Document Actions

IJEGE-11_BS-Mergili-et-alii

background image
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
973
DOI: 10.4408/IJEGE.2011-03.B-106
GLACIAL LAKE OUTBURST FLOODS IN THE PAMIR OF TAJIKISTAN:
CHALLENGES IN PREDICTION AND MODELLING
m
aRtin
MERGILI
(*)
, d
emian
SCHNEIDER
(**)
, R
aPHael
WORNI
(***, ****)
& J
ean
F. SCHNEIDER
(*)
(*)
Institute of Applied Geology, Department of Civil Engineering and Natural Hazards, University of Natural
Resources and Applied Life Sciences (BOKU Vienna), Peter-Jordan-Strasse 70, A-1190 Vienna, Austria
(**)
Glaciology, Geomorphodynamics and Geochronology, Department of Geography, University of Zurich,
Winterthurerstrasse 190, CH-8057 Zurich, Switzer-land
(***)
Institute for Environmental Sciences, University of Geneva, Switzerland, Chemin de Drize 7, CH-1227 Carouge, Switzerland
(****)
Institute of Geological Sciences, University of Berne, Baltzerstrasse 1-3, CH-3012 Berne, Switzerland
INTRODUCTION
Natural dams of different size and origin do exist
in mountain areas all over the world (C
osta
& s
CHus
-
teR
, 1988). They often retain lakes which, in the case
of a dam failure, may drain as powerful floods. If the
outbursting lake is located within the glacial or per-
iglacial area, such events are called Glacial Lake Out-
burst Floods (GLOFs). They can evolve in different
ways (Fig. 1), for example:
• rock/ice avalanches or calving glaciers that pro-
duce flood waves in a pro-, supra- or periglacial
lake which may overtop and breach glacial or mo-
rainic dams (t
inti
et alii, 1999);
• rising pro-, supra-, sub- or periglacial lake levels,
leading to overflow, progressive incision or me-
chanical rupture of a moraine or ice dam, as well
as to retrogressive erosion of a moraine dam;
• enhanced ground water flow (piping) through mo-
raines, or hydrostatic failure of ice dams which
can cause sudden outflow of accumulated water
(i
tuRRizaGa
, 2005a; 2005b);
• degradation of glacier dams or ice-cores in morai-
nic dams leading to loss of stability and to subsi-
dence resulting in internal failure or progressive
erosion if a certain threshold is reached.
R
iCHaRdson
& R
eynolds
(2000) provide an over-
view of failure mechanisms and case studies. GLOFs
often have a highly destructive potential because a
large amount of water is released within a short time,
with a high capacity to erode loose debris, potentially
ABSTRACT
Glacial lake outburst floods (GLOFs) are poten-
tially highly dangerous events and have contributed to
numerous disasters in history. Today, computer models
are standard tools to estimate the magnitude of hazard-
ous events in the future and to support risk mitigation.
The present paper explores the potentials and limitations
of modelling for predicting the motion of potential fu-
ture GLOF events, based on examples from the Pamir
(Tajikistan). Since the flow behaviour of GLOFs is in
between debris flows and floods, different model ap-
proaches come into consideration, though none of them
is perfectly suitable for GLOFs. RAMMS as a mass
movement model and FLO-2D as a river hydraulics
model were employed comparatively for the same ar-
eas. The friction parameters for RAMMS and rheologic
parameters for FLO-2D were first calibrated by back-
calculation with the well-documented Dasht event from
summer 2002, and then applied to other areas. However,
the applicability of such parameters to GLOFs of dif-
ferent volume and over a different topography remains
questionable. The results may nevertheless be a valuable
input for risk mitigation efforts, but due to the complex
nature of GLOFs and the connected uncertainties, par-
ticular care is required when interpreting the model re-
sults. The critical points and potential approaches to deal
with the limitations are discussed in the paper.
