# IJEGE-11_BS-Laigle&Peteuil

*DOI: 10.4408/IJEGE.2011-03.B-100*

**ANALYSIS OF THE DEBRIS-FLOW HAZARD ON THE RIOULONG**

**TORRENT (HAUTES-PYRéNéES, FRANCE) ALLUVIAL FAN**

**USING A SCENARIO-BASED APPROACH**

and potential damage) and probability of occurrence

(once every 10 years, 100 years, 1000 years?) of events

likely to affect a given point in the area (a house, for

instance). It is therefore crucial to take this variability

into account. In this context, the scenario-based ap-

proach exemplified here can be considered a practical

method of hazard assessment. It consists first in a di-

agnosis of debris-flow triggering processes at work in

the catchment considered. Then the ranges of variation

of all the inputs of a debris-flow simulation model are

deduced from the diagnosis. Each of the input values

is then assigned a probability of occurrence, mostly

qualitative (frequent, rare, exceptional, etc.). Numeri-

cal simulations, covering these predefined ranges of

variation of input data are then carried out. Each simu-

lation result can subsequently be assigned a probabil-

ity of occurrence, once again mostly qualitative.

vial fan, with a few houses, a camp site and a local road

within the hazard zone. First, the catchment is diagnosed

considering erosion processes at work in the catchment.

This diagnosis means that the probability of debris-flow

triggering cannot be ruled out. It also provides an evalu-

ation of the magnitude of debris flows likely to affect the

alluvial fan. In the second part of this article, a scenario-

based analysis, using a numerical model dedicated to the

computation of debris-flow spreading on alluvial fans, is

carried out in order to assess the debris-flow hazard. In

**ABSTRACT**

based approach, is applied to a torrent of the French

Pyrenees. A preliminary analysis of the catchment

shows that some risk of debris-flow occurrence is

present and that scenarios should be constructed ac-

cordingly. Numerical simulations are carried out on

the basis of these scenarios. They produce maps of

maximum extension of debris flows in relation to

qualitative levels of probability of occurrence. Simu-

lations also provide information that is useful for the

design of a protection structure. Finally, the maximum

extension of debris flows resulting from the presence

of the protection structure is analyzed.

**K**

**ey**

**words***: debris-flow, mudflow, erosion processes, hazard*

*assessment, numerical simulation, protection structures*

**INTRODUCTION**

ditions. This variability originates, of course, in the

occurrence of triggering rainfall conditions, but also

in the nature and availability of solid material present

in the debris-flow-prone catchment. These two factors,

once combined, will determine not only the volume of

each individual event, but also the mechanical proper-

ties, and more generally speaking, all flow conditions.

For people in charge of protection strategies, it is nec-

*D. LAIGLE & C. PETEUIL*

m/m (Figure 1). In this area, the stream cuts into ancient

morainic deposits. Covering about one-quarter of the

catchment, this sector resembles a large hanging humid

zone with more or less active instabilities. It is also the

confluence point of several gullies of the upper catch-

ment hydrographic network (Figure 2). Given the small

size of the basin, the steepness of slopes, gullies and

stream bed, the large number of gullies, and the pres-

ence of humid or impervious zones, the catchment is

likely to respond very quickly to any rainfall event.

*et alii*, 2003), but other models are able to compute the

spreading of debris flows as well, for instance, FLO2D

(o’b

*et alii*, 1993) could similarly be employed.

levee is also analyzed within the scenario-based ap-

proach. This analysis provides useful elements for the

design of the protection structure. Finally, the conse-

quences of the presence of the protection levee on the

flow downstream, and thus on the hazard on the al-

luvial fan, are also analyzed.

**PRELIMINARY ANALYSIS OF THE RIOU-**

**LONG CATCHMENT**

*MAIN FEATURES OF THE CATCHMENT*

district in France (Fig. 1). This torrent catchment covers

1.72 km² and ranges from 1920 m to 960 m a.s.l. The

catchment is for the most part well covered by vegeta-

tion and partly subjected to anthropogenic effects. Six

per cent of the catchment area is impervious because

of the presence of the Val Louron ski resort in its up-

per part. The channel in the medium reach is rather

*Fig. 1 - General overview of the Rioulong catchment*

*(photo: ONF - RTM 65-64)*

*Fig. 2 - Limits of the catchment and map of sediment production areas*

**ANALYSIS OF THE DEBRIS-FLOW HAZARD ON THE RIOULONG TORRENT (HAUTES-PYRéNéES, FRANCE)**

**ALLUVIAL FAN USING A SCENARIO-BASED APPROACH**

1). These assessments show very large uncertainties

but given the present state of the catchment, they can

be considered realistic.

