Environmental Modelling & Software 26 (2011) 817e821
Contents lists available at ScienceDirect
Environmental Modelling & Software
journal homepage: www.elsevier.com/locate/envsoft
Short communication
A GIS-based decision support system for integrated flood management under
uncertainty with two dimensional numerical simulations
Honghai Qi a, *, M.S. Altinakar b,1
a
b
National Center for Computational Hydroscience and Engineering (NCCHE), University of Mississippi, University, MS 38677, USA
NCCHE, University of Mississippi, University, MS 38677, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 19 January 2010
Received in revised form
19 November 2010
Accepted 23 November 2010
Available online 8 January 2011
A new decision support system has been developed for integrated flood management within the
framework of ArcGIS based on realistic two dimensional flood simulations. This system has the ability
to interact with and use classified Remote Sensing (RS) image layers and other GIS feature layers like
zoning layer, survey database and census block boundaries for flood damage calculations and loss of life
estimations. It also provides a user friendly interface which allows construction of user defined criteria,
such as stage-damage curves, running computations and visualization of the results. Monte Carlo
Simulation method is used to take into account uncertainties in various variables and parameters, and
event tree analysis is used to estimate the population dynamics. The analysis of a dam break flood
management strategy for Sinclair Dam in Georgia, USA is chosen as a case study to demonstrate the
capabilities of the decision support system. The test results compared with HEC-FDA software indicate
that this new system provides a very versatile and reliable environment for estimating various flood
damage, and may greatly enhance decision making process for future design of the flood proofing
facilities.
Ó 2010 Elsevier Ltd. All rights reserved.
Keywords:
Integrated flood management
Decision support system
2D simulation
ArcGIS
Remote sensing
Census block
1. Introduction
Natural or man-made disasters produce damages with varying
degrees of intensity in a given area. Flooding, a frequently
occurring hazard, costs human hardship and economic loss
whose intensity varies in space. The flood management can be
considered as a spatial problem (Simonovic, 2002; Luino et al.,
2009). Flood Protection can be achieved through various structural measures such as dykes, diversion channels, reservoirs, and
non-structural measures such as flood warning, mass evacuation
and etc. Spatial representation of the consequences of these flood
mitigation alternatives provides a better insight into their relative
effectiveness and facilities decision making (Jonkman et al., 2008;
Ernst et al., 2008). Usually, comparison of different flood
protection measures and evaluation of their impacts are based on
multiple criteria (Bouwer et al., 2009; Pingel and Watkins, 2010).
* Corresponding author. Present address: NMP Engineering Consultants, Inc.,
Executive Plaza III, Suite 300, 11350 McCormick Rd, Cockeysville, MD 21031, USA.
Tel.: þ1 410 771 99808; fax: þ1 410 771 9809.
E-mail addresses: hqi@nmpengineering.com (H. Qi), altinakar@ncche.olemiss.
edu (M.S. Altinakar).
1
Tel.: þ1 662 915 3783; fax: þ1 662 915 7796.
1364-8152/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.envsoft.2010.11.006
These criteria usually include loss of life and urban flood damage.
Recent advancements in reliable two dimensional (2D) flood
modeling, geographical information system and remote sensing
technology open a promising new path for flood hazard
management since it takes into account of spatial variability of
flood hazard.
The main objective of the research described in this paper is to
develop a decision support system in widely used GIS environment.
Based on the computational results of a 2D flood analysis model,
crucial criteria that are usually implemented for evaluating flood
hazard, such as loss of life and flood damage (in urban and rural area)
can also be calculated within the framework of this system, by
interacting with other GIS feature layers like census block layer and
remote sensing image. The case study of the dam break flood case of
Sinclair Dam in Milledgeville of Georgia, USA is used to test the
capabilities of the system. The versatile environment for construction different criteria and use of other GIS feature layer demonstrate
that the GIS-based decision support system can provide the user
better decision making aid. The results of the case study validated by
HEC-RAS and HEC-FDA software (US Army Corps of Engineers, 2010)
clearly show that this new approach based on 2D simulation results
allows the stake holders to have a better appreciation of the
consequences of the flood.