K
ey
words
: glacial lake outburst floods (GLOFs), modelling,
Central Asia
background image
M. MERGILI, D. SCHNEIDER, R. wORNI & J. F. SCHNEIDER
974
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
cial lake had suddenly released an estimated volume of
250,000 m³ of water (s
CHneideR
et alii, 2004). The vol-
ume of debris deposited on the cone was estimated 1.0-
1.5 million m³, meaning that the ratio between entrained
debris and water would be 4-6. This is a very high value
compared to the ratio of 2-3 suggested by H
uGGel
et alii
(2004b). However, an even higher ratio than observed
for Dasht was reported by b
Reien
et alii (2008) for a
GLOF in Norway. Possibly, subglacial water reservoirs
connected to the superficial lake and highly saturated
erodible material was involved in both events.
It was reported that the flood wave arrived in Dasht
in three stages, a phenomenon that can be explained by
temporary backwater in the canyon of the lower transi-
tional zone due to blockage of large boulders transport-
ed by the GLOF or by lateral slope failures followed by
vigorous breakthroughs (s
CHneideR
et alii, 2004). The
event destroyed a large portion of the village of Dasht,
killed a few dozens of people, and dammed a small lake
at the Shakhdara river. The event hit the village com-
pletely unexpected, as there was no awareness of the
hazard and preparedness for the event.
Even though potentially hazardous supra-, pro- and
periglacial lakes can be identified relatively easily with
remote sensing tools and field work (e.g. k
aeaeb
et alii,
2005; Q
uinCey
et alii, 2007), modelling and prediction
of the motion and reach of GLOFs still remain a chal-
lenge. Like many other GLOFs, the characteristics of
the Dasht event underwent pronounced changes dur-
ing the flow, converting from normal runoff to a hy-
perconcentrated flow and finally to a granular debris
flow. Changes in flow behaviour imply some difficulties
when using computer models to predict the flow path
and velocities of such events. Simple empirical rules for
debris flows travel distances show a large scatter among
themselves and generally underestimate the travel dis-
tance of GLOFs (Fig. 3). C
oRominas
et alii (2003) as-
leading to a powerful flow with a long travel distance.
Peak discharges are often some magnitudes higher than
in the case of “normal” floods (C
endeRella
& w
oHl
,
2001). The source area is usually far away from the
area of impact and events occur at very long time in-
tervals or as singularities, so that the population at risk
is often not prepared for such events (s
CHneideR
et alii,
2004). Deficiencies in risk communication are often
responsible that events evolve into disasters (C
aRey
,
2005). A number of significant GLOFs resulting in
fatalities and severe damage have occurred during
the previous decades, particularly in the Himalayas,
the mountains of Central Asia, the North American
mountains, New Zealand, and the Alps. Case studies
are provided e.g. by C
laRke
(1982); H
ewitt
(1982);
w
atanabe
& R
otHaCHeR
(1996); R
iCHaRdson
& R
ey
-
nolds
(2000); s
CHneideR
et alii (2004) and v
ilimek
et alii (2005). Climate change, with its impact on the
glacial extent, the hydrological cycle and the condition
of ice-bearing dams, may condition the occurrence of
GLOFs in manifold ways and on different time scales
(e
vans
& C
laGue
, 1994; d
ussaillant
, 2009).
The present paper deals with computer model-
ling of the flow path of GLOFs. Using test areas in the
Pamir (Tajikistan), the general potentials and limita-
tions of such approaches as well as the suitability of
different model concepts are explored and discussed.
Particular emphasis is put on the capabilities of the
models for the prediction of future events.
