Académique Documents
Professionnel Documents
Culture Documents
Abstract: A direct-processing approach to river system floodplain delineation is developed. Floodplain zones of part of
the South Nation River system, located just east of Ottawa, Ontario, are mapped in two dimensions and three dimen-
sions by integrating the hydraulic model of the choice with geographic information systems (GIS). The first objective
was to construct and validate a Hydrologic Engineering Center’s River Analysis System (HEC-RAS) river network
model of the system using existing HEC-2 model-generated data. Next, HEC-RAS simulations were performed to gen-
erate water surface profiles throughout the system for six different design storm events. The in-channel spatial data of
HEC-RAS were then geo-referenced and mapped in the GIS domain and integrated with digital elevation model (DEM)
over-bank data to build a triangular irregular network (TIN) terrain model. In the final step, floodplain zones for the
six design storms were reproduced in three dimensions by overlaying the integrated terrain model for the region with
the corresponding water surface TIN.
Key words: river, floodplain, delineation, GIS-approach, HEC-2 model, HEC-RAS model, data query.
Résumé : Une approche de traitement direct pour la délimitation des plaines inondables des bassins hydrographiques
est mise sur pied. Les zones de plaines inondables d’une partie du réseau hydrographique de la rivière South Nation,
située tout juste à l’est d’Ottawa, Ontario, sont cartographiées en 2D et en 3D par l’intégration du modèle hydraulique
choisi avec les données du SIG (système d’information géographique). Le premier objectif était de construire et de va-
lider un modèle de réseau hydrographique HEC-RAS du bassin utilisant les données existantes générées par le modèle
HEC-2. Ensuite, des simulations HEC-RAS ont été effectuées pour générer les profils des eaux de surface dans le ré-
seau pour six différents événements d’averse de projet. Les données spatiales HEC-RAS pour le lit ont ensuite été réfé-
rencées géographiquement et cartographiées dans le domaine SIG puis intégrées aux données de lit d’inondation du
modèle altimétrique numérique (DEM) afin de bâtir un modèle du terrain selon un réseau triangulé irrégulier (TIN). À
la dernière étape, les zones de plaines inondables pour les six événements d’averse de projet ont été reproduites en 3D
en superposant le modèle intégré de terrain pour la région et la surface d’eau du TIN correspondant.
Mots clés : rivière, plaine inondable, délimitation, approche SIG, modèle HEC-2, modèle HEC-RAS, demande de ren-
seignements.
[Traduit par la Rédaction] Yang et al. 28
Can. J. Civ. Eng. 33: 19–28 (2006) doi:10.1139/L05-102 © 2005 NRC Canada
20 Can. J. Civ. Eng. Vol. 33, 2006
Fig. 1. (a) South Nation River watershed and (b) phase 1 study area of the Bear Brook sub-watershed.
(a) (b)
Lower
Watershed
data types for display, query, and analysis. The Avenue to perform specific tasks, or develop an application that re-
Scripts and ArcView extensions HEC-GeoRAS, Spatial An- lies on the ArcView graphical user interface.
alyst, and 3-D Analyst were applied here in the floodplain
delineation exercises. The HEC-GeoRAS interface (USACE Methodology and implementation
1999) processes geo-spatial data between HEC-RAS and
GIS. Users can then create import files containing geometric Once the HEC-2 data were successfully converted and re-
attribute data from existing DTM and selected complemen- produced in HEC-RAS, a complete river network model was
tary data sets, such as stream centerlines and banks. Spatial generated. The network model was created to provide in-
Analyst, which enables the creation, display, and querying stream geometry data and water surface elevations for terrain
of rastar data, is necessary for the continuous display of sur- model and data query model construction, as well as for the
face modeling, e.g., a DEM, which cannot be modeled using 2-D and 3-D floodplain mapping. The various steps in the
vector data. Spatial Analyst also performs integrated raster- hydraulic modeling and 3-D floodplain delineation exercises
vector theme analysis, thereby allowing for the integration are summarized in Fig. 2.