818
H. Qi, M.S. Altinakar / Environmental Modelling & Software 26 (2011) 817e821
2. Developing multiple flood management decision making
criteria
2.1. Loss of life estimation with census block boundary
According to Graham (1999), loss of life resulting from flooding
is highly influenced by three factors: 1) The number of people
occupying the floodplain, which is also called Population at Risk
(PAR); 2) The amount of warning that is provided to the people
exposed to dangerous flooding and 3) The severity level of the
flooding. For this research, spatial variability of the above factors is
taken into account for each location of the area of interest, which is
a completely two dimensional approach. Fig. 1 shows the flowchart
to estimate the loss-of-life due to a flood within the decision
support system.
In order to determine the PAR value, the census block data,
which is usually a vector polygon layer (for example, in TIGER
format) are used in GIS environment (Poser et al., 2009). Census
blocks are small in area showing population variation. After
importing this layer into ArcGIS, the population density is first
calculated by using the total population of each census block and its
area. This feature polygon layer is then converted to a raster layer
which has the same cell size as the flood computation results. The
cell value, which represents the PAR living and working inside each
cell, is reclassified according to the product of population density
and the cell area. This operation would obtain a raster layer
showing the PAR distribution (Qi et al., 2005).
The typical definition of warning time of a flood is the length of
time from when the first public warning is issued until the flood
wave reaches the first person in the PAR (Aboelata et al., 2002). The
flood severity definition is usually associated with the flood depth.
Low, medium and high severity can be categorized according to
Graham (1999). For instance, the equation for retrieving the life loss
of a sub category layer which satisfies a medium flood severity with
no warning time in the domain is:
½LossLife ¼
½D < Hhigh &ð½D > ¼ Hlow Þ
&ðð½AT
Wissue Þ < Wnw Þ ½Cs Rf
(1)
where [LossLife] [D], [AT] and [Cs] represent raster layer of loss of
life, flood depth, arrival time and census block information
respectively, Hhigh/Hlow are the limits for high/low severity flood
depth, Wissue is the initial time of a public warning, Wnw is the time
limit for no warning, and Rf is the corresponding fatality rate. The
total life loss distribution layer is then obtained by summing up the
entire sub-category layers. Parameters of Hhigh, Hlow, Wissue, Wnw
and Rf can be found from Graham (1999).
GIS and
Remote
Sensing
Census Block
boundary layer
User
Awareness
level of
population
2.2. Urban flood damage calculation with remote sensing image
The second criterion used in the evaluation of the alternatives
is the dollar value of damage to the flooded structures within the
region of interest. Field surveys or interviews and expert panel
opinions are the two primary sources of the data used to develop
depth-damage relationships for structural and its content groupings and alternative residence types. Flood depth here refers to the
depth of the flooding above or below the first floor of the structure. The percentage damage to the structure refers to the percent
of the total depreciated replacement cost of the structure that is
damaged (US Army Corp of Engineers, 1997). Remote Sensing (RS)
is the recently developed science and technology of obtaining
information about an object, area or phenomenon through the
analysis of data acquired by a device that is not in contact under
investigation (Lillesand, 1999). The satellite image can provide
important information by showing various urban land cover
features, like vegetation, residential area or water body. Since
different land feature types have their inherent spectral reflectance and emittance properties, the RS image is usually classified
so that all the pixels in this image fall into certain land over classes
or themes. Each class of the land features manifests a unique
Digital Number (DN) value. With overlaying of this classified RS
image with the flood inundation image, the flood damage calculation can be achieved using arithmetic and relational raster map
algebra (Qi et al., 2006).
