geomorphic approaches for the delineation of flood prone areas

1/37 2014 Junior Enrico Marchi Lecture Dipartimento di Ingegneria – ROMA TRE Roma – 23 May 2014

Salvatore Manfreda

Università degli Studi della Basilicata

[email protected]

Geomorphic Approaches for the Delineation of Flood Prone Areas


Collaboration

The present activity have involved a significant number of colleagues that I would like to acknowledge

n  Domenico Capolongo, n  Francesco De Paola, n  Margherita Di Leo, n  Vito Iacobellis, n  Mauro Fiorentino, n  Andrea Gioia, n  Maurizio Giugni, n  Salvatore Grimaldi, n  Fernando Nardi, n  Alberto Refice, n  Giorgio Roth, n  Caterina Samela, n  Aurelia Sole, n  Angela Celeste Taramasso, n  Tara Troy.


Flood Exposure at the Global Scale

Flooding is evident in more than 1/3 of the world’s land area, in which some 82% of the world’s population resides. Dilley et al. (2005)


Flood Monitoring

Rela5vely poor density of gauging sta5ons in some regions, such as South America, Asia and Africa.

Herold and Mouton (HESSD, 2011)


Are Natural Hazard increasing? There is a significant increase of the economic losses for weather related events

Flooding 21%

CuJer and Emrich (EOS-‐ 2005)


Number of Reported Events Much of the increase in number may be due to improvements in information access, but the rising of flood and cyclones events is dramatic compared to others.

Peduzzi (2005)


Flood Mapping

ü There is a significant effort in flood mapping at different scales collecting all available information (e.g. Dilley et al., 2005; Moel et al., 2009) or using large scale physically based models of rainfall-runoff and river routing (e.g. Pappenberger et al., 2012; Winsemius et al., 2013).

ü The full mosaic is not available yet, but it may be extremely useful in reinsurance, large scale flood preparedness and emergency response (e.g. Kappes et al., 2012).


Flood Mapping: Limitations

Current methods or products have some limita5ons

•  Extent of the study area •  Reference basin scale •  Spa5al resolu5on adopted •  Time consuming computa5on

•  Calibra5on problems

Pappenberger et al.: Deriving global flood hazard maps (HESS -‐ 2012)


River basin morphology intrinsically contains an extraordinary amount of informa5on on flood-‐driven erosion and deposi5onal phenomena, cons5tu5ng a useful indicator of the flood exposure of a given area (e.g. Arnaud-‐FasseJa et al., 2009; Tucker et al., 2001; Tucker and Whipple, 2002)

Geomorphic Approaches

Flood Plain


Geomorphic Approaches

ü Following this theoretical principle, several authors have shown that the delineation of flood prone areas at the large scale can be carried out using simplified methods that rely on basin geomorphologic feature characterization (e.g., McGlynn and Seibert, 2003; Gallant and Dowling, 2003; Dodov and Foufoula-Georgiou, 2006).

ü The advent of new technologies to measure topographic surface elevation (e.g., GPS, SAR, SAR interferometry, and laser altimetry) has given a strong impulse to the development of geomorphic approaches for valley bottoms identification using Digital Elevation Models (DEMs).


Digital Elevation Models

ü The increasing availability of digital terrain models has given a strong impulse to the development of so called distributed and DEM-based models. ü Digital terrain model obtained through interferometric data gathered by the space shuttle campaign by NASA with a cell-size of 90m. (CGIAR-CSI: http://srtm.csi.cgiar.org/) ü ASTER GDEM 30m available from June 2009 (http://asterweb.jpl.nasa.gov/gdem.asp )


Research Questions

i)  What are the geomorphological signatures useful for the delineation of flood prone areas? ii)  Is it possible to define a simplified approach for the delineation of flood prone areas starting from DEMs? iii)  is it possible to use such procedure to map the flood exposure over large scale? iv)  What is the impact of scale/resolution on Geomorphic Approaches?


