0.8
057 5328.2 5327.4 0.8
058 5328.2 5326.7 1.5
059 5328.0 5325.1 2.9
060 5328.0 5323.8 4.2
70


TABLE 5.4 (Continued) Structure Bmshed Floor Elevations Standard Methods and GIS UN Modeling
STRUCTURE FFE City Topo FFE -ARC/INFO Absolute Value
Mapping TIN Model of Elevation
(feet) (feet) Difference (feet)
061 5327.7 5323.4 4.3
062 5328.4 5326.7 1.7
063 5328.7 5322.8 5.9
064 5329.0 5321.5 7.5
065 5329.3 5323.4 5.9
066 5329.5 5325.7 3.8
067 5329.7 5326.7 3.0
068 5330.2 5326.7 3.5
069 5330.2 5326.7 3.5
070 5330.2 5328.7 1.5
071 5330.2 5328.0 2.2
072 5330.1 5328.0 2.1
073 5330.1 5327.4 2.7
074 5330.0 5327.0 3.0
075 5329.7 5326.7 3.0
076 5328.5 5326.7 1.8
077 5329.7 5327.7 2.0
078 5330.5 5327.7 2.8
079 5330.5 5328.7 1.8
080 5330.5 5329.7 0.8
081 5330.5 5330.7 0.2
082 5330.5 5331.0 0.5
083 5330.5 5334.9 4.4
084 5333.2 5333.0 0.2
085 5332.2 5331.3 0.9
086 5332.2 5332.3 0.1
087 5332.2 5332.3 0.1
088 5333.4 5332.3 1.1
089 5333.0 5334.6 1.6
090 5328.8 5339.8 11.0
71


Table 5.5 Flood Damage Computations Standard Methods and CIS TIN Analysis
Frequency Structural Damages Standard Methods Structural Damages GIS TIN Analysis
0-2 Year $0 $150,000
2-5 Year $0 $450,000
5-10 Year $0 $450,000
10-25 Year $450,000 $2,250,000
25-50 Year $1,650,000 $3,300,000
50-100 Year $6,900,000 $4,650,000
100-500 Year $4,200,000 $1,650,000
>500 Year $0 $300,000
Total $13,200,000 $13,200,000
Table 5.5 Number of Flooded Structures Standard Methods and GIS TIN Analysis
Frequency Flooded Structures Standard Methods Flooded Structures GIS TIN Analysis
0-2 Year 0 1
2-5 Year 0 3
5-10 Year 0 3
10-25 Year 3 15
25-50 Year 11 22
50-100 Year 46 31
100-500 Year 28 11
>500 Year 0 2
Total 88 88
72


CHAPTER 6
CONCLUSIONS AND DISCUSSION
Ultimately, the final task in any major flood damage study is to
analyze the Expected Annual Damages (EAD) for the existing conditions of
the channel and floodplain and the proposed floodplain management
alternatives. The frequency method of EAD was described earlier in this
report. It is based on the idea that flood damages are calculated by
determining the monetary values of damages for various magnitude floods
and computing the percent chance of exceedance of each of these floods.
The damages are weighted by the probability of the flood event being
exceeded and are added together to obtain the EAD, defined as the
damages expected in any single year. The most common ways to express
these relationships are to relate the stream stage and/or flow to damage
and to relate damage to the exceedance frequency (Figure 1.1).
Benefit-cost ratios are computed for floodplain alternatives using
data on the reductions in expected annual damages (benefits) and the
annual costs of the alternative. These benefit-cost ratios are used to
indicate whether the alternatives to improve the channel and overbank
conditions in a floodplain can be economically justified. This process is
repeated for all of the proposed floodplain improvements in a study area.
An effective EAD analysis requires an accurate assessment of the
finished floor elevations of the floodplain structures. The stream stage,
flood damage, and exceedance frequency variables are the nucleus of any
73