3. Case study of a dam break flood analysis
A dam break flood analysis of Sinclair Dam near Milledgeville,
Georgia of the United States has been chosen to test the capability
of the designed decision support system. Completed in 1953 by the
Georgia Power Company, the Sinclair Dam is located on Oconee
River upstream of historical Milledgeville, GA, which was the state
capital of Georgia from 1803 to 1868. The dam consists of concrete
non-overflow and spillway sections, flanked by reinforced earthen
embankments. The span of the dam is 2988 ft (910.74 m). The dam
height is 104 ft (31.70 m). Sinclair Dam is classified as a high-hazard
dam. The river at the dam drains a total basin area of 13,805 km2.
The volumes at the normal storage level and the maximum storage
level are 333,000.0 acre-foot (410,749,457 m3) and 490,000.0 acrefoot (604,406,109 m3), respectively. The spillway has a width of
870.0 ft (265.2 m). The maximum discharge is 479,000.0 cfs
(13,563 m3/s). The floodplain downstream of the dam has a surface
area of about 669 km2. The population (2000 Census) of Milledgeville is about 44,700 (Fig. 2).
Census Raster
layer (PAR)
Number of
fatalities
Flood warning
time
CCHE2D FLOOD
Flood arrival
times
Flood depths
Time available
for escape
Fatality Rate
Flood severity
Empirical Relationship based
on past events (USBR)
Fig. 1. Flowchart showing the procedure to estimate probable loss-of-life due to a flood.
H. Qi, M.S. Altinakar / Environmental Modelling & Software 26 (2011) 817e821
819
Fig. 2. Study area: Milledgeville, Georgia in Southeastern United States (Scale 1:3000).
4. Results and validations
The main objective of the floodplain analysis is to evaluate the
consequences of the flood resulting from complete Sinclair Dam
failure happened on a sunny weekday. The dam break case was
assumed to occur at when the reservoir has normal storage level of
water. Loss-of-life and urban damage analysis were carried out
using the decision support system.
The loss-of-life potential is quite high due to the assumption of
sudden and complete failure of the entire dam. This is a highly
improbable scenario, and the value should be regarded as an upper
bound. A gradual failure would probably result in much smaller
loss-of-life potential due to not only the reduced peak discharge,
thus lower inundation depths and velocities, but also due to longer
escape time for the PAR. In fact loss-of-life potential is extremely
sensitive to the time when the alarm was given. Based on the user
defined probability distributions, the Monte Carlo simulation
method was used to take into account the uncertainties. The
number of Monte Carlo runs was set to 5000 for each analysis. The
Table 1
Input data for loss-of-life estimation with probability distribution functions.
Category
Name
Primary value
Probability
distribution
Related
parameters
Flood
severity
Warning
time
High
Low
Initial
Normal
Normal
Uniform
Variance ¼ 0.35
Variance ¼ 0.5
Range (15, 35)
Census
Adequate
No
PAR
>20 ft (6.1 m)
<15 ft (4.6 m)
20 min (after
flood)
>70 min
<25 min
Census raster
Uniform
Uniform
Normal
Fatality rate
FR
Default value
N/A
Range (60, 80)
Range (15, 35)
Variance ¼ 6%
PAR
N/A
input data used for the spatial Monte Carlo Analysis of loss of life
and flood damage for both urban and rural areas are presented in
Tables 1 and 2. “No of houses” means numbers of house units
within one cell of DEM used for the analysis.
Fig. 3 Left shows a map of the expected values of the estimated
loss-of-life potential after 5000 Monte Carlo Simulation runs.
Although not shown here, one can easily create maps for other
statistical properties such as standard deviation, skewness,
kurtosis, etc. Right shows a map of the expected values of the
estimated urban damage potential in US dollars based on the three
categories of building stock: High Intensity Residential Area (HIRA),
Low Intensity Residential Area (LIRA), and Commercial Industrial
Transportation Area (CITA). This analysis was carried out using the
urban flood damage toolbox of the GIS-based decision support
system.