Geomorphic Methods

ü Three different geomorphic approaches to the identification of flood prone areas are investigated.

ü The selected algorithms are: n  GM1 - modified Topographic Index (TIm) (Manfreda et al., 2011);

n  GM2 - linear binary classifier applied on different geomorphic features related to the location of the site under exam with respect to the nearest hazard source (Degiorgis et al., 2012);

n  GM3 - hydro-geomorphic method by Nardi et al. (2006) simulating inundation flow depths along the river valley with the associated extent of surrounding inundated areas.


GM1 Modified Topographic Index

Stream line

Contour line

The concept of contributing area per unit contour length

This index is commonly used to quan5fy topographic control on numerous hydrological processes, but it is also a good indicator of flood-‐prone areas, as recently highlighted by Manfreda et al. (2011) and Jalayer et al. (2014).

( ))tan(/ln βna

(Beven and Kirkby, 1979)


GM2: The linear binary classifiers

GM2 identifies areas subject to the flooding hazard through pattern classification techniques using five morphologic features: ü 1. the contributing area, A [m2]; ü 2. the surface curvature, ∇2H [-]; ü 3. the local slope, S [-]; ü 4. the distance of each cell from the nearest stream, D [m]; ü 5. the relative elevation to the nearest stream, H [m].


GM3 - Hydro-geomorphic method

The inundation model is defined as a function of the hydrologic characteristics of a predefined design flood event designed based on the flow peak discharge at the basin outlet.

ü GM3 is an automated GIS-based procedure, implementing a set of terrain analysis algorithms for flooded area delineation by linking a simplified inundation model with the geomorphic properties of the stream network (Nardi et al., 2006; Nardi et al., 2013).


Schematic description of the different algorithms analysed


Some Terms

FALSE POSITIVE

FALSE NEGATIVE


Accuracy, sensitivity, specificity

sensitivity (rtp) = true positive fraction = 1 – false negative fraction = TP / (TP + FN)

specificity (rtn) = true negative fraction = 1 – false positive fraction = TN / (TN + FP)

accuracy = (TP + TN) / (TP + TN + FP + FN)


Case Study: The Upper Tiber River

Alluvial Plain DEM Flood Map

Upper Tiber Basin 5000 km2

Chiascio River 727 Km2


Flood Map used for Calibration

ü The “Piano di Assetto Idrogeologico” or PAI developed by Tiber River Basin Authority (TRBA) contains flood hazard maps based on detailed standard hydrologic and hydraulic models (TRBA PAI, 2010).

ü The TRBA PAI was developed using high precision bathymetric surveys of the channel surveyed as cross sections with average spacing interval of 200-400 meters. This detailed fluvial morphology was used as main input of a 1D hydraulic models (HEC-RAS and FRESCURE). simulating the effect of the design hydrographs considering return periods of 50, 200, and 500 years.


Results: GM1

n=0.020; τ=3.1


Results: GM2 Single Features


Results: GM2

ü ROC curves




Results: GM3


Flood Maps according to the three considered methodologies


Visual comparison of the performances of the GM1, GM2 and GM3

Zone 2

Zone 1

Zone

2

Zone

1


Model Intercomparison

Mod.

Topographic Index

Single-‐feature binary classifier

(H)

Hydrogeomorphic Method

Ideal value

True posiHve rate, 90.2% 90.1% 60.1% 100% False negaHve rate, 9.8% 9.9% 39.9% 0% True negaHve rate, 51.4% 81.2% 97.2% 100% False posiHve rate, 48.6% 18.8% 2.8% 0%

rfp + rfn 58.4% 28.7% 42.7% 0%

Mod.