EAD investigation of floodplain improvement alternatives. The HEC-2
hydraulic model of Clear Creek shows that a two or three foot variation of
the water surface elevation in the stream channel can make a considerable
difference in the classification of the exceedance frequency of the flood.
For example, an increase of three feet in the water surface elevation can
mean the difference between a 10-year flood event and a 50-year flood
event.
GRID Modeling
The results from the previous chapter show that the GRID surface
modeling techniques do not accurately predict the finished floor elevations
needed for detailed flood damage modeling. Since many of the water
surface elevation differences between the GRID surface model and the
baseline conditions are in excess of 0.6 meters (2 feet), it would be difficult
to pinpoint the level of flood protection that an improvement alternative is
intended to provide. The variability of the finished floor elevations
estimated by the GRID surface introduces an error into the flood damage
estimates computed by the SID model. The output from the SID model is
used as input for the EAD analysis and as a result, the elevation differences
are inherited by the EAD process.
The inability of the GRID model to accurately predict the finished
floor elevations is inherent in the very data structure of the original DEM. A
DEM consists of an array of elevations stored as profiles that are spaced 30
meters (98.4 feet) apart. The elevations in each profile are also spaced 30
74


meters apart and are defined in 1 meter (3.3 feet) elevation intervals. The
GRID model is comprised of a horizontal and vertical array of grid elements,
or cells. Each grid cell contains specific information pertaining to the field
of study. In the case of the DEM, the database contains the elevation in
meters for each grid cell. When the DEM is converted into an ARC/INFO
GRID coverage, the grid cells are sized according to the characteristics of
the data input. For the Arvada, Colorado DEM, the grid cells are 30 meters
long on each side.
As described earlier in this report, the outlines of the structures in the
floodplain were digitized and converted into an ARC/INFO coverage. The
structures were superimposed onto the GRID surface model and a simple
on-screen query of the database was performed to determine the finished
floor elevations at each structure. This process presents a severe limitation
in the calculation of the structure elevation. The elevation interval for the
DEM and the GRID is 1 meter (3.3 feet). With this increment, the actual
elevation range in the grid cell is +/- 0.5 meters (1.65 feet) from the
assigned elevation of the cell. This implies that the actual finished floor
elevation for a structure in the floodplain exists within this cell elevation
range of +/- 0.5 meters. However, each on-screen query of the GRID
database accesses a single grid cell with an elevation that is rounded off to
the nearest one meter elevation increment.
The average difference in the finished floor elevations between the
data developed by the standard methods and the GRID was computed to be
approximately 0.9 meters (2.8 feet). Calculations outlined in the previous
chapter show that the GRID model predicted only 17% of the finished floor
elevations within 0.3 meters (1.0 feet) of the baseline conditions. As Table
75


5.1 shows, there are 19 structures in the floodplain (22% of total structures)
that have an elevation difference of 1.2 meters (4.0 feet) or more when the
GRID model is compared to the baseline conditions. If these structures are
considered to be outliers and are removed from the analysis, the average
difference in the finished floor elevations between the GRID model and the
baseline model decreases to 0.6 meters (2.0 feet).
The designed structure of a DEM is the primary reason for the wide
variation in the computation of the finished floor elevations. Surface
modeling using GRID analysis techniques is very effective in such areas as
optimum habitat location, zone and proximity analyses, general terrain
evaluation, noise modeling, and watershed studies. A successful flood
damage analysis requires the accurate computation of finished floor
elevations. This data is used extensively to develop the other functions
shown in Figure 1.1. Because of the wide variation in the computation of
the finished floor elevations, the GRID surface model was not an effective
tool to use for a detailed flood damage analysis.
However, the GRID model was successful in predicting the total flood
damages for the Clear Creek study area. As Table 5.2 shows, the
computed flood damages using both methods were estimated to be
$13,200,000. The GRID model was also effective in predicting the number
of flooded structures within general ranges of frequency floods. For
example, Table 5.3 shows that all 88 of the structures in the study area are
above the 10-year floodplain boundary when the models developed by
standard methods are used. The results from the GRID analysis show that
approximately 77 structures are above the 10-year floodplain. Although the
difference in the estimates is approximately 12.5%, the results suggest that
76


