1. Effect of DEMs generated from different
sources on flood inundation mapping
Presented by:
Banamali Panigrahi (11AG62R16)
Under the Guidance of
Dr. C. Chatterjee
Agricultural and Food Engineering Department
Indian Institute of Technology
Kharagpur ,721 302
3. Introduction
Odisha is one of the major flood affected states of India.
Flood inundation mapping plays a major role in conveying
flood risk information.
One of the major issues in producing accurate flood inundation
maps is uncertainty.
Flow estimation from hydrologic model
Input data
Modeling type
Model set up and assumption
Model parameters
Lack of data
From the input data, topography is a key factor which affects
flood inundation mapping
4. Objectives
To compare the errors in elevation arising
from different sources of DEMs, and
To study the effect of DEMs generated from
different sources on flood inundation mapping
5. Study Area
Balianta Gauge
Devi Station
i
ha
ak
Ku
Salient Features Prachi
Lat:19˚ 50'38.53˝-20˚ 26'21˝
Long:85˚ 51̍ 5.62-86˚ 1̍ 12.61˝
̎ Daya Nimapada
Study
Boundaries extent: Gauge Station
N-Jagatsinghpur
S-Bay of Bengal & Area
Ganjam District
E-Khurda Ratnachira
W-Bay of Bengal
Areal Extent: 520 Km2
Kushabhadra
Mean Elevation: 10 m Bhargavi
Figure: Drainage network of Study Area
6. Data Used
Satellite Stereo pair
CartoSat-1(Image paired Captured in 2008 and 2009)
Field Survey Data
Differential Global Positioning System (DGPS) data
Cross-section Survey data (Superintendence Co. (P) Ltd.)
DEM Data
90m- Shuttle Radar Topography Mission (SRTM)
[http://srtm.csi.cgiar.org]
30m-Advanced Space borne Thermal Emission and
Reflection (ASTER) [https://earthexplorer.usgs.gov]
30m-CartoSat-1 [http://bhuvan.nrsc.gov.in]
Discharge and Water level Data
Balianta & Nimapada Gauge Station
7. DGPS Survey Elevation Data
Nabatia
Banamali
pur
Haripur
Balanga Nimapada
Suanda
Gop
Chandanpur
Figure : Ground control points on the CARTOSAT-1 satellite data for respective base
stations
8. Cross-section Survey data
( By Superintendence Co. (P) Ltd.)
90 cross-section
survey was carried out
in the entire study area.
•Kuakhai-13 Cs
•Bhargavi-26 Cs
• Daya-6 Cs
• Kushabhadra-45 Cs
Cumulative chainage
length and length from
left bank to right bank
was surveyed.
All the data sets are
available in hard copy as
well as soft copy.
Cs profile of rivers are
available in auto-CAD
format.
Figure : Cross-section points on the CartoSat-1 satellite data
9. Methodology
Satellite Data (Stereo pair) DGPS Survey Data
Center R3 Software
Trimble Business
Block Development
Post Processing of Field Data
Import Image
Datum Transfer
Block Adjustment
Ortho-rectification
Addition of GCP
River
Triangulation Network
Boundary
Block Processing (DTM
Conditions Comparison of
Generation) Modeling Water level Extent
of flow for different
DEM Generation Cross in river
Sections DEMs
Validation of DEM
Hydrodynami
c Parameters
Figure: Flowchart for Methodology
10. Results and Discussion
(i) DEM Generation
(i) Hydrodynamic Modeling
Table: Statistical Analysis for Elevations of DEMs and DGPS data for Study Area
Statistical GPS Google ASTER DEM SRTM BHUBAN
Parameters elevation earth elevation DEM CartoSat-1
elevation elevation elevation
No. of locations 122 122 122 122 122
MIN (m) 03.63 09.00 05.00 07.00 -61.00
MAX (m) 16.60 19.00 37.00 20.00 -37.00
MEAN (m) 09.17 13.63 13.60 13.68 -46.32
SD (m) 03.13 02.70 05.70 02.94 04.65
SEM (m) 00.28 00.24 00.50 00.26 00.42
RMSE (m) 04.98 05.50 04.53 55.53
11. Figure. Comparison of elevation of DGPS points with elevation of Google earth, SRTM
and ASTER DEM