The results of loss of life and urban damage computed by
CCHE2D-Flood and Decision Support System (DSS) are summarized
in Table 3. As compared with the results obtained from same
scenario analysis by the widely used and accepted HEC-RAS dam
break simulation and HEC-FDA, it is found that the CCHE2D-Flood
and DSS system provide very close and accurate estimates (error
<5%). The major reason for the difference results from the difference from 1D flood simulation by HEC-RAS and 2D flood simulation
by CCHE2D-Flood. Downstream of Sinclair Reservoir sits within the
wide FEMA 100-year floodplain of Oconee River, where there is not
a well defined valley. 1D model assumes that the flow fills the entire
cross section instantly. When flood flow enters a flat area with
poorly defined channel, 1D flow hypothesis breaks down, and it
may lead to significant errors in estimating arrival time, flow depth,
and flow duration. At those locations, 2D model can better simulate
the hydrodynamics of flood situation in flat area. The advantages of
using the DSS system also lie in that various maps showing spatially
820
H. Qi, M.S. Altinakar / Environmental Modelling & Software 26 (2011) 817e821
Table 2
Input data for urban flood damage analysis with probability distribution functions.
Category
Name
Primary value
Prob. distribution
Related parameters
High intensity residential
area (HIRA)
No of houses
Structure value
Content value
No of houses
Structure value
Content value
No of houses
Structure value
Content value
Other value
HIRA
10 Units
$98,000
$75,000
6 Units
$120,000
$90,000
8 Units
$150,000
$100,000
$30,000
f ¼ 0.68h4 þ 7.78h3 34.52h2
þ 78.7h þ 4.95
f ¼ 2.61h4 þ 25.37h3
78.42h2
þ 95.89h þ 4.31
f ¼ 0.81h4 þ 9.25h3 38.13h2
þ 71.20h þ 4.77
Triangular
Normal
Normal
Triangular
Normal
Normal
Triangular
Normal
Normal
Normal
Triangular
min ¼ 8, max ¼ 14
Variance ¼ 0.45
Variance ¼ 0.35
min ¼ 4, max ¼ 10
Variance ¼ 0.45
Variance ¼ 0.35
min ¼ 6, max ¼ 12
Variance ¼ 0.25
Variance ¼ 0.35
Variance ¼ 0.4
min ¼ 5%f,
max ¼ þ5%f
min ¼ 5%f,
max ¼ þ5%f
Low intensity residential
area (LIRA)
Commercial industrial
transportation area (CITA)
Depth e % Damage relationship
(h: flood depth in feet,
f: damage value in $)
LIRA
CITA
Triangular
Triangular
min ¼ 5%f,
max ¼ þ5%f
Fig. 3. Map of the expected values of the estimated loss-of-life potential and urban flood damage obtained following a Monte Carlo simulation with 5000 runs (Scale 1:3000).
Table 3
Summary of estimated loss-of-life potential and flood damage results compared
with HEC-RAS and HEC-FDA analysis.
Category
Items
CCHE2D-flood
and DSS results
HEC-RAS and
HEC-FDA results
Loss of life
Number of fatalities
147 person
3.45 10 $
N/A (no such
function)
3.50 108 $
2.62 108 $
2.71 108 $
5.81 108 $
5.72 108 $
1.18 109 $
1.19 109 $
Urban flood High intensity residential
damage area (HIRA)
Low intensity residential
area (LIRA)
Commercial, industrial
transportation area (CITA)
Total
8
distributed loss of life and flood damage information can be provide
rather than damage report and plan performance in forms and
charts by HEC-FDA.
5. Conclusions
The present study shows risk and uncertainty analysis based on
2D numerical simulation results, GIS and remote sensing technologies can significantly improve the accuracy of flood hazard
assessment. This approach efficiently assists in evaluation and
ranking of flood control management strategies, and future design
of flood proofing works. The resulting raster/vector maps of the case
study showing spatial distribution of loss of life and flood damage
can greatly enhance decision making process for future planning of
emergency management operations. Currently, the studies are
underway to incorporate the discharge-exceedance probability
functions and flood frequency probability functions into the spatial
risk and uncertainty analysis. It is also important to underline the
fact that the currently available relationships used for estimating
property damage are generally expressed as a function of flood
depth only. The accuracy of predictions can probably be improved
by taking into account other detailed information provided by 2D
flood modeling, such as flood velocity and duration. Further
research is needed to develop such improved damage relationships.