the GRID surface model can be used to predict general ranges of flood
damages and flooded structures.
TIN Modeling
The results show that the TIN surface modeling techniques predict
the finished floor elevations somewhat better than the GRID analysis.
However, the results indicate that the TIN model used in this study does not
predict the finished floor elevations within the accuracy range needed for a
detailed flood damage analysis. Like the GRID model, many of the water
surface elevation differences between the TIN surface and the baseline
conditions are in excess of 2 feet. Again, this makes the analysis of
channel or floodplain improvements in an EAD investigation difficult
because the data does not allow an accurate return period to be assigned
to the damage functions.
The structure of the TIN model also causes a variability in the
computation of the finished floor elevations. The TIN model consists of a
set of adjoining triangles that do not overlap. The triangles are computed
from the DEM data points with x and y coordinates and elevation (z) values.
The TIN model stores the topological relationship between the triangles.
This structure allows the generation of surface models for the analysis and
display of terrain and other surface types. The TIN model consists of a
continuous surface where features can be represented without the sharp or
abrupt changes that are characteristic of the GRID model. The x, y, and z
values of the input DEM are joined by a series of lines to form a mosaic of
77


triangles. The advantage of the TIN data structure is that the elevations
between adjacent points can be interpolated due to the triangular
composition of the surface. A graphical comparison of the GRID and TIN
data structures is shown in Figure 6.1.
The TIN model used in this study was generated from the DEM
obtained for the Arvada, Colorado 7.5 minute quadrangle. As previously
discussed, the DEM consists of an array of elevations that are spaced 30
meters (98.4 feet) apart. The elevations in the DEM consist of 1 meter
(3.28 feet) increments. The TIN model in this study utilized a standard
linear interpolation between the regularly spaced data input points from the
DEM with no additional adjustments made to smooth the surface or change
the interpolation methods between the data points.
The database for the TIN model was also accessed using the on-
screen query process. A comparison of the finished floor elevations
computed by the baseline methods and the TIN model is shown in Table
5.4. The average difference between the data estimates was computed to
be approximately 0.7 meters (2.2 feet), 0.2 meters (0.6 feet) lower than the
average difference between the baseline model and the GRID surface.
Previous calculations determined that the TIN model predicted the
finished floor elevations within 0.3 meters (1 feet) of the baseline conditions
for approximately 32% of the structures in the floodplain. The model
predicted the elevations within 0.6 meters (2 feet) of the baseline conditions
for 57% of the structures. Although these results are considerably better
than the predictions obtained from the GRID model, the linearly interpolated
TIN surface was also not effective in predicting the finished floor elevations.
Table 5.4 shows that a total of 11 structures in the floodplain (12.5% of total
structures) have an elevation difference of 1.2 meters (4.0 feet) or greater
78


GRID Structure
D
TIN Structure
Source: ESRI (1994)
FIGURE 6.1 GRID and TIN Data Structures
May 1996


when the TIN model is compared to the baseline conditions. If these
structures are classified as 'outliers and are removed from the results
shown in Table 5.4, the average difference in the finished floor elevations
between the TIN model and the baseline conditions decreases to 0.5
meters (1.65 feet).
As in the GRID analysis, the structure of the TIN model is the primary
reason for the variation in the computation of the finished floor elevations.
A TIN model can be used to generate elevation contours, calculate
watershed slope, aspect, and flow lengths, and perform cut and fill
analyses. One of the more powerful features of a TIN model is the ability to
perform sophisticated visual analysis and surface interpretation. However,
for a detailed flood damage and EAD analysis, the TIN model used in this
study could not be considered an effective tool to use for the development
of detailed data.
The linear interpolation between data points in the TIN model for this
study does not take into account the local variations in topography that
could occur within the 30 meter (98.4 feet) spacing of the DEM data points.
The TIN data is used in the processes outlined in Figure 1.1. With the wide
elevation variations and the difficulty in determining the actual elevation-
damage-frequency relationships, the TIN data generated for this study
could not be used for a detailed flood damage analysis.
However, the TIN model was also successful in calculating the total
flood damages for the Clear Creek study area. As Table 5.5 shows, the
computed flood damages using the baseline and TIN modeling techniques
were estimated to be $13,200,000. The analysis from the TIN modeling
also predicted the general number of flooded structures for various flood
80