*Possible magnitude of debris flows*

analysis of the lengthwise profile of the stream show that

the risk of debris-flow triggering cannot be neglected.

deposits, are present in the vicinity of the channel and

gullies and are likely to mobilize. Furthermore, the

steep longitudinal slope of the channel and gullies in

this area are compatible with the triggering and propa-

gation of mudflows down to the alluvial fan.

potentially mobilized, it appears pertinent to consider

at least three debris-flow scenarios associated with

a probability of occurrence qualified as follows: in-

frequent, rather rare and rare. The absence of recent

events, however, gives substantial uncertainty to the

estimations. We therefore considered the possible oc-

currence of an exceptionally rare event. Table 2 gives

the debris-flow volume values associated with these

qualitative terms and an order of magnitude of the an-

nual probability of occurrence of these volumes.

scribed. These values were established based on the

catchment hydrology and the empirical relationship

between peak discharge and volume of a debris-flow

event proposed by R

1929, 1936, and 1987, according to the unpublished ar-

chives of the ONF-RTM department who is in charge of

the torrent control). The consequences were severe for

human activities in the vicinity of the stream. Charac-

terized by huge sediment transport with gravels, blocs,

and mud of glacial origin, some of these floods probably

included debris flows. This is confirmed by the typical

shape of ancient deposits still present on the alluvial fan.

early as the late 19

alluvial fan, thus increasing the vulnerability already

present in this area (several houses and a local road).

A drainage network and check dams were built dur-

ing the 1990s (unpublished archives of the ONF-RTM

department who is in charge of the torrent control).

*EROSION AND DEBRIS-FLOw MAGNITUDE*

*Erosion processes*

phenomenon is a landslide in morainic terrains, drained

by the Paulède gully, a small tributary of the Rioulong

torrent. Consequently, sediment transport in the chan-

nel remains very limited for common discharge values,

occurring once a year on average. Nevertheless, for in-

tense flooding, bed destabilization is likely to occur in

the steep channel between the upper basin and the al-

luvial fan, involving at least 700 m of channel where

the bed grain size distribution appears very small com-

pared to the stream steepness. Only one-third of this

channel has been protected by check dams.

z

*et alii*(1997) and completed by detailed

triggered under long and intense rainfall conditions.

*Tab. 1 - Assessment of potential mobilized volume of mate-*

*rial considering erosion processes at work*

*Tab. 2 - Reference scenarios considered: qualitative probability of occurrence, volume and tentative value of the annual prob-*

*ability of occurrence*

*D. LAIGLE & C. PETEUIL*

invite the reader to refer to l

*et*

*alii*(2003) and R

*et alii*(2006).

**CONSTRUCTING THE SCENARIOS**

square meter), using the LAVE2D model (l

*et*

*alii,*2006) required two additional input data: the de-

bris-flow hydrograph at the alluvial fan apex and the

rheological parameters of the flowing material.

*INPUT HYDROGRAPH*

on the shape of this hydrograph, we assume a linear evo-

lution of the discharge between its peak value at time t =

0 to a zero value at time t

The scenarios studied hereafter have been inferred from

the diagnosis presented previously. The hydrograph is

“injected” into the computation domain (Fig. 3) in the

channel located at the apex of the alluvial fan.

*RHEOLOGICAL PARAMETERS*

tions (shape of ancient deposits, grain size distribu-

tion of soils present on the slopes and in the gullies).

(including events with a low discharge value) is quite

high. Some spreading of the flow towards human set-

tlements on the alluvial fan is therefore likely to occur.