BACKGROUND
In summer 2002, the village of Dasht (Shakhdara
Valley, Pamir, Tajikistan; Fig. 2a) was hit by a GLOF. 10
km upstream in the headwaters of the valley, a supra-gla-
Fig. 1 - Schematic representation of a glacial lake out-
burst flood (GLOF)
Fig. 2a - Left: The debris cone resulting from the GLOF in
Dasht in summer 2002, covering most of the vil-
lage and damming a small lake upstream
Fig. 2b - Right: Lake dammed by a rock glacier in the up-
per khavrazdara Valley
background image
GLACIAL LAKE OUTBURST FLOODS IN THE PAMIR OF TAJIKISTAN:
CHALLENGES IN PREDICTION AND MODELLING
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
975
Several physically based model approaches and
software packages are potentially suitable for GLOF
runout modelling, some of which were developed
within the mass movement research community, oth-
ers within the river hydraulics community.
Many mass movement models go back to the
v
oellmy
(1955) approach and were developed for
snow avalanches, but are also applicable to other
types of mass movements. A remaining problem is the
entrainment of material that is an important character-
istic of GLOFs (b
Reien
et alii, 2008; x
u
, 1988). Some
models include entrainment modules, but rather on an
empirical-statistical than on a physical base. b
Reien
et
alii (2008) emphasize the lack of appropriate data and
knowledge on entrainment issues.
River hydraulics models commonly use flood
routing algorithms based on volume conservation and
a roughness parameter (usually Manning’s n) for esti-
mating the extent and the depth of river flow and flood-
ing events. Most of the widely used software packages
(e.g. FLO-2D, HecRAS) include modules for sediment
transport, hyperconcentrated flows, and debris and
mud flows. In contrast to mass movement models,
they require input hydrographs. Therefore, they allow
accounting more detailed for the onset mechanism,
which plays a crucial role for the flow propagation and
the magnitude of the resulting flood wave (w
aldeR
&
C
osta
, 1996). This type of model is particularly bet-
ter suited for modelling the initial stage and flow path
section of the event that depends more strongly on the
input hydrograph. b
eRtolo
& w
ieCzoRek
(2005) com-
pare models following different concepts for the same
set of debris flows. For an appropriate modelling of the
motion of GLOFs, a combination of mass movement
and river hydraulics models is suggested.
sume an average runout angle of 21° for debris flows
on unobstructed flow paths. H
uGGel
et alii (2003), em-
ploying the Modified Single Flow direction model MSF,
used an angle of 11° proposed by H
aebeRli
(1983) as a
minimum for observed granular debris flows. However,
in the case of the Dasht event, both values underestimate
the maximum travel distance of the debris flow which
reached a runout angle as low as 9.3°. The debris flow
actually did not stop before reaching the main valley.
R
iCkenmann
(1999) suggested the following empirical
relationship for the travel distance of debris flows:
L = 1.9V
0.16
Z
0.83
Eq. 1,
where L is the travel distance of the flow, V is the in-
volved volume, and Z is the loss of elevation. Using
the release volume of 250,000 m
3
in Eq. 1, the Dasht
travel distance is again strongly underestimated, while
the estimated deposition volume of 1.5 million m
3
leads to a travel distance closer to the observation.
However, it is not the ‘fault’ of these empirical
models not to fully capture the Dasht event, but rather a
conceptual problem related to the characteristics of the
event: The GLOF - as many others - was not a classical
debris flow, it was characterized by several flow trans-
formations (hyperconcentrated to debris flow and back).
Semi-deterministic approaches, using a friction
model (e.g. P
eRla
et alii, 1980 for snow avalanches)
in combination with random walk routing techniques
go one step further than strictly empirical models and
are often applied in combination with GIS (e.g. G
amma
,
2000; w
iCHmann
, 2006; m
eRGili
et alii, 2008). They can
be used for back-calculating GLOFs and other types of
mass flows, but are only partly suitable for prediction
purposes. Reliable physically based dynamic models
are therefore required when trying to predict the motion
of potential future mass flows (H
unGR
et alii, 2005).
Fig. 3 - Empirical approaches developed for the reach of debris flows and the observed travel path of the Dasht 2002 event
background image
M. MERGILI, D. SCHNEIDER, R. wORNI & J. F. SCHNEIDER
976
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
breach of the dam followed by a flood wave down
the valley is possible.