of properties in a raster theme based on overlaid vector
theme. A 3-D Spatial Analyst allows model representation in Hydraulic modeling
three dimensions as well as real-time perspective viewing. The first steps in the modeling process were modifica-
Users can therefore create and visualize spatial data in three tion and importation of the existing HEC-2 single reach
dimensions to provide insight, reveal trends, and solve prob- data into HEC-RAS. Several issues had to be addressed
lems. Avenue Script is used to customize ArcView, direct it here, including:
Established
HEC-RAS river network model
HEC-2 model
(1) HEC-RAS does not allow importation of HEC-2 data in ing the equivalent HEC-RAS model, and constructing the
free format, and the river-station order sequence of the river network model are described in greater detail else-
imported reach must start at the upstream end and pro- where (Yang 2004).
ceed downstream.
(2) Some HEC-2 options are not available in HEC-RAS, Three-dimensional floodplain delineation
e.g., (i) compute Manning’s n from high water marks A 10 m interval DEM was used to represent study area
(J1), (ii) define internal rating curves (RC), (iii) create terrain. Compared to HEC-RAS, a DEM offers lower accu-
archive (AC), (iv) allow the input data to be free format racy when describing in-stream channel geometry. There-
(FR), and (v) create storage outflow data (J4). fore, an integrated digital terrain model in the form of a TIN,
(3) To have HEC-RAS reproduce HEC-2 results for channel containing both HEC-RAS in-stream data and DEM over-
flows in reaches without bridges or culverts, the same bank data, was developed. The resulting TIN model was ap-
conveyance (HEC-2 conveyance subdivision method) and propriate for 3-D visualization of the terrain and floodplain
critical depth (parabolic) computational methods used in zones. A set of Avenue Scripts (Tate 1999), was used to
HEC-2 must be selected for the HEC-RAS simulations. transfer the HEC-RAS output data corresponding to the se-
(4) More detailed bridge and culvert information is required lected design storms to the GIS domain. The HEC-RAS in-
for HEC-RAS flow analyses in these type of locations stream geometry and flood elevation information were asso-
including bridge low chord data, bridge deck data, and ciated with each cross-section cutline in the GIS domain us-
culvert geometry parameters. ing the stream centreline as the connectivity.
The Bear Brook River system examined in phase 1 con- Since HEC-RAS and DEM data were collected at differ-
tained 11 (main channel and 10 tributary) reaches. The river ent times and had different resolutions, the elevation data
network is represented by stream centrelines, which are de- were not the same in the overlapped parts of the floodplain
noted as straight lines in the established single reach HEC- zones. To smooth the elevation differences between the two
RAS model. To connect the 11 reaches in HEC-RAS, the data sources, the cross section was modified using Avenue
GIS stream centreline information (containing the coordi- Script (Tate 1999). The modified cross section extracted
nates defining the real-world location of the stream network) from the DEM-RAS integrated terrain TIN is identical to the
had to be imported into HEC-RAS. This required (i) digitiz- HEC-RAS in-channel cross section and begins to approxi-
ing the stream centrelines in ArcView using an aerial photo mate the DEM-extracted cross section as the outer limits of
base map, (ii) creating the river network using GeoRAS and the floodplain zones are approached. Figures 3a and 3b
using the snap tool in ArcView to connect reaches, (iii) pro- compare the DEM-RAS integrated TIN with the terrain TIN
ducing a HEC-RAS import file of stream centreline theme derived from DEM data alone, and Fig. 3c shows a typical
containing geometric attribute data from a DTM, (iv) im- modified stream cross-section profile together with corre-
porting GIS format stream centreline data to HEC-RAS, and sponding profiles for the DEM and HEC-RAS data. As
adding other geometry data (banks, station ID, cross-section shown in the figure, the DEM-derived terrain TIN and the
geometry, etc.) from the aforementioned 11 single each DEM-extracted cross section appear flat in the stream chan-
models, (v) performing HEC-RAS simulations, and (vi) fine nels when compared with the integrated terrain TIN and the
tuning the HEC-RAS model by comparing corresponding modified cross section. This is because the 10 m interval
HEC-2 and HEC-RAS output data. The various procedures DEM data do not provide sufficient detail for proper channel
followed in processing the HEC-2 model input data, validat- representation when using 10 cells. With its higher density
© 2005 NRC Canada
Yang et al. 23
Fig. 3. (a) Digital elevation model (DEM)-derived terrain TIN, (b) digital elevation model - river analysis system (DEM-RAS) inte-
grated terrain TIN, and (c) example of a modified cross section (C/S) 161585.