**MODEL PRESENTATION**

*et alii*, 2003) is a numerical

free-surface spreading of materials with complex rhe-

ology. It is based upon the 2D steep-slope-shallow-wa-

ter-equations which are solved by using a finite volume

technique. It takes into account viscous dissipation in-

side the flowing material, assumed homogeneous, by

the use of the wall shear stress expression. This expres-

sion from C

sented by a Herschel-Bulkley model mainly applying

to mudflows or so-called viscous debris flows. Apart

from values of the rheological parameters, model in-

puts are: boundary conditions (imposed discharge ver-

sus time at the point where the flow enters the zone

of spreading), a computation mesh combined with a

digital elevation model of the alluvial fan and a set of

numerical parameters (i.e. stability criterion of the nu-

merical scheme, simulation duration). For more details

about this model and its evaluation by comparison to

*Fig. 3 - Maximum simulated flow depth for a debris-flow volume of 10 000 m*

*3*

*, a peak discharge of 100 m*

*3*

*/s and a τ*

*c*

*/ρ = 0.5 m*

*2*

*/s*

*2*

*ratio*

**ANALYSIS OF THE DEBRIS-FLOW HAZARD ON THE RIOULONG TORRENT (HAUTES-PYRéNéES, FRANCE)**

**ALLUVIAL FAN USING A SCENARIO-BASED APPROACH**

schel-Bulkley model. These material properties are de-

termined by two parameter values: the yield-stress to

density ratio,

*τ*

*c*

*/ρ*(m

*k/τ*

*c*

the following assumptions. The value of parameter

*k/*

τ

τ

*c*

value of parameter

*τ*

*c*

*/ρ*is considered parametrically

- An average value

*τ*

*c*

*/ρ*= 1.0 m

der of magnitude is frequently observed on many

torrents. Its probability of occurrence is conside-

red high. Furthermore, the observation of the an-

cient deposits on the Rioulong alluvial fan tends

to confirm this assumption.

*τ*

*c*

*/ρ*= 2.0 m

in the parametric approach because it gives high

flow and deposit thickness values (considered

here as the highest thickness likely to be observed

for a given set of discharge and volume values).

*τ*

*c*

*/ρ*= 0.5 m

parametric approach because it gives high flow ve-

locities and extensions (considered here as the hi-

ghest velocity and extension likely to be observed

for a given set of discharge and volume values).

*SYNTHETIC PRESENTATION OF SCENARIOS*

CONSIDERED

CONSIDERED

in Table 3. They do not cover all possible cases, for

practical purposes unlimited in number. For each of

the cases considered, perturbing phenomena are likely

to modify the result substantially. As an example of

*Tab. 3 - Synthetic presentation of all scenarios considered in the study. Not all of them are illustrated in the present paper*

*D. LAIGLE & C. PETEUIL*

other torrents (Rickenmann

*et alii*, 2006). However,

one can see an important trend on Rioulong: the lat-

eral overflow at the alluvial fan apex of all the mate-

rial coming from upstream with a distribution between

the right and the left bank. This trend is confirmed by

simulations presented below.

*INFLUENCE OF THE PEAk DISCHARGE*

tions are carried out considering a given volume of

10 000 m

*τ*

*c*

*/ρ*= 1.0 m

maximum extension of the flow. The peak discharge

value essentially influences the flow velocities and

to a lesser extent the flow depth, mainly in the chan-

nelized area at the apex of the alluvial fan.

*INFLUENCE OF THE VOLUME*

lations were carried out considering a peak discharge

of 100 m

*τ*

*c*

*/ρ*=

sumption of log jamming occurring in the channel at

the apex of the alluvial fan. Two protection levee as-

sumptions are also considered.

**HAZARD ASSESSMENT: ANALYSIS OF MA-**

**XIMUM EXTENSION OF DEBRIS FLOWS**

sidered their maximum extent as the most pertinent

criterion for the hazard analysis on the Rioulong al-

luvial fan. These extensions are mapped for each of

the scenarios considered. However, rather than con-

sidering only the extensions it is more interesting to

consider their variation related to any variation of the

model’s input parameters. This is why the results are

presented as a sensitivity analysis.

*INFLUENCE OF THE YIELD-STRESS VALUE*

ing material. We consider a given volume of 10 000

m

fluid (

*τ*

*c*

*/ρ*= 0.5 m

*τ*

*c*

*/ρ*= 1.0

*τ*

*c*

*/ρ*= 2.0 m

*Fig. 4 - Maximum simulated flow depth for a debris-flow volume of 10 000 m*

*3*

*, a peak discharge of 100 m*

*3*

*/s and a τ*

*c*

*/ρ = 1.0 m*

*2*

*/s*

*2*

*ratio*

**ANALYSIS OF THE DEBRIS-FLOW HAZARD ON THE RIOULONG TORRENT (HAUTES-PYRéNéES, FRANCE)**

**ALLUVIAL FAN USING A SCENARIO-BASED APPROACH**

*τ*

*c*

*/ρ*= 0.5 m

less thick. The simulations are carried out considering

a peak discharge of 100 m

(represented as black rectangles in the figures) located

ume has a substantial influence on the flow extension.

This result is coherent with previous results obtained

on other torrents (R

*et alii*, 2006).