frequency ranges. The baseline conditions model shows that all 88 of the
structures in the study area are above the 10-year floodplain boundary.
The results from the TIN modeling reveal that approximately 81 structures
are above the 10-year floodplain. A comparison of the number of flooded
structures is shown in Table 5.6. The difference in these estimates is less
than 8%, suggesting that the TIN data in this study provides general
information about the total flood damages and the number of flooded
structures above a given floodplain level.
The GRID and TIN data models are very effective for the
manipulation, analysis, and display of numerous surface modeling
problems. Both of the data structures are digital abstractions or
representations of a given surface in the real world. The addition of the
third dimension in a surface model can generate new information and
insights about a study area that are not possible in a two-dimensional
analysis. The surface analysis techniques have a very high information
content and can be a powerful tool for analyzing geographic information.
Coordinate Transformation Issues
The original DEM for the Arvada, Colorado quadrangle is referenced
horizontally in the UTM coordinate system. The reference datum is the
North American Datum of 1927 (NAD27). Figures from the 1979 Clear
Creek FHAD study were used as the base maps for digitizing the outlines of
the structures in the floodplain, the local streets, and the 100-year
floodplain boundary. The coordinates shown on the FHAD maps are from
81


the State Plane Projection for the central zone of the State of Colorado. A
comparison of the State Plane coordinates on the FHAD maps with the
Arvada, Colorado 7.5 minute quadrangle map revealed a difference of
approximately 183 meters (600 feet) in the x (easting) direction and
approximately 9.1 meters (30 feet) in the y (northing) direction.
The State Plane Coordinate system shown on the 1979 FHAD maps
was adjusted to fit the State Plane reference points on the Arvada, Colorado
7.5 minute quadrangle map. These adjustments were made by measuring
common reference points on both map sources to determine the average
variability in the x and y directions. These corrections were then applied
during the digitizing of the structure outlines and the 100-year floodplain
boundary. The horizontal error in the FHAD reference system was a cause
for concern. As a result, the latest aerial topographic maps of the study
area were obtained from the City of Wheat Ridge. The city topographic
maps were generated in 1985 and are based on the North American Datum
of 1983 (NAD83). The finished floor elevations were re-determined for the
baseline conditions using these updated maps. The DEM and the digitized
structure outline and floodplain coverages were converted from NAD27 to
NAD83 to remain consistent with the re-evaluation of the finished floor
elevations.
The data shift of the original FHAD maps and the projections of the
DEM data and the digitized coverages are potential sources of error in the
computation of the flood damage input data. Another potential error source
is the transformation of all of the data from NAD27 to NAD83. The
ARC/INFO GIS software provides a detailed and accurate series of steps
for the transformation of one coordinate system into another. However,
82


because of the shift of the original FHAD data, the age of the FHAD report,
and the lack of information on the coordinate system used for the FHAD
mapping, the figures from the FHAD report used to digitize the structure and
floodplain coverages are a potentially significant source of error in this
analysis.
Probability of FFE Differences
The variations in the finished floor elevations for both the GRID and
TIN models were grouped into classes at intervals of 0.15 meters (0.5 feet).
Using these elevation classes, the relative frequency and the cumulative
relative frequency of the elevations in each class interval were computed.
The relative frequency, interpreted as a probability estimate, is calculated
by taking the number of elevations grouped in each class interval and
dividing by the total number of elevation observations. The cumulative
frequency, used to approximate a cumulative probability distribution, is
determined by summing the relative frequencies in each class interval. The
cumulative distribution shows the frequency of events less than (or greater
than) a given elevation value. A summary of the relative and cumulative
frequency data used for the GRID and TIN elevation differences is
presented in Tables 6.1 and 6.2, respectively. The frequency and
cumulative frequency distributions are presented graphically in Figures 6.2
and 6.3.
The tabular and graphical data illustrates that the majority of the
elevation differences are below 1.52 meters (5.0 feet) for the GRID model
83


