Document Actions

IJEGE-11_BS-Horton-et-alii

background image
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
875
DOI: 10.4408/IJEGE.2011-03.B-095
FLOW-R, A MODEL FOR DEBRIS FLOW SUSCEPTIBILITY MAPPING
AT A REGIONAL SCALE – SOME CASE STUDIES
P
asCal
HORTON
(*)
, m
iCHel
JABOYEDOFF
(*)
, m
aRkus
ZIMMERMANN
(**)
,
b
enoit
MAZOTTI
(*)
& C
eline
LONGCHAMP
(***)
(*)
Institute of Geomatics and Risk Analysis, University of Lausanne, Lausanne, Switzerland
(**)
NDR Consulting GmbH, Thun, Switzerland
(***)
Nagoya University - Department of Civil Engineering - Nagoya 464-8601, Japan
INTRODUCTION
Physical modelling of debris flows in the frame-
work of regional mapping is difficult because of the
complexity of the phenomenon and the variability of
controlling factors. Flow-R (Flow path assessment of
gravitational hazards at a Regional scale) aims at giv-
ing a quick assessment of debris flows susceptibility
at a regional scale with minimum data requirement
(H
oRton
et alii, 2008). It identifies potential source
areas and delimits the zones in the path ofthe propa-
gation. The proposed method merges several existing
GISbased approaches, which makes it an evolution of
past ones (e.g. H
uGGel
et alii, 2003).
The three case studies provide examples of debris
flow susceptibility maps produced using the Flow-R
model in different contexts and based on very dif-
ferent datasets. Susceptibility mapping aims to give
an insight of existing or potential new susceptibility
zones. This is equally important in mountainous areas
of industrialized or developing countries
FLOW-R MODEL
Flow-R is a numerical model (compiled with Mat-
lab) that was first developed for the Canton de Vaud
case study. The characteristics of the software are (i)
limited requirement of datasets and (ii) customization
of the inputs, the algorithms and the parameters through
a graphical user interface. The model, originally devel-
oped for debris flows, has proved to be as well appli-
cable for other processes such as rockfall, floods and
ABSTRACT
Every year, debris flows cause huge damage
in mountainous areas. Due to population pressure
in hazardous zones, the socio-economic impact is
much higher than in the past. Therefore, the devel-
opment of indicative susceptibility hazard maps is
of primary importance, particularly in developing
countries. However, the complexity of the phenom-
enon and the variability of local controlling factors
limit the use of processbased models for a first as-
sessment. A debris flow model has been developed
for regional susceptibility assessments using digital
elevation model (DEM) with a GIS-based approach..
The automatic identification of source areas and the
estimation of debris flow spreading, based on GIS
tools, provide a substantial basis for a preliminary
susceptibility assessment at a regional scale. One of
the main advantages of this model is its workability.
In fact, everything is open to the user, from the data
choice to the selection of the algorithms and their
parameters. The Flow-R model was tested in three
different contexts: two in Switzerland and one in Pa-
kistan, for indicative susceptibility hazard mapping.
It was shown that the quality of the DEM is the most
important parameter to obtain reliable results for
propagation, but also to identify the potential debris
flows sources.
K
ey
words
: model, regional, debris flow, susceptibility map-
ping
background image
P. HORTON, M. JABOYEDOFF, M. ZIMMERMANN, B. MAZOTTI & C. LONGCHAMP
876
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
ity of material availability, would be excluded.
The primary dataset is the DEM. On this basis,
various morphological data can be derived. Three cri-
teria are considered as critical by R
iCkenmann
& z
im
-
meRmann
(1993), t
akaHasHi
(1981) and d
elmonaCo
et alii (2003): (1) sediment availability, (2) water input
and (3) gradient. The two last are directly derived from
the DEM, whereas (1) can be defined manually or using
DEM information such as curvatures
GRADIENT
The gradient is a decisive criterion for debris
flow susceptibility (t
akaHasHi
, 1981). Generally,
debris flows occur when the gradient of the bed is
above 15° (R
iCkenmann
& z
immeRmann
, 1993; t
aka
-
HasHi
, 1981). This is the default threshold in this
model.