(a) (c)
71
C/S 161585 70
69
68
Elevation (m)
67
66
65
64
(b)
63
62
0 50 100 150 200 250 300 350
Distance from left bank (m)
C/S 161585
DEM extracted C/S HEC-RAS C/S Modified C/S
of points within the channel, the integrated terrain TIN is Fig. 4. Steps in the 3-D floodplain visualization process for the
clearly more appropriate for creating reliable floodplain maps. 100-year flood event (a) water surface TIN, (b) integrated terrain
Yang (2004) provides further details on the various steps TIN, (c) floodplain 3-D view, and (d) floodplain 3-D photo view.
followed in synthesizing the terrain TIN.
(a)
For each flood event, water surface elevation and terrain-
water surface interface location data were extracted from
HEC-RAS and transferred to the GIS cross-section files to
generate water surface TIN. Flooded areas (occurring where Water surface TIN
Brook River are a fixed data set, its terrain TIN model may
be applied without adjustment to all floodplain simulations. (c)
Based on the respective water surface elevation and terrain-
water interface location data, separate water-surface TINs
were generated for each flood event considered. Figure 5
shows a sample portion of the inundated floodplain zones of
3-D floodplain visualization
the Bear Brook River for the 10, 50, and 100-year design
storms, projected onto an aerial photo of the region.
Flooding occurs to a depth equal to the difference be- (d)
tween the water-surface TIN and surrounding terrain TIN.
Converting both TIN themes into grids allows for subtrac-
tion of the elevation data using the raster calculator in GIS.
3-D floodplain photo view
The terrain raster theme is then subtracted from the water
surface grid theme to create a flood depth grid. Using the
query tool in GIS, the flow depth at any point within the in-
undated floodplain zones can be obtained by simply clicking
© 2005 NRC Canada
24 Can. J. Civ. Eng. Vol. 33, 2006
Fig. 5. A portion of the inundated floodplain zones of the Bear Brook River for the 10, 50, and 100-year design storms.
Legend
10-year flood
50-year flood
100-year flood
metres
Fig. 6. An application of the data query models relating to: (a) flood depth, and (b) cross-section information.
Fig. 7. Flood zone for the 100-year storm event, showing depth information.
Legend
Flood depth (100 years) metres
Table 2. Differences in water surface elevations between HEC-2 and HEC-RAS for the entire river network.
Maximum difference Number of cross sections for different ranges (%)
Reach Station Description Value (%) 0–0.1 0.1–0.2 0.2–0.3 0.3–0.4
Main 173860 HEC-2 special bridge 0.372 72 21 3 1
Tributary-1 17588 Junction 0.202 7 1 2 0
Tributary-2 9561 Culvert 0.231 91 5 2 0
Tributary-3 509 Junction 0.089 9 0 0 0
Tributary-4 13160 Culvert 0.220 6 2 1 0
Tributary-5 4110 HEC-2 special bridge 0.328 19 7 1 1
Tributary-6 21310 HEC-2 special bridge 0.279 28 2 1 0
Tributary-7 444 Junction 0.110 8 2 0 0
Tributary-8 250250 Cross section adjacent to junction 0.071 10 0 0 0
Tributary-9 270300 Cross section adjacent to junction 0.082 5 0 0 0
Tributary-10 283461 Culvert 0.272 10 0 1 0
Number of cross sections 265 40 11 2
achieved by comparing the HEC-RAS-generated results with Fig. 8. Differences in water surface elevations between HEC-2
the HEC-2 output data and re-calibrating the former where and HEC-RAS simulations.