12. Comparison of derived CartoSat-1 DEMs with available SRTM & ASTER DEM
(a) (b) (c)
(d)
(e)
Figure: 30m Generated CartoSat-1 DEMs for (a) DGPS, (b) Reduced Google Earth, (c) Google Earth,
and available DEMs of (d) 90m SRTM & (e) 30m ASTER
13. Quantitative Analysis of CartoSat-1 DEMs for Floodplain
Table: Error Analysis of Generated CartoSat-1 DEMs Elevation and DGPS Survey data
for Study Area
Statistical DGPS DGPS Reduced Google Google earth CartoSat-1
Parameters survey Cartosat-1 DEM earth CartoSat-1 DEM elevation
elevation elevation DEM elevation
MIN (m) 03.62 02.89 01.86 04.69
MAX (m) 16.60 16.89 16.82 20.82
MEAN(m) 10.08 09.97 10.25 12.76
SD (m) 03.09 02.89 02.76 02.82
RANGE (m) 12.98 13.99 14.95 16.13
Table: Analysis of Discrepancies (absolute values) between DGPS Survey and Generated
CartoSat-1 DEMs Elevation Data for Study Area
Statistical Parameters DGPS Cartosat-1 DEM Reduced Google earth Google earth CartoSat-1
elevation CartoSat-1DEM elevation DEM elevation
MIN (m) 00.00 00.00 00.00
MAX (m) 08.97 08.22 12.57
MEAN (m) 01.58 02.25 04.75
SD (m) 02.02 02.12 02.71
RANGE (m) 08.97 08.21 12.57
14. Quantitative Analysis of CartoSat-1 DEMs for River Bed
Table: Analysis of Cross-section Survey Elevation and Elevation of Generated CartoSat-1
DEMs for Study Area
Statistical Survey Cross DGPS Reduced Google Google earth
Parameters section elevation Cartosat-1 earth CartoSat-1 CartoSat-1 DEM
DEM elevation DEM elevation elevation
Number of 1189 1189 1189 1189
locations
MIN (m) -06.03 -01.91 -03.21 -03.26
MAX (m) 20.40 21.23 22.27 26.27
SD (m) 05.20 04.80 04.70 05.19
MEAN (m) 08.03 08.25 08.80 12.57
RANGE (m) 26.44 23.15 25.48 29.54
SEM (m) 00.15 00.13 00.13 00.15
Table: Analysis of Discrepancies (absolute values) between Cross-section Survey Elevation
and Elevation of Generated CartoSat-1 DEMs for Study Area
Statistical Parameters e DGPSCartosat-1 DEM Reduced Google Google earth
elevation earth CartoSat-1 DEM elevation
DEM elevation
Number of locations 1189 1189 1189
MIN (m) 00.00 00.02 00.01
MAX (m) 13.47 15.07 18.85
SD (m) 02.46 02.68 03.64
MEAN (m) 02.99 03.50 05.28
RANGE (m) 13.47 15.04 18.83
SEM (m) 00.07 00.07 00.10
15. Comparative Analysis of DEMs generated from different
sources
Table: Error Analysis Generated CartoSat-1 DEMs with SRTM and ASTER DEMs for Study
Area.
Sources of DEMs Floodplain River bed
RMSE MAE RMSE MAE
DGPSCartoSat-1 DEM 1.65 1.04 3.56 2.76
Reduced Google earth CartoSat-1
DEM 2.94 2.04 4.41 3.50
Google earth CartoSat-1 DEM 5.46 4.75 6.41 5.28
SRTM DEM 4.89 3.93 4.41 3.50
ASTER DEM 6.31 4.21 8.18 5.81
16. Qualitative Analysis of DEMs generated from different DEM
sources
1:1 Line 1:1 Line 1:1 Line
1:1 Line
1:1 Line
Figure: Scatter plots of floodplain elevation for different DEMs
17. Qualitative Analysis of DEMs generated from different DEM
sources
1:1 Line 1:1 Line
1:1 Line
1:1 Line
1:1 Line
Figure: Scatter plots of river bed elevation for different DEMs
18. Comparison of derived CS from different DEMs with Survey CS
Figure :Cross-section Profile for Bhargavi River
19. Comparison of derived CS from different DEMs with Survey CS
Figure :Cross-section Profile for Kushabhadra River.
20. Simulation setup of MIKE-11
Figure :Generated data bases for Kushabhadra river
system
21. Simulation setup of MIKE-11
H-point (blue) Cross section (C/S)
Q-point (Red ) System
generated point at the
mid of two Cross section
(C/S)
Figure : Locations of H-point (Cross section) and Q-point (H-Q relation can be obtained)
22. Calibration of Kushabhadra River System
Table : Error function values for Nimapada gauging station during calibration for the year
2003 for MIKE 11.