wATER INPUT (FLOw ACCUMULATION)
The upslope contributing area, processed by
means of flow direction algorithms (see Section
“Flow Direction Algorithms”), can represent the
amount of water flowing through the cell. The flow
directions and flow accumulation is processed for
the whole DEM to produce a map of the values of
flow accumulation at every cell. It is used widely in
distributed hydrological models (t
aRboton
, 1997;
e
Rskine
et alii, 2006). A threshold that is related to
the slope angle is commonly used in Switzerland
(H
einimann
, 1998). Based on this threshold and the
observation of an extreme event (R
iCkenmann
& z
im
-
meRmann
, 1993), two thresholds were considered for
rare and extreme events in Switzerland (Fig. 2). Both
threshold curves are limited by the 15° slope angle
avalanches. The model and its various concepts have
been and are tested in various countries and contexts by
partners in France, Italy, Austria, Canada and Norway
(b
laHut
et alii, 2010; l
aRi
et alii, in review; k
aPPes
et
alii, in review; l
aRi
et alii, this volume).
PRINCIPLES
The modelling comprises two distinct steps: 1) the
identification of the potential source area and 2) the
assessment of the spreading (Fig. 1). First, the sources
are identified on the basis of the Digital Elevation
Model (DEM) and user-defined datasets. From these
sources mass points are propagated on the topography,
using a probabilistic and energy approach, respective-
ly (H
oRton
et alii 2008). The mass of the sources is
not taken into account because of the difficulties in
assessing volume at a regional scale and because of
significant mass changes occurring through deposition
and erosion (i
veRson
& d
enlinGeR
, 2001) which are
difficult to estimate.
IDENTIFICATION OF SOURCE AREA
The source areas are identified by means of a
combination of criteria based on the morphology of
the terrain and on user-defined datasets. Those layers
are processed to classify every cell as potential source,
or as an excluded or ignored area. The cell is consid-
ered as source if it was at least once considered as a
potential source, but never excluded (H
oRton
et alii,
2008). For example, a cell with considerable slope and
large water inflow, makes a good candidate for debris
flows, but having a bedrock that reduces the possibil-
Fig.1 - Scheme of the method (J
ABoyeDoff
et alii, in re-
view)
Fig 2 - River bed gradient in relation to upslope area:
threshold values for rare and extreme events.
After HORTON et al. (2008), based on data by
HEINIMANN (1998) and RICkENMANN & ZIM-
MERMANN (1993)
background image
FLOW-R, A MODEL FOR DEBRIS FLOW SUSCEPTIBILITY MAPPING AT A REGIONAL SCALE – SOME CASE STUDIES
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
877
tion methods (Q
uinn
et alii, 1991, f
Reeman
, 1991,
H
olmGRen
, 1994). Holmgren's algorithm (Eq. 3) is
frequently chosen in our case studies as it allows con-
trolling the spreading by means of an exponent, and
thus enables replacement of most of the other algo-
rithms. And the Holmgren algorithm fits the real cases
better. The higher the exponent, the more convergent
the flow becomes. When x = 1, it turns into the basic
multiple flow direction, and when x → ∞, it becomes
a single flow direction.
where i, j = flow directions (1..8) with angular inter-
vals of 45 degrees, f
si
= flow proportion (0..1) in direc-
tion i, tan(β
i
) slope gradient between the central cell
and cell in direction i and x the variable exponent.
Based on G
amma
(2000), inertia was implement-
ed under the form of persistence (Eq. 4), which is a
weighting function of the change in direction.
where i = flow directions (1..8), f
pi
= flow proportion
(0..1) in direction i, α
i
= angle between the previous
direction and the direction from the central cell to cell
i, w
0,45,90,135
= weights for the corresponding change in
direction.
The final probabilities (Eq. 5) are function of the
flow direction algorithm and the persistence (H
oRton
et alii, 2008):
where i, j = flow directions (1..8), fi = total flow pro-
portion (0..1) in direction i, f
si
= flow proportion from
the slope-related algorithm, f
pi
= flow proportion from
the persistence, f0 = previously determined flow pro-
portion of the central cell.
From a given source, the spreading is processed
once, integrating every possible path and attributing
given by t
akaHasHi
(1981)
The rare event threshold is given in Eq. 1:
where tan β
lim
= slope gradient, SUA = surface of the
upslope contributing area.
The extreme event threshold is given in Eq. 2:
SEDIMENT AVAILABILITY (CURVATURE)
Debris flows sources are found where curvature
is concave (d
elmonaCo
et alii, 2003; w
ieCzoRek
et
alii, 1997). The plan curvature, perpendicular to the
steepest slope derived from the DEM, provides iden-
tification of gullies that are considered as potential
sources.