*ANALYSIS OF THE MAXIMUM EXTENSION OF*

A FLUID MATERIAL

A FLUID MATERIAL

*Fig. 5 - Maximum simulated flow depth for a debris-flow volume of 25 000 m*

*3*

*, a peak discharge of 100 m*

*3*

*/s and a τ*

*c*

*/ρ = 1.0 m*

*2*

*/s*

*2*

*ratio*

*Fig. 6 - Maximum simulated flow depth for a debris-flow volume of 15 000 m*

*3*

*, a peak discharge of 100 m*

*3*

*/s and a τ*

*c*

*/ρ = 0.5 m*

*2*

*/s*

*2*

*ratio*

*D. LAIGLE & C. PETEUIL*

alluvial fan, are likely to be reached by debris flows.

Additionally, as for the previous simulations, the risk

is high on the left bank of the channel (notably the

camp site area).

*EXAMPLE OF THE INFLUENCE OF LOG*

JAMMING

JAMMING

sumption of log jamming. It is unrealistic to consider

all possible assumptions of log jamming, since this

phenomenon is likely to occur almost anywhere in the

channel and the resulting plug dimensions can vary

greatly. Consequently, we illustrate this phenomenon

using an example. The assumption is that the log jam-

ming occurs at the alluvial fan apex, where previous

simulations show that the flow tends to spread later-

ally. Simulations are based on a peak discharge of 100

m

at the alluvial fan apex.

of the log jamming tends to increase the overflow

this overflow is not fundamentally modified. Conse-

quently, this plausible phenomenon tends to increase

the volume of material driven to the left part of the

alluvial fan and increase the risk in this sector, no-

tably in the camp site area. This phenomenon can

explain the numerous old flow traces still visible in

this area.

**A PROTECTION LEVEE AGAINST OVER-**

**FLOW: DESIGN ELEMENTS**

*COMPARING TwO POSSIBLE ORIENTATIONS*

OF THE PROTECTION LEVEE

OF THE PROTECTION LEVEE

the RTM service, in charge of controlling the torrent,

considered of building a protection levee within a

more general protection strategy. The objective of the

levee, located at the apex of the alluvial fan where the

flow tends to diverge, is to eliminate the overflow risk

towards the left bank. We examine two orientations of

the levee. The first orientation (type A levee in Fig-

ure 8) presents a 45° angle with the upstream channel

and a second one (type B levee in Fig. 8) presents a

15-20° angle with the upstream channel. The follow-

ing assumptions are considered for the comparison of

these orientations: an average viscosity (

*τ*

*c*

*/ρ*= 1.0 m

*Fig. 7 - Maximum simulated flow depth for a debris-flow volume of 10 000 m*

*3*

*, a peak discharge of 100 m*

*3*

*/s and a τ*

*c*

*/ρ = 1.0*

*m*

*2*

*/s*

*2*

*ratio assuming log jamming of the stream at the apex of the alluvial fan*

**ANALYSIS OF THE DEBRIS-FLOW HAZARD ON THE RIOULONG TORRENT (HAUTES-PYRéNéES, FRANCE)**

**ALLUVIAL FAN USING A SCENARIO-BASED APPROACH**

*CINITY OF THE LEVEE*

- analyze the flow in the vicinity of the levee in

velocity under several assumptions.

To do this, we consider:

carried out under this assumption will provide in-

formation on maximum flow velocities likely to

occur in the vicinity of the levee.

lations carried out under this assumption will pro-

vide information on maximum flow depths likely

to occur in the vicinity of the levee (as well as

elements for the future design of the levee).

value. Simulations carried out under these assum-

ptions will provide information on the maximum

flow extension in presence of the levee.

The volume has a very limited influence on the

tion leading to the maximum velocity (peak discharge

ciency and then in terms of the required dimensions.

Levees are numerically considered to be cells imper-

vious to any flow.

of flow extension, both types of have similar effects.

However, flow depths and velocities in the vicinity of

the type B levee are lower than for the type A levee

(maximum flow depth: 2.8 m and maximum veloc-

ity: 5.6 m/s for type A and maximum flow depth: 1.5

m and maximum velocity: 1.0 m/s for type B). This

presents two advantages in favor of type B: for simi-

lar efficiency, the type B levee can be lower and the

risk of impact and erosion is also lower because of

a lower velocity. Furthermore, the type A levee cre-

ates a bottle-neck for the channel flows immediately

downstream of the levee, producing local acceleration

of the flows (maximum local velocity under the as-

sumptions considered: 8.2 m/s). The type B levee does

not generate this bottle-neck effect and thus no local

acceleration of the flow (maximum local velocity un-

der considered assumptions: 5.3 m/s). A type B levee

is therefore preferred.