necessary. There are 318 designated cross sections in the
Bear Brook River system, including 16 bridge and 14 cul-
vert sections. According to the results presented in Table 2
and Fig. 8 when compared with water surface elevations
generated by HEC-2, 83% of the HEC-RAS cross sections
have water surface elevation differences less than 0.1% (0–
0.2 ft (1 ft = .3048 m)), 13% have differences ranging from
0.1%–0.2% (0.2–0.4 ft), and only 4% have differences larger
than 0.2% (>0.4 ft). Maximum differences (0.071%–
0.372%) between the two sets of results for all reaches were
observed at channel junctions, cross sections adjacent to
these junctions, special bridges, and culvert locations. This
is mainly because (i) HEC-RAS computes energy losses re-
sulting from confluence (junction) flows, whereas HEC-2 one for the bottom), whereas HEC-2 uses the bottom coeffi-
does not account for such losses; (ii) the HEC-RAS low cient only in its culvert flow simulations (USACE 1991,
flow bridge analysis method is superior to the HEC-2 proce- 2001). Certain differences noted between corresponding
dure. (HEC-RAS relies on the actual bridge opening geome- computed water surface elevations clearly resulted from the
try, whereas HEC-2 assumes the presence of a trapezoidal aforementioned different assumptions and computational
section); (iii) HEC-RAS computes two roughness coeffi- routines in HEC-2 and HEC-RAS regarding the modeling of
cients inside a culvert barrel (one for the top and sides and bridge, culvert, and junction flows.
Fig. 9. Three-dimensional stream centerline and bank theme (a) before interpolation and (b) after interpolation.
(a) 2-D stream centreline
3-D stream centerline and bank
theme (before interpolation)
The accuracy of the floodplain delineation process was in- same event provided by SNC (Fig. 10). The SNC flood
fluenced by polygon, which is based on HEC-2 model water surface sim-
• Accuracy of the HEC-RAS model: The accuracy of the ulations, was produced by manually plotting the flood
HEC-RAS water surface simulations directly impacts the boundaries on a topographic map. While the 2-D and 3-D
resulting flood zones. The selection of computational floodplain simulations display similar plan-form shapes, the
methods and modifications to the original HEC-2 dataset 3-D simulation also provides detailed flood depth informa-
are important factors in the floodplain delineation process. tion. Accordingly, the latter is viewed as a superior
• Accuracy of the terrain: The terrain TIN model developed floodplain management tool.
in phase 1 integrates two sources of data: HEC-RAS and
DEM data. Large differences are observed between the
DEM-extracted and HEC-RAS cross sections (Fig. 3c). Conclusions
The differences in elevation between the DEM and HEC-
RAS cross sections may widen or narrow the resulting The modeling procedures developed in this research were
flood zones. Accordingly, accuracy of the river network initially undertaken to assist the South Nation Conservation
model is dependent on both the field survey data for the Authority in simplifying and reducing the time factor in-
HEC-RAS in-channel geometry and the resolution of the volved in the floodplain delineation process. The notable ac-
DEM data. A sensitivity analysis is needed to investigate quirement from this project is to automate and simplify the
the degree of precision of the terrain model. Model per- flood mapping process for an entire river system based on an
formance and accuracy should be examined by comparing established HEC-2 model.
DEM elevation data with measurements in the field. The HEC-RAS river network model developed through
• Cross-section interpolation: A reasonable number of cross this research provides improved simulations with its en-
sections were required to closely represent the shape of hanced computational routines (analysis of flow at junctions,
the stream channel, especially in the vicinity of stream bridges, and culverts), supports both the importation and ex-
bends. To artificially increase the number of cross sec- portation of GIS data, allows for simple and quick cross-
tions, the cross-section interpolation option must be em- section interpolation and allows the user to view the river
ployed in both HEC-RAS and GIS. Figures 9a and 9b reach and cross-section data in three dimensions.