The following information was compiled for Kha-
vrazdara as well as for all the other case studies:
• DEMs of different resolution were prepared for
the investigation areas in order to allow the de-
termination of the effect of resolution. SRTM-4
(90 m) was used as well as 10 m and 20 m DEMs
derived from CORONA imagery and 5 m DEMs
derived from WorldView1 imagery.
• Susceptible glacial lakes were identified using
analysis of multitemporal satellite imagery, helicop-
ter surveys, and field investigations. The surface of
the relevant lakes was computed using ASTER ima-
gery and the lake volumes were estimated.
• The peak discharges of potential outburst events
were estimated from empirical rules (e
vans
1986;
C
osta
, 1988; C
osta
& s
CHusteR
, 1988; m
anville
,
2001; H
uGGel
et alii, 2004a). Scenarios of outburst
hydrographs were then created, based on the esti-
mated peak discharge and the lake volume.
• The characteristics of the flow path and the area
of deposition were mapped from satellite imagery,
from the helicopter, and in the field (morphology
of the valley, type of surface material, indicators
for former outburst flood events).
METHODS
The first model used is the physically based mass
movement model RAMMS (Rapid Mass Move-
ments), developed at the WSL Institute for Snow
and Avalanche Research SLF Davos, Switzerland
(see C
HRisten
et alii, 2010a, 2010b for a more de-
tailed description and for case studies). The frictional
resistance S is based on the v
oellmy
(1955) model
that combines dry Coulomb friction μ with a velocity-
squared dependent turbulent friction ξ.
where g is the gravitational acceleration, H is the flow
depth, φ the slope angle, and U is the depth-averaged
flow velocity. The maximum velocity U
max
is defined
by v
oellmy
(1955) as:
If μ equals zero, Eq. 3 can be further transformed
OBJECTIVES
The general objective of the study presented was
to elaborate a way to estimate the travel distance and
travel times of potential future GLOFs by comparing
the results of two different models for mass flows.
Each of them partially represents certain characteris-
tics of GLOFs but cannot fully reproduce the flow be-
haviour. The results and the model settings and param-
eters suitable for GLOFs, but also the needs for further
research and model development are high-lighted, us-
ing examples from the Pamir (Tajikistan).
The paper concentrates on the movement of the
flood wave itself, the breaching process of the dam is
not considered. For the on-set of the GLOF process, sce-
narios for the outburst volume and hydrograph, as well
as for the finally deposited volume (including entrained
debris) were elaborated. The scenarios are based on the
lake volume, the dam characteristics, and the suscepti-
bility to rock and ice avalanches into the lake.
STUDY AREAS AND DATA
The modelling was performed for five study areas
in Tajikistan (one for back-calculation, four for pre-
diction; Fig. 4). All areas are located in the Pamir, a
heavily glaciated high mountain area culminating in
7,495 m a.s.l. The lakes used in the case studies are
distributed between 3,800 m and 4,800 m a.s.l.
The results for Khavrazdara, a Northern tribu-
tary of the Bartang Valley, will be discussed in detail
below. 20 km upstream from the valley outlet, the
tongue of a rock glacier dams a lake with a surface of
2 km², approximately, and an estimated volume of 40
million m³ (Fig. 2b). In the case of a climate-change-
induced degradation of the rock glacier tongue, a
Fig. 4 - Map of Tajikistan with the five selected areas for
modelling
Eq. 2
Eq. 3
background image
GLACIAL LAKE OUTBURST FLOODS IN THE PAMIR OF TAJIKISTAN:
CHALLENGES IN PREDICTION AND MODELLING
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
977
The following work flow was applied for the
modelling:
1. Back-calculation of a well-documented recent
GLOF: the Dasht event from summer 2002 was
used to test the models and to find suitable parame-
ter values. Travel distance, the spatial distribution of
the deposit, and the travel time from the start to the
deposit were used as reference for the calibration;
2. Scenarios of possible future outburst events of se-
lected lakes (see Table 1 for an example) were ela-
borated. Outburst volumes, peak discharges, and
flow rheologies were varied among the different
scenarios. The friction parameters of RAMMS
with the best fit for Dasht were taken as a referen-
ce, but adapted according to the outburst volumes
and water content so that the worst case scenario
reached the former debris flow fan of the main
valley (compare Discussion and Conclusions).