*ANALYSIS OF FLOw SIMULATIONS IN THE VI-*

*Fig. 8 - Maximum simulated flow depth for a debris-flow volume of 10 000 m*

*3*

*, a peak discharge of 100 m*

*3*

*/s and a τ*

*c*

*/ρ = 1.0*

*m*

*2*

*/s*

*2*

*ratio assuming the presence of a protection levee showing a 45° angle with the channel axis*

*D. LAIGLE & C. PETEUIL*

*τ*

*c*

*/ρ*= 0.5 m

very locally 13 m/s in the vicinity of the downstream

edge of the levee (Fig. 9). The assumption leading to

the maximum flow depth (peak discharge of 180 m

*τ*

*c*

*/ρ*= 2.0 m

ity of the levee (Fig. 10). Even though these figures

to the future design of the levee.

*CONSEQUENCES OF THE PRESENCE OF THE*

LEVEE

LEVEE

and subsequent extensions on this part of the al-

*Fig. 9 - Maximum simulated flow velocity for a debris-flow volume of 10 000 m*

*3*

*, a peak discharge of 180 m*

*3*

*/s and a τ*

*c*

*/ρ = 0.5*

*m*

*2*

*/s*

*2*

*ratio assuming the presence of a protection levee showing a 15° angle with the channel axis*

*Fig. 10 - Maximum simulated flow depth for a debris-flow volume of 10 000 m*

*3*

*, a peak discharge of 180 m*

*3*

*/s and a τ*

*c*

*/ρ = 2.0*

*m*

*2*

*/s*

*2*

*ratio assuming the presence of a protection levee showing a 15° angle with the channel axis*

**ANALYSIS OF THE DEBRIS-FLOW HAZARD ON THE RIOULONG TORRENT (HAUTES-PYRéNéES, FRANCE)**

**ALLUVIAL FAN USING A SCENARIO-BASED APPROACH**

fected by debris-flow hazards can be predicted. It

also provides practical information on the design of

a protection levee, including guidelines on the levee’s

orientation and dimensions based on modeled maxi-

mum flow depths and velocity and on evaluation of

the potential changes to the debris-flow hazard area on

the alluvial fan due to the presence of the levee.

the variability of debris-flow hazards on an alluvial fan

without the costs of more rigorous approaches. Such

rigorous approaches do exist, however they all require

the assessment of statistical properties for model input

parameters which is difficult to achieve in practice.

houses located downstream. However, the houses

are potentially hit in only one of the two scenarios

considered with a volume of 25 000 m

*τ*

*c*

*/ρ*=

**CONCLUSION**

ysis of the torrent catchment for producing debris flows.

This analysis was a necessary step prior to the numeri-

cal modeling for the scenario-based analysis employed

in the second phase of this study. This analysis made it

possible to define flow volume and discharge scenarios

and to assign a qualitative probability of occurrence. In

conjunction with the assumptions of flow rheology, a

*Fig. 11 - Maximum simulated flow depth for a debris-flow volume of 25 000 m*

*3*

*, a peak discharge of 180 m*

*3*

*/s and a τ*

*c*

*/ρ = 1.0*

*m*

*2*

*/s*

*2*

*ratio assuming the presence of a protection levee showing a 15° angle with the channel axis*

**REFERENCES**

*Steady laminar flow of concentrated mud suspensions in open channel*. Journal of Hydraulic Research,

**32**

*Les laves torrentielles. Connaissances à l’usage du praticien*, Cemagref Editions, Antony, 177 pp. (in French).

*A two-dimensional model for the study of debris flow spreading on a torrent debris fan*. In proceedings of the

*Comparison of numerical simulation of muddy debris flow spreading*

*to records of real events*. In proceedings of the 3

*D. LAIGLE & C. PETEUIL*

*Two-dimensional water flood and mudflow simulation*. Journal of Hydraulic

**119**(2): 244-261.

*Empirical relationships for debris flows*. Natural Hazards,

**19**(1): 47-77.

*Comparison of 2D debris-flow simulation models with field*

*events*. Computational Geosciences,

**10**(2): 241-264.

*Murganggefahr und klimaänderung - ein*

*GIS-basierter Ansatz*, In. Schlussbericht Nationalen Forschungs-Programms, NFP 31, Zürich, Vdf Hochschulverlag AG an

der ETH, 161 pp. (in German).