Lecture given at the Ferdowsi University of Mashhad

1. Salvatore Manfreda and Caterina Samela
1Università degli Studi della Basilicata, Potenza, 85100, Italy.
E-mail address: salvatore.manfreda@unibas.it
Water Authority
Mashhad (Iran), 16 October 2018
LARGE SCALE FLOOD MAPPING
USING GEOMORPHIC METHODS
Salvatore Manfreda and Caterina Samela
1Università degli Studi della Basilicata, Potenza, 85100, Italy.
E-mail address: salvatore.manfreda@unibas.it
Water Authority
Mashhad (Iran), 16 October 2018
LARGE SCALE FLOOD MAPPING
USING GEOMORPHIC METHODS

2. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
FLOOD HAZARD
Many of these countries
are considered least
developed or developing.
More people are affected by floods than by any other type of natural disaster.
Recent analysis (World Resources Institute, 2016) shows that approximately 21 million people
worldwide are affected by river floods each year on average.

3. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
Hydrological and hydraulic models are the best approach for deriving detailed
inundation maps, but they require:
 large amounts of money and time
 significant amount of data and parameters not available for all areas.
Poor density of gauging
stations in some regions
(Asia, Africa, Australia).
(Herold and Mouton,
2011)
Flood risk assessment in data poor environments poses a great deal of challenge.
MOTIVATION: WHY SIMPLIFIED PROCEDURES?

4. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
GEOMORPHIC PROCEDURES
Basin morphology contains an extraordinary amount of information.
Research questions:
1) Does exist a physical attribute of the surface able to reveal if a portion
of a river basin is exposed to flooding?
2) Is it possible to use such descriptor to map the flood exposure over
large scale/unstudied areas?
Aim:
Develop a practical and cost-effective way to efficiently characterize
floodplains in non studied areas using readily available data.

5. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
LINEAR BINARY CLASSIFICATION
 SRTM: NASA Space Shuttle Radar Topography Mission, 30-meter digital
elevation model (download: https://earthexplorer.usgs.gov/)
 ASTER GDEM: METI (Japan) and NASA Advanced Spaceborne Thermal Emission
and Reflection Radiometer Global Digital Elevation Model (download:
https://earthexplorer.usgs.gov/).
 JAXA’s Global ALOS 3D World: Japan Aerospace Exploration Agency’s (JAXA)
Global Advanced Land Observing Satellite “DAICHI” (ALOS), 30-meter resolution
(download: https://www.eorc.jaxa.jp/ALOS/en/aw3d30/)
DEM (SRTM) of the Upper Tevere River Basin at 90m of resolution.
INPUT DATA
1. Digital representations of the terrain morphology (DEMs)
Data source: remote sensed elevation datasets available
at moderate resolution over the entire globe:
2. Standard Flood hazard maps
Data source: Flood inundation maps obtained by
hydraulic simulations.
A linear binary classification is performed to divide the basin cells in flood-prone and not
prone to flood areas. Linear classifiers separate the data sets using a “linear boundary”.
(Degiorgis et al., 2012; Manfreda et al., 2014)

6. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
TESTS
11 DEM-derived morphologic descriptors have been turned
into linear binary classifiers and tested, to analyse the
sensitivity of the classifiers to changes in the input data in
terms of:
i. DEM resolution;
ii. standard flood maps adopted (1-D or 2-D hydraulic model);
iii. dominant topography of the training area;
• Are the thresholds identified for the various indices related to basin
characteristics, such as topography?
• How robust are the thresholds and how transferrable?
iv. size of the training area;
• What percentage of a basin's area requires calibration?

7. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
Ethiopia, AFRICA
(Samela et al., 2015)
8 river basin in (ITALY):
• Tiber
• Chiascio
• Basento
• Cavone
• Agri
• Sinni
• Noce
• Bradano
TESTING THE RELIABILITY IN DIFFERENT CONTEXTS
United States of
America
(Samela et al., 2017)
(Manfreda et al.,
2015)

8. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
BEST PERFORMING CLASSIFIER: THE GFI
(A)
hr ≈ aAr
n
H
D
(B)
Location under exam
Nearest element of
the river network
along the flow path
Flow path
Example of a hydraulic cross-section with the description of the parameters:
• H is the difference in elevation from the point under exam and the closest element of
the river;
• the river depth hr (‘r’ stands for river) is calculated for each basin location as a
function of the upslope contributing area (in the section of the river network
hydrologically connected to the point under exam)using a hydraulic scaling
relationship proposed by Leopold and Maddock (1953) : hr ≈ 𝒂𝑨 𝒓
𝒏
;
Geomorphic
Flood Index, GFI ln(hr/H)

9. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
BEST PERFORMING CLASSIFIER: THE GFI
Performance less sensitive to:
 the different topography of the training area.
 the size of the calibration area;
 the resolution of the adopted Digital Elevation Model (DEM);
 the reference hydraulic map used for calibration.
The geomorphic procedure proposed can be a useful tool to have a preliminary but efficient
delineation in unstudied areas with simple data requirements, low costs and reduced
computational times.
This kind of simplified approach is generally of high interest to both researchers and decision-
makers since increasing portions of the population affected by flooding live in developing countries
where data availability is often poor.
Geomorphic
Flood Index ln(hr/H)

10. Pictorial representation of the 100 yr flood-
prone areas for the continental U.S.
according to the linear binary classifier
based on the GFI.
The large-scale map allows to see that the
index produces a realistic description of the
flood prone areas, with the possibility to
extend the flood hazard information in
those portions of the Country where the
FEMA’s maps are lacking (grey areas).
EXAMPLE
OF APPLICATION

11. The plugin can bee downloaded for free from the
QGIS repository and also from:
https://github.com/HydroLAB-UNIBAS/GFA-
Geomorphic-Flood-Area.
GEOMORPHIC FLOOD AREA (GFA) TOOL

12. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
GFA-TOOL ALGORITHM
DEM
Comparison:
Potential Flood-prone areas VS Standard flood hazard map
Standard
Flood Hazard map
Optimal threshold τ*:
minimizing the sum of the errors RFP + RFN
Geomorphic Flood
Prone areas
INPUT
DATA
Computes the GFI
Derive the
normalized GFI
Detection of areas above
a given threshold τ
OUTPUT
DATA
GFI, H, S, performance
measures
GFA TOOL ALGORITHM

13. GFA TOOL
EXAMPLE OF APPLICATION: Danube river sub-basin, ROMANIA

14. GFA TOOL

15. GFA TOOL

16. GFA TOOL

17. GFA TOOL

18. GFA TOOL

19. GFA TOOL

20. GFA TOOL

21. GEOMORPHIC FLOOD INDEX MAP

22. CALIBRATION MAP

23. GEOMORPHIC FLOOD AREAS MAP

24. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
WATER DEPTH ESTIMATE
In addition to flood extent, the inundation depth is a key factor in many
riverine settings for estimation of flood induced damages.
The years of research about the relationships between geomorphic and
hydraulic factors and the good performances obtained by the GFI method
led us to further exploit the potential of geomorphic features to obtain an
approximate, but immediate, estimate of the water surface elevation in a
river and surrounding areas.
Nepal, July 2018 Japan, July 2018

25. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
WATER DEPTH ESTIMATE
Geomorphic Flood Index:
𝑮𝑭𝑰 = 𝒍𝒏
𝒉 𝒓
𝑯
= 𝒍𝒏
𝒂𝑨 𝒓
𝒏
𝑯
= 𝒍𝒏 𝒂 + 𝒏 𝒍𝒏(𝑨 𝒓) − 𝒍𝒏 𝑯
Hydraulic scaling relationship (Leopold and Maddock, 1953):
𝒉 𝒓 ≈ 𝒂𝑨 𝒓
𝒏
 The hydraulic scaling might be difficult to calibrate, since it requires streamflow
observations and paired values of (ℎ 𝑟, 𝐴 𝑟) from a number of gauging stations sufficient
to carry out a linear regression.
 Results of linear binary classification are not influenced by the parameter a of the scaling
relationship. Therefore, in case calibration is not possible, the exponent n may be
estimated using literature values and assuming the parameter a equal to one.
 This implies that the ln(a) will be included in the computed GFI and therefore its value
will be incorporated in the calibrated threshold.

26. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
Along the calibrated boundary of decision:
𝑙𝑛
ℎ 𝑟
𝑎𝐻
= 𝜏 and
ℎ 𝑟
𝐻
= 1;
1
𝑎
= exp(𝜏) → 𝑎 =
1
𝑒𝑥𝑝(𝜏)
In this way, we can correct the values of river stage depth 𝒉 𝒓 = 𝒂𝑨 𝒓
𝒏
.
WATER DEPTH ESTIMATE
In this case, the obtained GFI’ may be expressed as:
𝐺𝐹𝐼′ = 𝐺𝐹𝐼 − 𝑙𝑛(𝑎) = 𝑙𝑛
𝐴 𝑟
𝑛
𝐻
= 𝑙𝑛
ℎ 𝑟
𝑎𝐻

27. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
WATER DEPTH ESTIMATE
(A)
hr ≈ aAr
n
H
WD
(B)
Location under exam
Nearest element of the river network along the flow path
Flow path
At this point, we can use the hr values to estimate, in a simple and direct way, the water
depth (WD) cell by cell of the flood-prone areas:
𝑾𝑫 = 𝒉𝒓
– 𝑯
Schematic description of the parameters used to derive the GFI and the water level depth
estimated in a hypothetical cross-section:

28. CASE STUDY: BRADANO RIVER, ITALY (OUTLET)
HYDRAULIC MODEL VS GEOMORPHIC APPROACH

29. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
CASE STUDY: BRADANO RIVER (ITALY)
FLOOD EXTENT PERFORMANCES
τ RTP RFN RTN RFP RFP + RFN AUC
-0.423 0.885 0.115 0.902 0.098 0.213 0.941
Table 1. Results of the linear binary classification based on the GFI.
INUNDATION DEPTH PERFORMANCES
Comparison over the
1D-domain of the
hydraulic simulation
Comparison within the
2D-domain of the
hydraulic simulation
Linear correlation
coefficient, r
0.917 0.906
Root Mean Square
Error, RMSE (m)
0.620 0.335
Table 2. Summary of the performance measures calculated comparing water depths
estimated for the mouth of the Bradano River basin (Italy) using the Geomorphic Flood
Index method and the FLORA-2D hydraulic model.

30. HYDRAULIC MODEL VS GEOMORPHIC APPROACH

31. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
FLOOD RISK AS THE CONJUNCTION OF HAZARD AND VULNERABILITY
As next step, we are working with stage – damage functions, which relate inundation
depth to damage level.

32. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
CONCLUSIONS
The great advantage of this methodology is the possibility to extend flood hazard
studies over large areas starting from the study of a small portion of a basin.
Procedures characterized by low cost and simple data requirement can be very
useful for local authorities and planners to realize effective management strategy.
33
100-year flood inundation map
of Romania derived using the
GFI approach.

33. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
FLOOD RISK
34

34. SALVATORE MANFREDA “Large scale flood mapping using geomorphic methods”
16 October 2018 | Mashhad (Iran)
RELATED PUBLICATION
1. Samela, Albano, Sole, & Manfreda (2018). A GIS tool for cost-effective delineation of flood-prone areas. Computers,
Environment and Urban Systems.
2. Samela, Troy, & Manfreda, (2017). Geomorphic classifiers for flood-prone areas delineation for data-scarce
environments. Advances in Water Resources.
3. Manfreda, Samela, Troy, (2018). The use of DEM-based approaches to derive a priori information on flood-prone
areas, in Flood monitoring through remote sensing, Springer Remote Sensing/Photogrammetry, 61-79.
4. D’Addabbo, Refice, Capolongo, Pasquariello, Manfreda, (2018). Data fusion through Bayesian methods for flood
monitoring from remotely sensed data, in Flood monitoring through remote sensing, Springer Remote
Sensing/Photogrammetry.
5. Samela, Manfreda, Troy, (2017). Dataset of 100-year flood susceptibility maps for the continental U.S. derived with
a geomorphic method. Data in Brief.
6. D’Addabbo, Refice, Capolongo, Pasquariello, Manfreda, (2018). Data fusion through Bayesian methods for flood
monitoring from remotely sensed data, in Flood monitoring through remote sensing, Springer Remote
Sensing/Photogrammetry.
7. Samela, Manfreda, De Paola, Giugni, Sole, & Fiorentino, (2016). Dem-based approaches for the delineation of flood
prone areas in an ungauged basin in Africa, Journal of Hydrologic Engineering.
8. Manfreda, Samela, Gioia, Consoli, Iacobellis, Giuzio, Cantisani, & Sole, (2015). Flood-Prone Areas Assessment Using
Linear Binary Classifiers based on flood maps obtained from 1D and 2D hydraulic models, Natural Hazards.
9. Manfreda, Nardi, Samela, Grimaldi, Taramasso, Roth, Sole (2014). Investigation on the Use of Geomorphic
Approaches for the Delineation of Flood Prone Areas, Journal of Hydrology.
10. Manfreda, Samela, Sole & Fiorentino (2014). Flood-Prone Areas Assessment Using Linear Binary Classifiers based
on Morphological Indices, ASCE-ICVRAM-ISUMA 2014.
11. Di Leo, Manfreda, Fiorentino, (2011). An automated procedure for the detection of flood prone areas:
r.hazard.flood, Geomatics Workbooks n. 10 - "FOSS4G-it: Trento 2011".
12. Manfreda, Di Leo, & Sole, (2011). Detection of Flood Prone Areas using Digital Elevation Models. Journal of
Hydrologic Engineering.