OTHER DATASETS
Any other user-defined input can be taken into
account, should it be numerical or classified data. In
the first case, the selection is based on ranges, and in
the second, on values. This allows, for example, to
asses sediment availability using geological or litho-
logical information or land use data.
SPREADING ASSESSMENT
The assessment of the spreading area is based on
the one hand on a probabilistic spreading by means
of flow direction algorithms, and on the other hand
on a basic energy balance, which defines the maxi-
mal runout distance. Both are customizable and can
be combined according to the user choice.
FLOw DIRECTION ALGORITHMS
The flow direction algorithms allow distributing
the probabilities to the neighbouring cells, according
to a relationship using gradient. In contrast to its use
for water contribution in the source detection, the di-
rection of the debris flow is here processed step by
step for every cell. Various algorithms are implement-
ed, like the D8, D(x) (t
aRboton
, 1997), p8 (f
aiRfield
& l
eymaRie
, 1991) and different multiple flow direc-
(1)
(2)
(3)
(4)
(5)
background image
P. HORTON, M. JABOYEDOFF, M. ZIMMERMANN, B. MAZOTTI & C. LONGCHAMP
878
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
Where V
i
B
is the velocity at the end of the ith
segment, V
i
A
is the velocity at the beginning of the ith
segment α
i
= g (sin θ
i
- μ
i
cos θi ) , βi = -2L
i
/(M / D)
i with θ the slope angle of the segment, μ the coef-
ficient of friction, M/D the mass-to-drag ratio (m) and
L the horizontal distance of the segment.:
SIGNIFICANCE OF THE SIMPLE ENERGY
LOSS MODEL
The velocity limit, adapted from PERLA (1980),
for an infinite slope with a constant slope angle (θ)
and a friction coefficient μ = tan Φ, can be seen in
equation 9:
Fixing the mass to drag ratio, the velocity limit is
depending on the slope angle and the friction coeffi-
cient that can also be expressed by means of an angle
(Fig. 3). The Perla model and the constant friction mod-
probabilities to every cell. The spreading areas of
all sources are combined by keeping the maximum
probability values. It is, however, not a mathematical
probability in a strict sense, but it has to be to be in-
terpreted in a qualitative way (H
uGGel
et alii, 2003).
RUNOUT DISTANCE ASSESSMENT
Energy based algorithms are used to assess the ru-
nout distance. This constraint defines if a cell can be
reached by the debris flow, or if the actual energy of
the flow portion is not high enough. Thus,they control
the distance reached by the debris flow and in addition
may reduce lateral spreading.
As the source mass is unknown, the energy bal-
ance is processed on a unit mass (Eq. 6), which is
clearly simplistic compared to the complex real physi-
cal processes.
where i = time step, E
kin
= kinetic energy, ∆E
pot
=
change in potential energy and E
loss
= constant loss.
Two main algorithms are available for energy
balance assessment: a slope angle concept or single
parameter friction model, and a two parameters fric-
tion model.
The single parameter friction model assumes an
incrementaz energy loss by a friction coefficient using
where ∆x is the increment of horizontal displacement,
Φ
b
is the basal angle of friction linked to the friction
coefficient μ and g the acceleration due to gravity.
This algorithm is available in the model along with a
kinetic energy limit that aims at keeping a realistic ve-
locity of the propagation. This permits the use of very
simple arguments to run the model using the observed
mean slope angle of the path for which the debris flow
stops as the lower limit for Φ
b
, and the maximum ex-
pected velocity of the debris flow.
The two parameter friction model is described in
P
eRla
(1980). It was then established for avalanches,
but can be used for other hazards with the correspond-
ing parameters. In that case, the equation governing
the velocity is given by P
eRla
(Eq. 8) (1980)
(6)
(7)
(8)
(9)
Fig. 3 - Relationship between the slope angle and the
maximum velocity for different friction angles and
a mass to drag ratio of 30
Fig. 4 - Comparison of the two models of the velocity and
distance reached over time
background image
FLOW-R, A MODEL FOR DEBRIS FLOW SUSCEPTIBILITY MAPPING AT A REGIONAL SCALE – SOME CASE STUDIES
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
879
CASE STUDY OF THE CANTON DE VAUD,
SWITZERLAND
The Canton de Vaud (western Switzerland) (Fig.