show the 3-D stream centerline and bank theme before The methodology used to automate the 3-D floodplain de-
and after interpolation, respectively. Both the HEC-RAS lineation process results in (1) providing a reliable and more
and GIS cross-section interpolations are based on a linear economical approach to delineate flood zones; (2) enabling
assumption. If the stream-channel geometry between two users to quickly update flood maps, with changes in the hy-
adjacent surveyed cross sections does not change linearly, draulic conditions; (3) visualizing floodplain zones in three
then the intermediate (interpolated) cross section will not dimensions; and (4) enabling quick data query.
closely represent the actual (field) geometry. Additional Using an aerial photo as a base map, the impacts of se-
surveyed cross sections would be required to more accu- lected design storms, on both river and nearby structures,
rately represent field condition, especially, where the can be quickly assessed. Figure 11 shows an example of
stream channel is tortuous. how the resulting flood map can be used for flood emer-
• Cross-section extent: Since the outer limits of a cross sec- gency prediction. The 100-year floodplain modeling result
tion were assumed to be the outer boundaries of the flood, provides information concerning the flood extent and depth
an interactive process between HEC-RAS and GIS is re- at the building of interest. This will possibly help in locating
quired to extend the cross sections in the HEC-RAS those areas in need of remediation work and identify suitable
model to cover the entire flood zone. locations and arrangements for flood prevention and (or)
To validate the flood mapping exercise, a 3-D 100-year control structures. The data query models constructed in this
flood zone for the study portion of the Bear Brook sub- research can be applied to the entire watershed for river net-
watershed was compared with the 2-D flood polygon for the work data management purposes.
Fig. 10. Comparing the model-generated 100-year flood zone with corresponding SNC data.
Legend
100-year flood (SNC) 0 500 1000 2000 3000 4000
metres
100-year flood (model)
Fig. 11. Flood emergency prediction at river station 166310 for the 100-year storm event.
Cross-section 166310
cutline
structure of interest
100-year
floodplain
boundary
0 5 10 20 30 40
metres
This research deals only with integrating the hydraulic data be modified to best serve the goal of the user with regards to
with GIS to map the desired inundated floodplain zones. An existing hydraulic models.
important future research effort in this area would be to in-
corporate the hydrological aspects of the watershed into the
modeling process. Such a contribution would lead to a more
effective flood emergency warning and management tool.
References
The present HEC-GeoRAS extension allows the user to ex- Beavers, M.A. 1994. Floodplain determination using HEC-2 and
tract cross-section geometry from a terrain model and map geographic information systems. M.Sc. thesis, Department of
the flood zones based on these extracted data. However, the Civil Engineering, The University of Texas at Austin, Tex.
user cannot import the hydraulic data directly into the GIS Boss International. 2000. RiverCAD user’s manual. Boss Interna-
domain without real-world coordinates. The program should tional, Madison, Wis.
ESRI. 1992. ArcView user’s guide. 2nd ed. Environmental Systems USACE. 1991. HEC-2: water surface profiles, user’s manual. U.S.
Research Institute, Redlands, Calif. Army Corps of Engineers, Hydrologic Engineering Center, Da-
ESRI. 1998. AvRAS extension. Environmental Systems Research vis, Calif.
Institute, Redlands, Calif. USACE. 1999. HEC-GeoRAS: an extension for support of HEC-
Jenson, S.K., and Dominique, J.O. 1988. Extracting topographic RAS using ArcView, user’s manual. U.S. Army Corps of Engi-
structure from digital elevation data for geographic information neers, Hydrologic Engineering Center, Davis, Calif.
systems analysis. Photogrammetric Engineering and Remote USACE. 2001. HEC-RAS: river analysis system, user’s manual.
Sensing, 54(11): 1593–1600. U.S. Army Corps of Engineers, Hydrologic Engineering Center,
Shamsi, U.M. 2002. GIS tools for water, wastewater, and storm- Davis, Calif.
water systems. ASCE Press, Reston, Va. Yang, J. 2004. Applying HEC-RAS and GIS techniques in river
Tate, E.C. 1999. Mapping using HEC-RAS and ArcView GIS. floodplain delineation. M.Sc. thesis, Department of Civil Engi-
M.Sc. thesis, Department of Civil Engineering, University of neering, University of Ottawa, Ottawa, Ont.
Texas at Austin, Tex.