3. The scenarios were run with RAMMS and FLO-
2D. The resolution of the DEM and the computa-
tion were varied in order to estimate the influence
of this setting on the model results.
RESULTS
BACk-CALCULATION FOR DASHT
First, the Dasht (2002) event was back-calcu-
lated using RAMMS (Fig. 5). The purpose was to
calibrate the model for this type of event and to find
suitable values for the friction parameters μ and ξ.
The model was run on the CORONA DEM (10 m),
and on the SRTM-4 DEM (90 m) with a calculation
resolution of 20 m in order to figure out the influ-
ence of different levels of smoothing of the terrain.
into the Chézy equation. Therefore, by applying low
μ-values, an approximation to turbulent clear water
open channel flow can be established.
RAMMS was originally designed to predict the
maximum travel distance and velocity of snow ava-
lanches. Calibrated parameters are available for this
type of process. They are only valid for the front of
the avalanche, so that the deposition geometry can-
not be predicted in a straightforward way (C
HRisten
et alii, 2010a). The model is able to compute entrain-
ment of material by the flow, governed by an empiri-
cally determined scaling factor and an entrainment
law. RAMMS has recently been used for modelling
other types of mass movements. s
CHneideR
et alii (ac-
cepted) successfully used it for the back-calculation
of large rock-ice avalanches and P
ReutH
et alii (in
press) simulated various large rock avalanches in the
European Alps. It has further been used for the simula-
tion of debris flows in Switzerland (n
aef
et alii, 2006;
R
iCkenmann
et alii, 2006; a
Rmento
et alii, 2008) but
not yet for modelling GLOFs.
The second model - FLO-2D - was developed
by J. O’Brien (e.g. o’b
Rien
et alii, 1993; o’b
Rien
,
2001). It is a volume conserving model for flow rout-
ing of clear water floods, hyperconcentrated flows,
or debris flows over floodplains or through confined
channels. Topography, input hydrograph, and resist-
ance to flow determine the flow behaviour. Case
studies are provided e.g. by H
uebl
& s
teinwendtneR
(2001) or b
eRtolo
& w
ieCzoRek
(2005). For clear
water flow, the governing equations are:
where h is the flow depth, U is the depth-averaged
flow velocity in one flow direction x, i is rainfall in-
tensity, S
f
is the friction slope component (based on
Manning’s Equation), and α is the bed slope.
Both programs - RAMMS and FLO-2D - need a
DEM as input. RAMMS further requires the spatial
distribution and depth of the release volume, the co-
efficients and possible areas for entrainment, and the
friction parameters μ and ξ. FLO-2D needs an input
hydrograph and values of Manning’s n. When using
FLO-2D for debris flow modelling the rheologic flow
parameters viscosity and yield stress must be specified.
Eq. 4,
Eq. 5,
Fig. 5 - Back-calculation of the Dasht 2002 event using
RAMMS
background image
M. MERGILI, D. SCHNEIDER, R. wORNI & J. F. SCHNEIDER
978
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
Friction parameters of μ = 0.14 and ξ = 1,300 proved
to be the best guess for reconstructing the event,
though it was necessary to assume lower values of μ
(0.01) and higher values of ξ (2,000) in the flat start-
ing area (representing the lake surface) in order to
initiate the movement. The velocities and the extent
of spreading in the area of deposit were larger when
using the SRTM DEM (smoother terrain).