6) susceptibility mapping was established at a scale
of 1:25000 on the entire territory. Amongst other haz-
ards, debris flows were studied along with mud flows
and hyperconcentrated flows (J
aboyedoff
et alii,
2008). Only debris flows are presented hereafter.
The state has an area of 2'822 km². The elevation
ranges approx. from 400 m to more than 3000 m a.s.l.
USED DATASETS
A 1 m resolution digital elevation model (DEM)
was available, derived from aerial laser scanning, with
an accuracy of 30 cm and a standard deviation of 5 cm
(MNT-MO Swisstopo factsheet, 2005). Because of the
time required for processing, the high resolution DEM
was scaled down to 10 m.
A “geotypes” map (t
uRbeRG
et alii, 2008), which
contains uniform and complete information about out-
cropping geological formations, was available for the
whole region. This map is based on the Swiss Atlas of
Geological maps 1:25’000 (swisstopo.ch). However, it
has some limitations for debris flow susceptibility as-
sessment, as it does not consider the tectonic origin of
the different rock types. This simplification does not
allow differentiating the disparity in fracture and the
weathering degree within the same rock type.
A land use map based on the product Vector 25
from SwissTopo (www.swisstopo.ch) was available
and has been considered to identify certain inaccurate
sources, that are located in built-up areas or that are
man-made infrastructures
SOURCES CRITERIA
The standard morphological data (slope, flow ac-
cumulation and curvature) were integrated with the
following parameters:
Slope threshold: 15°
Flow accumulation threshold: 1 ha
Flow accumulation – slope relationship: extreme
threshold
Curvature: -2/100 m-1
The lithology was considered by means of the
geotypes map. The selected lithologies are debris flow
prone rocks (marl, slate, siltstone) and slope deposits.
The land use was integrated to remove zones
located in built-up areas or that are man-made in-
el with limiting velocity were compared for an infinite
slope (Fig. 4). It can be shown that the behaviour of
both models provide very similar results for the same
μ, when the velocity limit for the second model is set to
V∞, making the velocity limit quite efficient.
The difficulty is to choose the velocity limit, be-
cause the slope angle changes along the flow path. But
from Fig. 4, it can be seen that the order of magnitude
can be compared for both models. It is clear that the re-
lationship between both models can be tuned by chang-
ing both μ and the max velocity value according to the
topographic profile.
MODEL OUTPUTS
The outputs of the model are the source areas, the
propagation kinetic energy and its probability, as illus-
trated on Fig. 5.
Fig. 5 - Illustration of the model outputs: possible sourc-
es, kinetic energy and spreading probability (J
A
-
BoyeDoff
et alii, in review
)
Fig. 6 - Resulting susceptibility map for the Canton de
Vaud. Red identifies debris flows and yellow,
hyperconcentrated flows and mud flows
background image
P. HORTON, M. JABOYEDOFF, M. ZIMMERMANN, B. MAZOTTI & C. LONGCHAMP
880
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
frastructures. Outcropping or suboutcropping rocks
were also excluded.
PROPAGATION PARAMETERS
Propagation parameters were taken from litera-
ture or calibrated on the basis of traces of old debris
flows visible on orthophotographs:
1 Holmgren’s exponent was set to 4, as established in
C
laessens
et alii (2005).
2 The energy loss function is the slope angle algo-
rithm (Eq. 7) with a value of 11° according to
H
uGGel
et alii (2002) for the probable maximum
runout.
3 The velocity threshold, considered along with the
slope angle
algorithm, is 15 m/s. It is based on the observed
maximum velocity of 13 to 14 m/s among various
debris flow events in Switzerland (R
iCkenmann
&
z
immeRmann
, 1993).
CASE STUDY OF THE BAGNES VALLEY,
SWITZERLAND
A susceptibility map was established with a
1:25000 scale for the Bagnes valley (Fig. 7). The
method is similar to the case study of the Canton de
Vaud, with improvements by means of field work to
check the accuracy of the sources. Four test sites in
the valley were chosen to calibrate the model param-
eters for the source areas detection and the propaga-
tion. This was made possible due to the small size of
the catchment. Then, source areas were divided into
two classes: proven and potential, resulting in two
propagation area types
The valley has an area of 282 km². The elevation
ranges from 600 m to 4200 m a.s.l.. The geology is
characterized by three main paleogeographical do-
mains of the Alps: Helvetic, Penninic and Austro-Al-
pine (t
RümPy
, 1980). The rocks are then diverse, from
some Cambrian polycyclic basements to Mesozoic-
Cenozoic sedimentary covers (s
aRtoRi
et alii, 2006).