and below 1.37 meters (4.5 feet) for the TIN model. The cumulative relative
frequency data in Figure 6.3 shows that the likelihood of an elevation
difference occurring below 1.52 meters (5.0 feet) is approximately 89.9% for
the GRID modeling. Figure 6.3 shows that the likelihood is 90.9% that the
elevation difference for the TIN modeling will be below 1.37 meters (4.5
feet).
The relative frequency curves of the GRID and TIN elevation
differences shown in Figure 6.2 are generally skewed to the right and
display a considerable amount of variability between the models. The
relative frequency curve for the GRID model is also multimodal, illustrated
by the alternating peaks and valleys in the elevation difference classes.
Both curves show that there are a minimal number of elevation differences
greater than 2.1 meters (7 feet), except for one outlier of 3.4 meters (11.0
feet) shown in both surface models.
Earlier in this chapter, it was stated that the average difference in the
finished floor elevations determined by the standard methods and the GRID
model is approximately 0.9 meters (2.8 feet). Further analysis of this data
shows that the standard deviation of the elevations is approximately 0.58
meters (1.89 feet). The average difference in the elevations between the
standard methods and the TIN model data is approximately 0.7 meters (2.2
feet). The standard deviation for the TIN model differences was computed
to be approximately 1.83 feet.
The results of the statistical analyses outlined above combined with
the tendency of the relative frequency curves to be skewed to the right
further supports the conclusion that the GRID and TIN modeling techniques
in this study do not provide the detail required for an accurate assessment
84


of flood damages. The relative frequency and cumulative frequency curves
display a high variability in the data points, making it difficult to accurately fit
the data to a statistical distribution.
Table 8,1 Frequency of Elevation Differences GtS GRID Model
Elevation # of Occurrences Interval (feet) Relative Frequency Cumulative Relative Frequency
0.0- -0.5 11 0.125 0.125
0.5- -1.0 4 0.046 0.171
1.0- -1.5 14 0.159 0.330
1.5- -2.0 7 0.080 0.410
2.0- -2.5 4 0.046 0.456
2.5- -3.0 12 0.136 0.592
3.0- -3.5 8 0.091 0.683
3.5- -4.0 8 0.091 0.774
4.0- -4.5 1 0.011 0.785
4.5- -5.0 10 0.114 0.899
5.0- -5.5 2 0.023 0.922
5.5- -6.0 4 0.046 0.968
6.0- -6.5 2 0.023 0.991
6.5- -7.0 0 0.000 0.991
7.0- -7.5 0 0.000 0.991
7.5- -8.0 0 0.000 0.991
8.0- 8.5 0 0.000 0.991
8.5- -9.0 0 0.000 0.991
9.0- -9.5 0 0.000 0.991
9.5- 10.0 0 0.000 0.991
10.0- -10.5 0 0.000 0.991
10.5- -11.0 1 0.011 1.000
85


Frequency of Elevation Differences GIS TIN Model
Elevation Interval (feet) # of Occurrences Relative Frequency Cumulative Relative Frequency
0.0-0.5 14 0.159 0.159
0.5-1.0 13 0.148 0.307
1.0-1.5 11 0.125 0.432
1.5 2.0 12 0.136 0.568
2.0-2.5 9 0.102 0.670
2.5-3.0 11 0.125 0.795
3.0-3.5 5 0.057 0.852
3.5-4.0 2 0.023 0.875
4.0-4.5 3 0.034 0.909
4.5-5.0 2 0.023 0.932
5.0-5.5 1 0.011 0.943
5.5-6.0 2 0.023 0.966
6.0-6.5 1 0.011 0.977
6.5-7.0 0 0.000 0.977
7.0-7.5 1 0.011 0.988
7.5-8.0 0 0.000 0.988
8.0-8.5 0 0.000 0.988
8.5-9.0 0 0.000 0.988
9.0-9.5 0 0.000 0.988
9.5-10.0 0 0.000 0.988
10.0-10.5 0 0.000 0.988
10.5-11.0 1 0.011 1.000
86


TIN Data
Elevation Difference
FIGURE 6.2 Relative Frequency
May 1996