The simulated travel time from the onset of the flow to
the village was 55 minutes with the SRTM DEM and
76 minutes with the CORONA DEM. These values
correspond reasonably with local reports concerning
the time difference between the acoustic detection of
the GLOF and its arrival at the village.
As the GLOF event in Dasht propagated as a debris
flow, this case study was used to define the rheologic pa-
rameters for debris flow modelling in FLO-2D. It was
found that values for viscosity η = 279 poises and yield
stress τ = 798 dynes/cm
2
represented best the debris flow
in Dasht. Consequently these values were also used in
the scenario modelling. FLO-2D was run on the CORO-
NA DEM only. The simulated travel time, flow heights
and extent matched well with the field observations.
kHAVRAZDARA
Different scenarios for lake outburst floods were
then computed for Khavrazdara, Varshedzdara, Up-
per Rivakdara, and Rivakkul (see Fig. 4). The mod-
elling results for Khavrazdara (see Fig. 2b) are dis-
cussed in detail.
The scenarios defined for an outburst flood in
Khavrazdara are shown in Table 1. A cell size of 20
m was used for the RAMMS simulations and 40 m
for the FLO-2D simulations, respectively. Whilst the
GLOFs simulated with FLO-2D reached the outlet of
the valley, the RAMMS simulations indicated a stop
of the flow in the middle portion of the valley when
using the friction parameters calibrated with the Dasht
event. It was then tested how much the friction would
have to be reduced to allow the flow to reach the val-
ley outlet and to cover the debris cone there. Friction
values of μ = 0.04 and ξ = 1,000 were found to be suit-
able. Decreased μ-values of μ = 0.03 were used to ac-
count for the lower sediment concentration expected
in the upper section (before entrainment takes place),
whilst increased values of μ = 0.05 were applied to ac-
count for the higher sediment concentration expected
in the lower section. ξ-values were held constant for
the entire flow path. The spatial distribution of the
maximum flow height simulated with RAMMS and
FLO-2D for selected scenarios is illustrated in Fig. 6.
With FLO-2D all scenarios were modelled as a
hyperconcentrated flow with a volumetric sediment
concentration of 20% on the one hand, and as a de-
bris flow with volumetric sediment concentrations up
to 50% on the other hand. Applying a range of dif-
ferent flow rheologies is a strategy to deal with the
uncertainty regarding the flow type produced by the
released water from a lake. The amount of water is the
same in both flow types, but the debris flow is bulked
with much more material. This is why its total flow
volume and peak discharge are higher than for hyper-
concentrated flows of the same outburst scenario. As
the peak discharge has the highest influence on calcu-
lated maximum flow depths, and the higher viscosity
results in lower flow velocities, inundation depths are
higher when modelling the GLOF as a debris flow.
According to the simulation, the flow would reach
the village of Pasor at the outlet of the valley between 48
minutes and 4.5 hours after the onset, depending on the
scenario (see Table 1; average velocity between 1 and
6 m/s, respectively). This wide range shows the uncer-
Fig. 6 - Maximum flow depth computed with FLO-2D and
RAMMS for dif-ferent lake outburst scenarios of
khavrazdara
background image
GLACIAL LAKE OUTBURST FLOODS IN THE PAMIR OF TAJIKISTAN:
CHALLENGES IN PREDICTION AND MODELLING
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
979
valley outlet) provides useful information in two ways:
• A comparison of the assumed friction parameter
values with those derived from the back-calcu-
lation of documented events allows an asses-
sment of how realistic the assumed parameters
are, and therefore the likelihood of the flow to
reach the outlet of the respective tributary val-
ley. Table 2 shows the friction parameters used
in the RAMMS calculation, based on the assum-
ption that the flow would reach the outlet of the
respective valley.
• Approximate travel times to the outlet can be de-
rived, given that the assumed parameters are con-
sidered as realistic.