Various schists can be found in some areas, as on the
Merdenson catchment, where fine mobilizable mate-
rial is present in important quantities.
USED DATASETS
A 2 m resolution digital elevation model (DEM)
was available, derived both from aerial laser scanning
for altitudes below 2000 m and from a 25 m DEM
(MNT25, swisstopo) built from the national maps at
1:25000 (CN25, swisstopo) for altitudes above 2000
m. The resolution for modelling was also 10 m.
Geological and tectonic vector maps at 1:500000
were available to assess the sediment availability,
which is an important criterion for the debris flow initi-
ation (R
iCkenmann
& z
immeRmann
, 1993; t
akaHasHi
,
1981). For this study, only 11 different lithologies were
taken into account based on those maps
As for the Canton de Vaud, a land use map based
on the product Vecteur25 (Swisstopo) was available
and improved the determination of the source areas.
SOURCES CRITERIA
As for the case of the Canton de Vaud, the mor-
phological data were integrated with the following pa-
rameters:
1 Slope threshold: 15 - 40°. The upper threshold was
found useful to remove cliffs where debris flow
cannot occur.
2 Flow accumulation threshold: 1 ha
3 Flow accumulation – slope relationship: extreme
threshold
4 Curvature: -2/100 m
-1
The selected lithologies from the structural map
are debris flow prone rocks (limestone, conglomer-
ate, flysch, meta-felsic, marble,quartzite, schist) and
slope deposits.
The land use and the geological information
were integrated in order to suppress source areas de-
tected on man-made structures. Bedrock was consid-
Fig. 7 - Resulting susceptibility map for the Bagnes valley.
Red identifies proven, orange probable and yellow
potential debris flows
background image
FLOW-R, A MODEL FOR DEBRIS FLOW SUSCEPTIBILITY MAPPING AT A REGIONAL SCALE – SOME CASE STUDIES
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
881
A geological map existed at scales of 1:50’000 or
1:1’000’000 according to the region.
SOURCES CRITERIA
The land cover and the geology layers were not
considered. Land cover information derived from the
Landsat image has too low resolution and many in-
consistencies. Geological data are not taken into ac-
count due to the fact that no mapped geology type can
be excluded as debris flow source areas. Indeed, there
is no main geology type in the studied area that is not
susceptible to debris flows. This has been confirmed
by field observations.
Thus, only the DEM and its derivatives remain.
Based on observation debris flows were divided in two
classes (1) common and (2) major. The first phenom-
enon is found in torrents with a limited flow accumula-
tion and an important slope. The second is found in
riverbeds that are much gentler, but that have a larger
water input leading to debris flows being less viscous.
Those were considered with the following parameters:
COMMON DEBRIS FLOwS
- Slope threshold: 15°
- Flow accumulation threshold: 2 ha. New sources
found below that limit are not relevant, and in-
troduce new channels where no accurate debris
flow was found. Above that limit, some noticeable
source areas are missing.
ered as potentially susceptible.
After field work and aerial photos analyses, the
sources were classified into 3 classes (Tab. 1).
PROPAGATION PARAMETERS
Propagation parameters were calibrated on the
basis of field observations and numerical data analy-
sis (orthophotos, DEM, topographic maps). After
testing, the 2-parameters friction model (Eq. 8) was
chosen for its good reproduction of past events.
1 Holmgren’s exponent was set to 6.
2 The parameters of the Perla friction model were
optimal at μ = 0.09 and M/D = 30.
The classification made on the sources was also
applied to the propagation areas. Moreover, a dis-
tinction was made between the probabilities that are
below 2% (probable danger) and above 2% (strong
probable danger).
CASE STUDY IN PAKISTAN
The susceptibility map for Pakistan was estab-
lished on two pilot districts: Muzaffarabad and Man-
shera (Fig. 8). Two main hazards were mapped: on the
one hand common debris flows, and on the other hand
major hyperconcentrated flows or major debris flows
(called major debris flows hereafter) that take place in
large rivers with an important upslope area. The areas
of the two districts are respectively 2496 km² for Mu-
zaffarabad and 4579 km² for Manshera. The altitude
goes up to 4500 m a.s.l..