References
Aboelata, M. et al., GIS Model for Estimating Dam Failure Life Loss. Invited paper. In:
The Tenth Engineering Foundation Conference on Risk-Based Decision making
in Water Resources: Protection of the Homeland’s Water Resources Systems,
Santa Barbara, California. November 2002.
Bouwer, L.M., Bubeck, P., Wagtendonk, A.J., Aerts, J.C.J.H., 2009. Inundation scenarios
for flood damage evaluation in Polder areas. Natural Hazards and Earth System
Sciences 9, 1995e2007.
H. Qi, M.S. Altinakar / Environmental Modelling & Software 26 (2011) 817e821
Ernst, J., Dewals, BJ., Giron, E., W Hecq, W., Pirotton, M., 2008, Integrating Hydraulic
and Economic Analysis for Selecting Flood Protection Measures in the Context of
Climate Change. In: 4th International Symposium on Flood Defence: Managing
Flood Risk, Reliability and Vulnerability, Toronto, Ontario, Canada, May 6e8, 2008.
Graham, W.J., 1999. A Procedure for Estimating Loss of Life Caused by Dam Failure.
Report No. DSO-99e06. Dam Safety Office, US Bureau of Reclamation, Denver, CO.
Jonkman, S.N., Bo
ckarjovab, M., Kokc, M., Bernardinid, P., 15 May 2008. Integrated
hydrodynamic and economic Modelling of flood damage in the Netherlands.
Ecological Economics 66 (1), 77e90.
Lillesand, Thomas M., 1999. Remote Sensing and Image Interpretation, Fourth ed.
John Wiley & Sons, Inc. Press, ISBN 0-471-25515-7.
Luino, F., Cirio, C.G., Biddoccu, M., Agangi, A., Giulietto, W., Godone, F., Nigrelli, G.,
September 2009. Application of a model to the evaluation of flood damage.
GeoInformatica 13 (3), 339e353.
Pingel, Nathan, Watkins Jr., David, 2010. Multiple flood source expected Annual
damage computations. Journal of Water Resources Planning and Management
136 (3), 319e326.
Poser, Kathrin, Kreibich, Heidi, Dransch, Doris, 2009. Assessing Volunteered
Geographic Information for Rapid Flood Damage Estimation. In: 12th AGILE
821
International Conference on Geographic Information Science 2009. Leibniz
Universität Hannover, Germany.
Qi, Honghai, Altinakar, M.S., Ying, X., Wang, S.S.Y, 2005. Risk and Uncertainty
Analysis in Flood Hazard Management Using GIS and Remote Sensing Technology. In: Proc.. of American Water Resources Association Annual Conference,
Seattle, Nov 2005.
Qi, Honghai, Altinakar M.S., Youngho Jeon, 2006, A Decision Support Tool for Flood
Management under Uncertainty Using GIS and Remote Sensing Technology. In:
Proceedings of 7th International Conference on Hydroscience and Engineering,
Drexel University, Sept 2006, Philadelphia, United States.
Simonovic, S.P., 2002, A Spatial Fuzzy Compromise Programming for Management
of Natural Disasters, ICLR Research Paper Series e No. 24.
US Army Corp of Engineers, 1997, Depth-Damage Relationships for Structures,
Contents and Vehicles and Content-To-Structure Value Ratios (CSVRs) in
Support of the Jefferson and Orleans Flood Control Feasibility Studies, Final
Report, Baton Rouge, Louisiana, United States;
US Army Corps of Engineers., 2010. HEC-FDA: Flood Damage Reduction Analysis
User’s Manual, Version 1.2.5. Hydrologic Engineer Center, Davis, California,
United States.