A special characteristic of the RAMMS model is
the sudden release of the start volume (mass), that well
represents a sudden mechanical failure of a lake dam
or the overtopping of a large impact wave. However,
this is not always the way how GLOFs are triggered
and may therefore lead to exaggerated flow heights and
widths in the upper flow section. In contrast, the abil-
ity to erode material from the ground is an important
feature of RAMMS because it accounts for the often
observed fact that start and end volumes differ signifi-
cantly (b
eRti
et alii, 1999, b
Reien
et alii, 2008).
In general, RAMMS predicts higher values of
flow depth in the uppermost section of the flow path
than FLO-2D. This is due to the sudden release of
the mass (see above). FLO-2D makes use of an input
hydrograph that distributes the release volume over a
given time period, leading to lower flow depths for
given total volumes. This can better reproduce a dam
failure due to progressive incision.
RAMMS predicts the stop of the flow and the
deposition of the mass on the debris cone whilst FLO-
2D tends to predict a continuation of the flow along
the stream path of the main valley. The potential im-
pact areas derived with RAMMS are therefore smaller
and the inundation depths are larger than those calcu-
lated with FLO-2D.
Partially good correspondence is found in the flow
durations to the outlet of the valley. Table 1 shows
that the range of the flow durations calculated with
RAMMS are similar to those derived from the FLO-
2D calculations, at least regarding the - more critical
- lower boundary. This is remarkable because they are
computed completely independently (the adaptation
of μ and ξ is a purely frictional issue).
tainties connected to the scenarios, topographic data and
parameters used. The maximum velocity ranges around
10 m/s over most of the valley, with much higher values
yielded in the upper portion, particularly by RAMMS.
The influence of the DEM resolution on the model re-
sults is considerable: finer DEM resolutions generally
lead to a rougher surface which significantly reduces the
flow velocity and hence the reach of the debris flow (see
also C
HRisten
et alii, 2010a). The smoother the original
terrain is, the less this effect is observed.
DISCUSSION AND CONCLUSIONS
The present study illustrates that modelling of
GLOFs remains a challenge. Each case study area has
its individual characteristics and the results provided
by different model approaches sometimes diverged
considerably. These differences are not surprising as
the two models follow disparate concepts, each requir-
ing a specific definition of the initial conditions (sud-
den release of mass vs. discharge curve). In order to
homogenize the results and to account for the generally
larger amount of water, the friction parameters used in
RAMMS had to be reduced considerably in comparison
to those used for the Dasht event. This is of very high
importance when the results are interpreted or shown
to local authorities. However, adapting the friction pa-
rameters individually for each study area in a way that
the simulated flow reaches the area of interest (often the
Tab. 1 - Modelling of a potential GLOF from the large lake
in the upper khavrazdara: Scenarios, involved vol-
umes, maximum discharge, and travel times to the
village of Pasor (outlet of the valley). Hflw = Hyper-
concentrated flow, Dflw = Debris flow, C = CORO-
NA DEM , S = SRTM-4 DEM, N/A = not applicable
background image
M. MERGILI, D. SCHNEIDER, R. wORNI & J. F. SCHNEIDER
980
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
One has to conclude that, considering all relevant
aspects, the simulation of the motion of potential fu-
ture GLOFs remains a big challenge. Problems are
in particular:
• The knowledge about the onset of the process is of-
ten limited (properties of dam, type of dam breach,
understanding of process chains and interactions).
• The volume of water involved in the outburst flo-
od is unclear. The lake bathymetry is often unk-
nown and may change rapidly, whilst the ratio of
water actually bursting out has to be estimated.
Furthermore many lakes burst out within a short
time after their development without being detec-
ted as potential source of hazard (n
aRama
et alii,
2010). Continuous monitoring is required to keep
updated on the existing hazards.
• Uncertainties related to erosion and deposition are
a big unresolved issue. Erosion of the dam and the
bed as well as concomitant deposition can stron-
gly change the rheology and the moving volume
of the flow. These changes have a direct impact on
the spreading and reach.