USED DATASETS
The available Digital Elevation Model is the
SPOT DEM that has been furnished with a resolution
of 20 m.
A land cover was extracted from a Landsat image.
Six classes were defined: Forest, Water body, Snow
and Ice, Vegetation, Urban barren and Unclassified.
Tab. 1 - Classification of the debris flow source areas for
the Bagnes valley (after J
ABoyeDoff
et alii, 2010)
Fig. 8 - Resulting susceptibility map for the two pilot
districts in Pakistan. Red identifies debris flows
and yellow, hyperconcentrated flows
background image
P. HORTON, M. JABOYEDOFF, M. ZIMMERMANN, B. MAZOTTI & C. LONGCHAMP
882
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
through aerial photograph interpretation. No obvious
active torrent was missing. However, the lithological
map was not sufficient to exclude some unrealistic
source areas in regions where geology is the factor
discrediting the possibility of a debris flow.
The Val de Bagnes case study was interesting,
as fieldwork was possible due to the size of the val-
ley. Some occurred events were documented as they
affected roads (as the one depicted in Fig. 9). Those
could be compared to the model output detail. The
possibility to classify every source area as proven, po-
tential or incorrect, adds value to the indicative map-
ping process. The statistics on this classification (Tab.
2) shows that 86% of the identified source areas were
found to be pertinent. Most of the 14% remaining is
found in the bedrock class of the land use classifica-
tion. The identification of the source areas was com-
prehensive, as no actual debris flow source was found
outside of this susceptibility map.
During establishment of the susceptibility map for
Pakistan, field work was conducted at two different
regions to check the model outputs. No actual source
area in torrents seemed to miss on the map, and by far
the largest part of the modelled sources and spreading
could be confirmed. However, these field investigations
revealed the necessity to add the category of the major
- Flow accumulation – slope relationship: extreme
threshold. According to the high rainfall intensity
during monsoon in Pakistan, this threshold for ex-
treme events is accurate.
- Curvature: -1/100 m-1. It succeeded in identifying
the gullies prone to debris flows.
MAJOR DEBRIS FLOwS
- Slope threshold: 5-10°. Those events occur in riv-
erbeds that are relatively flat, as the water input is
higher than for common debris flows.
- Flow accumulation range: 500 ha to 5000 ha.
- Curvature: it was not taken into account for the ma-
jor debris flow source areas. Those happen in rela-
tively large riverbeds that are best identified by the
contributing area (flow accumulation).
PROPAGATION PARAMETERS
Propagation parameters were calibrated on the
basis of orthophotographs and field trips. For common
debris flows, the runout distance was found best rep-
resented by means of the 2 parameters friction model
(Eq. 8). Parameters were found optimal as following:
- Holmgren’s exponent was set to 6.
- The parameters of the Perla friction model were op-
timal at μ = 0.09 and M/D = 30.
- No velocity limitation was considered.
PARAMETERS FOR MAJOR DEBRIS FLOwS
ARE DIFFERENT
:
- Holmgren’s exponent was set to 1, meaning it cor-
responds to the multiple flow direction model. The
dispersion can be important for major debris flows.
- The runout distance was found best represented by
means of the energy line method with a 5° slope.
- No velocity limitation was considered
RESULTS
All three case studies resulted in susceptibility
maps, fully usable for a first assessment. However, it
remains to do field validation and detailed mapping.
The results should not be considered for local hazard
mapping without fieldwork. The results must be ana-
lysed at 1:25,000 scale. All interpretation of the model
at a finer scale makes little sense, particularly for the
source area assessment.
Results of the Canton de Vaud showed good
agreement with past debris flow features found
Fig. 9 - Example of spreading assessment in the Mau-
voisin region. A documented spreading zone is
highlighted, but the model identified a source area
and their respective propagation for every torrent
(modified after J
ABoyeDoff
et alii, in review)
Tab. 2 - Statistics of the source areas classification for the
Bagnes valley. (after J
ABoyeDoff
et alii, 2010)
background image
FLOW-R, A MODEL FOR DEBRIS FLOW SUSCEPTIBILITY MAPPING AT A REGIONAL SCALE – SOME CASE STUDIES
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
883
assessment. This was observed in every case study.
In Pakistan, where the DEM quality was lowest of all
case studies, some spreading were blocked by non-
existing dams present in the DEM (Fig. 10).