Topographic Index

Single-‐feature binary classifier

(H)

Hydrogeomorphic Method

Ideal value

True posiHve rate, 93.8% 93.4% 75.8% 100% False negaHve rate, 6.2% 6.6% 24.2% 0% True negaHve rate, 61.3% 66.4% 94.3% 100% False posiHve rate, 38.7% 33.6% 5.7% 0%

rfp + rfn 44.9% 40.2% 29.9% 0%




DEM Resolution affects on morphological indexes

(Wood, 1996)


Scale Dependence of the Modified Topographic Index

The spatial distribution of the topographic index is inevitably linked to the cell-size of the adopted DEM (Zhang and Montgomery, 1994). This d e p e n d e n c e i s i n v e s t i g a t e d comparing the errors rfp and rfn ob ta ined us ing the mod i f i ed topographic index computed from DEMs with different cell-size. Analyses were carried starting from DEM with cell-size of 20m up to 720m.


Scale Dependence and Associated Errors

11 10

2 3

4 5

6

7 8

9

1

ARNO Sub-catchments

Less sensi5ve to the change of resolu5on


Future directions

New strategies for flood modelling calibration

ü Notwithstanding the limitation of the flood map adopted for comparison, it would be extremely useful to have flood map of real event.

ü In this respect, a possible contribution may arise from the remote sensing that is developing a number of applications for flood mapping using Synthetic Aperture Radar (SAR).


Future Direction: Synthetic Aperture Radar Remote Sensing

J.A. Richards, Remote Sensing with Imaging Radar, Springer, 2009

Basic principle: water-covered surfaces appear darker than non-water

ComplicaHng factors: o  speckle o  water:

§  wind; waves; vegeta5on… o  non-‐water:

§  crops; forest; urban,…

SAR intensity thresholding


An Example

(h(At)-H)/D

Inundated area Inundated area

Orthophoto


Conclusion

ü Geomorphic approaches represent a useful tool for preliminary studies on flood prone areas or to extend flood mapping over large areas.

ü These methods can be improved or integrated using available knowledge on limitations and prerogatives of different algorithms.

ü This approaches may benefit from advances in remote sensing techniques and vice versa.

ü Sensitivity of Geomorphic approaches is not influenced by cell-size resolution changes (from 20m-100m) or basin size. This open the perspective of using these methods for downscaling procedures in global flood mapping.


Related Publication Samela, C., S. Manfreda, F. De Paola, M. Giugni and M. Fiorentino, Dem-based approaches for the delineation of flood prone areas in an ungauged basin in Africa, Journal of Hydrologic Engineering, 2014.

Manfreda, S., F. Nardi, C. Samela, S. Grimaldi, A.C. Taramasso, G. Roth, A. Sole, Investigation on the Use of Geomorphic Approaches for the Delineation of Flood Prone Areas, Journal of Hydrology, 2014.

Manfreda, S., C. Samela, A. Sole and M. Fiorentino, Prone Areas Assessment Using Linear Binary Classifiers based on Morphological Indices, ASCE-ICVRAM-ISUMA 2014, 2014.

Manfreda, S. and Sole, A. ”Closure to “Detection of Flood-Prone Areas Using Digital Elevation Models” by Salvatore Manfreda, Margherita Di Leo, and Aurelia Sole.” Journal of Hydrologic Engineering, 18(3), 362–365, 2013.

Manfreda, S., M. Di Leo, A. Sole, Detection of Flood Prone Areas using Digital Elevation Models, Journal of Hydrologic Engineering, Vol. 16, No. 10, September/October 2011, pp. 781-790 (10.1061/(ASCE)HE.1943-5584.0000367), 2011.

Fiorentino, M., S. Manfreda, V. Iacobellis, Peak Runoff Contributing Area as Hydrological Signature of the Probability Distribution of Floods, Advances in Water Resources, 30(10), 2123-2144, 2007.

Manfreda, S., A. Sole, e M. Fiorentino, Valutazione del pericolo di allagamento sul territorio nazionale mediante un approccio di tipo geomorfologico, L'Acqua, n. 4, 43-54, 2007 (In Italian).

Manfreda, S., A. Sole, M. Fiorentino, Can the basin morphology alone provide an insight on floodplain delineation?, on Flood Recovery Innovation and Response, WITpress, 47-56, 2008.