Station Name Nash Sutcliffe RMSE MAE R2
Coefficient (NSC)
Nimapada 0.91 0.62 0.43 0.91
Figure: Comparison of predicted and observed discharge at Nimapada during
calibration for the year 2003 for MIKE 11.
23. Validation of Kushabhadra River System
Table: Error function values for Nimapada gauging station during validation for
the year 2004 and 2005 for MIKE 11
Year of Nash Sutcliffe
RMSE MAE R2
Validation Coefficient (NSE)
2004 0.89 0.58 0.44 0.89
2005 0.88 0.70 0.55 0.88
Figure: Comparison of predicted and observed discharge at Nimapada during
validation for the year 2004 and 2005
24. Conclusions
Generated cartoSat-1 DEMs give better representation of elevation
of terrain than available ASTER and SRTM DEMs.
Cartosat-1 DEM is derived using DGPS points show better result
followed by reduced Google earth, Google earth, SRTM and
ASTER DEMs.
Kushabhadra river system for MIKE 11 is well validated for
Manning's roughness 0.0265.
25. Work to be Done
Calibration and validation of Kuakhai-Bhargavi river system for
MIKE-11.
Comparison of water level extent of river for different sources of
DEMs.
Quantification of effects of DEM on flood inundation modeling.
27. Review of Literature
Citation Major findings
Werner Estimating flood extent maps using a simple Inverse Distance
(2001) Weighted (IDW) interpolation easily avoids the local
depressions which are not directly connected to the main
channel.
Merwade et al. Locating the channel centerline along the thalweg , is a
(2005) reference for assigning s ,n-coordinates to the bathymetric
data. The resulting bathymetric data in the s ,n, z-coordinate
system are used to create a square mesh or FishNet.
Merwade et al. The variable-direction of the river channel bathymetry is
(2006) accounted for by using a flow-oriented curvilinear coordinate
system to establish a unidirectional flow channel.
Gorokhovich Absolute average vertical errors from CGIAR dataset is
&Voustianiouk significantly better than a standard SRTM by considering the
(2006) slope and aspect where the slope values greater than 10°.
28. Cont……
Citation Major findings
Wilson et al. Model accuracy is good at high water, while accuracy drops at
(2007) low water due to incomplete drainage of the floodplain resulting
from errors in topographic data and omission of floodplain
hydrologic processes from this initial model.
Merwade et al. Creating surface representations of river systems is a challenging
(2008a) task because of issues associated with interpolating river bathy-
metry and then integrating this bathymetry with surrounding
topography.
Merwade et al. It is unknown how the uncertainties associated with topographic
(2008b) representation, flow prediction, hydraulic model, and inundation
mapping techniques are transferred to the flood inundation map.
Cook and Providing an improved understanding of interplay among
Merwade topography, geometric description and modelling approach in
(2009) the final inundation mapping, showed that the flood inundation
area reduces with improved horizontal resolution.
29. Cont……
Citation Major findings
Getirana et al. Proposed a new approach based on ‘burning’ concepts to obtain
(2009a) better defined watershed delineation in floodplains of large
basins by using the spatial distribution of flooded areas from
satellite images.
Getirana et al. To overcome D8 algorithm failure, a double burning method,
(2009b) known as floodplain burning (FB) method has been proposed.
The proposed method introduces five coefficients requiring
adjustment in order to obtain a relevant watershed delineation
but minimizing DEM changes
Vaze et al. Demonstrates that the loss of details by re sampling the higher
(2010) resolution DEM to coarser resolution are much less compared to
the details captured in the commonly available coarse resolution
DEM derived from contour maps
Paz et al. Hydro-dynamic models require river cross-sectional profiles that
(2010) must comprise both main channel and floodplain in order to
better represent the river hydraulics, the floodplain commonly
being several times larger than the main channel when dealing
with large rivers.
30. Cont……
Citation Major findings
Alsdorf et al. Basically in flat river basins like the Amazon, surface waters are
(2010) actively exchanged between river channels and floodplains
Yamazaki et Floodplains have a key role as natural retarding pools which
al. (2011) attenuate flood waves and suppress flood peaks in large basins
Yamazaki et The flow connectivity is ensured by new Burning algorithm
al. (2012) which is found to be essential for representing realistic water
exchanges between river channels and floodplains in hydro-
dynamics modeling.
Jung and Topography plays a significant role in hydraulic modeling used
Merwade to derive water surface elevations corresponding to the design
(2012) flow, and the geometry that defines the flow domain, including
river cross sections and bathymetry mesh in a hydraulic model.