The DEM resolution and accuracy is a key ele-
ment which will condition the quality of outputs. It has
been observed in the Val de Bagnes, that for mountain
slopes above 2000 m, where the DEM is coarser, the
source areas identification is not as accurate as at lower
elevations. This is due to the fact that on the 1:25000
map data, gullies and small trenches are not repre-
sented, where they can be identified by aerial laser
scanning. This has been observed for Pakistan as well.
However, Pakistan is more concerned by large debris
flow hazards, starting from wider source areas that are
present in the low resolution DEM data.
Flow-R has been applied with success also in
Italy and France (b
laHut
et alii, 2010; l
aRi
et alii,
in review; k
aPPes
et alii, in review; l
aRi
et alii, this
volume). It permitted also to create a regional risk
analysis in Valentine area, showing the efficiency of
the model at a regional scale (l
aRi
et alii, this volume).
Its strengths are low requirements of data and a full
control for the user, namely it can provide meaningful
results with a DEM only and its derivatives, and all
data inputs, algorithms and parameters are open to the
user and can be changed.
ACKNOWLEDGEMENTS
Authors would like to thank the Canton de Vaud in
quality of mandating authority for susceptibility map-
ping project, the “Service des forêts, de la faune et de
la nature” (SFFN), Patrik Fouvy and Diane Morattel
for their involvement in the project. Thanks are due
also to Andrea Pedrazzini and Alexandre Loye for
important suggestions during the model development
phase. For the Pakistan project, thanks are due to the
Earthquake Reconstruction & Rehabilitation Author-
ity (ERRA) team, especially Air Cdre (R),Naunehal
Shah , Mujeeb Alam and Ejaz Karim for their precious
field knowledge and enjoyable collaboration, and the
UNDP for financial and logistical support. The authori-
ties and the communal geologist Dr. F. Baillifard are
thanked to authorized us to published results
hyperconcentrated flows and major debris flows that
occur in large riverbeds with a gentle slope.
DISCUSSION AND CONCLUDING RE-
MARKS
Results of the various case studies confirm the sig-
nificance of the model outputs although the model has
some limitations and cannot account for certain local
controlling factors. Observations of past events in the
field were satisfyingly reproduced during the calibra-
tion and validation procedures. Thus, for the purpose
of susceptibility mapping, the outputs can be consid-
ered as accurate. In the framework of susceptibility
mapping, the identified areas are often larger than the
observed events on the field. This is on purpose, as
the map should be representative of the worst cases,
soalways rather conservative (J
aboyedoff
et alii, in re-
view).Susceptibility hazard maps provide an excellent
overview to indicate where detailed field investigation
should be conducted to develop a hazard map
The DEM and its derivative are the most valu-
able data for the method. Morphological criteria were
found to be very pertinent for debris flow sources iden-
tification, whereas other inputs just helped removing
inaccurate source areas. However, those still improve
the mapping pertinence. This means that the results
are conditioned by the DEM quality both in terms of
resolution and accuracy. Indeed, artefacts can be mis-
leading for both sources identification and propagation
Fig. 10 - Effect of a DEM error above Nardajian (Muzaf-
farabad). The model stops due to an anomaly in
the DEM. Thus, the model misses a past event (de-
picted in blue)
background image
P. HORTON, M. JABOYEDOFF, M. ZIMMERMANN, B. MAZOTTI & C. LONGCHAMP
884
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
REFERENCES
b
laHut
J., v
an
w
esten
C.J & s
teRlaCCHini
s. (2010) - Analysis of landslide inventories for accurate prediction of debris-flow
source areas. Geomorphology, 119 (1-2): 36-51.
C
laessens
L. (2005) - DEM resolution effects on shallow landslide hazard and soil redistrib. modelling. Earth Surf. Proc. Land.,
30: 461-477.
d
elmonaCo
G., l
eoni
G., m
aRGottini
C., P
uGlisi
C. & s
PizziCHino
d. (2003) - Large scale debris-flow hazard assessment: a
geotechnical approach and GIS modelling. Nat. Hazard Earth Sys., 3: 443-455.
e
Rskine
R.H., G
Reen
t., R
amiRez
J. & m
aC
d
onald
l. (2006) - Comparison of grid-based algorithms for computing upslope
contributing area. Water Resour. Res., 42 (9).
f
aiRfield
J. & l
aymaRie
P. (1991) - Drainage Networks From Grid Digital Elevation Models. Water Resour. Res., 30 (6): 1681-
1692.
f
Reeman
T.G. (1991) - Calculating catchment area with divergent flow based on a regular grid. Computers & Geosciences, 17
(3): 413-422.
G
amma
P. (2000) - dfwalk - Ein Murgang- Simulationsprogramm zur Gefahrenzonierung. Bern, Switzerland.
H
einimann
H.R. (1998) - Methoden zur Analyse und Bewertung von Naturgefahren. Bundesamt für Umwelt, Wald und Land-
schaft (BUWAL).
H
olmGRen
P. (1994) - Multiple flow direction algorithms for runoff modelling in grid based elevation model. Hydrol. Process.,
8(4): 327-334.
H
oRton
P., J
aboyedoff
M. & b
aRdou
E. (2008) - Debris flow susceptibility mapping at a regional scale. 4th Canadian Conference on
Geohazards: From Causes to Management, Université Laval, Québec, Qc, Canada, Presse de l'Université Laval.
H
uGGel
C., k
ääb
A., H
aebeRli
W. & k
RummenaCHeR
B. (2003) - Regional-scale GIS-models for assessment of hazards from
glacier lake outbursts: evaluation and application in the Swiss Alps. Nat. Hazard Earth Sys., 3: 647-662.
i
veRson
R.m. & d
enlinGeR
R.P. (2001) - Mechanics of debris flows and debris-laden flash floods. Seventh Federal Interagency
Sedimentation Conference, Reno, Nevada, USA.
J
aboyedoff
M., P
edRazzini
a., l
oye
a., H
oRton
P. & s
uRaCe
i. (2008) - Cartes Indicatives de Danger des Mouvements de Ver-
sants du Canton de Vaud. Technical report, University of Lausanne, Switzerland.
J
aboyedoff
M., m
iCHoud
C., m
azotti
B., C
Hoffet
M., d
ubois
J., b
ReGuet
A., m
etRaux
V., d
eRRon
M.H., H
oRton
P., l
oye
A. &
P
edRazzini
A. (2010) - Cartes Indicatives de Dangers pour le Val de Bagnes et de Vollèges. Carnet méthodologique, 89 pp.
l
aRi
S., C
Rosta
G.B., f
Rattini
P., H
oRton
P. & J
aboyedoff
M. (this volume) - Regional-scale debris-flow risk assessment for
an alpine valley.
P
eRla
R. (1980) - A two-parameter model of snow-avalanche motion. J. Glaciol., 26 (94): 197-207.
Q
uinn
P., b
even
k., C
HevalieR
P. & P
lanCHon
o. (1991) - The Prediction Of Hillslope Flow Paths For Distributed Hydrological
Modelling Using Digital Terrain Models. Hydrol. Process., 5 (1): 59-79.
R
iCkenmann
d. & z
immeRmann
m. (1993) - The 1987 debris flows in Switzerland: docum. and analysis. Geomorphology, 8
(2-3): 175-189.
s
aRtoRi
m., G
ouffon
y. & m
aRtHaleR
m. (2006) - Harmonisation et définition des unités lithostratigraphiques briançonnaises
dans les nappes penniques du Valais. Eclogae Geologicae Helvetiae, 99: 363-407
t
akaHasHi
T. (1981) - Estimation of potential debris flows and their hazardous zones : soft countermeasures for a disaster.
Natural disaster science, 3 (1): 57-89.
t
aRboton
D.G. (1997) - A new method for the determination of flow directions and upslope areas in grid digital elevation mod-
els. Water Resour. Res., 33 (2): 309-319.
t
aRboton
D.G., b
Ras
R.l. & R
odRiGuez
-i
tuRbe
i. (1991) - On the extraction of channel networks from digital elevation data.
Hydrol. Process., 5 (1): 81-100.
t
uRbeRG
P., P
aRRiaux
a., k
albeRmatten
m., G
olay
f. & l
anCe
J.m. (2008) - The geotype concept to develop GIS oriented
analysis in engineering geology application. Swiss Geoscience Meeting conference, Lugano, Switzerland.
w
ieCzoRek
G.f., m
andRone
G. & d
eCola
l. (1997) - The Influence of Hillslope Shape on Debris-Flow Initiation. In: ASCE (Edi-
tor), First International Conference Water Resources Engineering Division, San Francisco, CA: 21-31.
Statistics