• The flow transformation processes of the natural
phenomena are a challenge for the models (and in
general for any assessment). Software developed by
the hydrological community is specialized to simu-
late floods or hyperconcentrated flows with input
hydrographs on moderately steep flow channels and
with lower sediment loads. In contrast to this, pro-
grams for rapid mass movements are better suited
for steeper slopes and sudden failure of the initial
volume. The typical characteristics of GLOFs are
in between and vary for different channel sections.
Sediment transport models properly computing ero-
sion and deposition are rather designed for less steep
slopes, so that they are hardly applicable to GLOFs.
Furthermore, the outburst scenario is very critical.
Flood dynamics are quite well understood and
model results can therefore be considered as confi-
dent. In contrast, debris flow modelling is a based on
empirical components and the results are therefore
more inaccurate compared to modelling pure water or
hyperconcentrated flows.
Nevertheless it is important not to model only the
outburst scenarios as hyperconcentrated flows, but
also as debris flows. With such a modelling strategy
a range of expectable flow rheologies can be covered.
This increases the robustness of the results and does
not pretend a wrong accuracy.
Existing programs also largely fail to simulate
process interactions and transformations such as the
development of a hyperconcentrated flow into a debris
flow, the effects of multiple flood waves (including the
modified topography after the first wave), or the ef-
fects of short-term storage of water and debris by self-
induced blockage of the valley.
Considering all these points, it has to be concluded
that up to now, no well suitable modelling approaches
do exist for GLOFs, as these represent highly variable
phenomena and often exhibit a behaviour in between de-
bris flows and floods. However, applying a combination
of different model approaches, as attempted in the study
presented, helps to estimate realistic process magnitudes,
areas of impact, maximum velocities, and travel times.
As a general conclusion for any kind of modelling effort,
a responsible interpretation of the results and a controlled
knowledge transfer to local authorithies is crucial.
ACKNOWLEDGEMENTS
The studies were supported by the FOCUS Hu-
manitarian Assistance, an affiliate of the Aga Khan
Development Network, by the Swiss Agency for De-
velopment and Cooperation (SDC), and the British
Department for International Development (DFID).
Special thanks go to Christian Huggel for his highly
valuable comments on the draft manuscript.
Tab. 2 - Involved volumes, valley characteristics, and friction parameters chosen for the RAMMS simulation. L = length of
valley; ΔZ = loss of elevation; φ
avg
= average inclination.
a
start volume
b
end volume * 1
st
section of flow path (upper)
** 2
nd
section of flow path (middle) *** 3
rd
section of flow path (lower)
background image
GLACIAL LAKE OUTBURST FLOODS IN THE PAMIR OF TAJIKISTAN:
CHALLENGES IN PREDICTION AND MODELLING
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
981
REFERENCES
a
Rmento
m.C., G
enevois
R. & t
eCCa
P.R. (2008) - Comparison of numerical models of two debris flows in the Cortina d’Ampezzo
area, Dolo-mites, Italy. Landslides, 5: 143-150.
b
eRti
m. & s
imoni
a. (2007) - Prediction of debris flow inundation areas using empirical mobility relationships. Geomorphology,
90: 144-161.
b
eRti
m., G
enevois
R., s
imoni
a. & t
eCCa
P.R. (1999) - Field observations of a debris flow event in the Dolomites. Geomorphology,
29: 265-274.
b
eRtolo
P. & w
ieCzoRek
G.f. (2005) - Calibration of numerical models for small debris flows in Yosemite Valley, California,
USA. Nat. Haz. Earth Syst. Sci., 5: 993-1001.
b
Reien
H., d
e
b
lasio
f.v., e
lveRHoi
a. & H
oeG
k. (2008) - Erosion and morphology of a debris flow caused by a glacial lake
outburst flood, western Norway. Landslides, 5 (3): 271-280.
C
aRey
M. (2005) - Living and dying with glaciers: people’s historical vulnerability to avalanches and outburst floods in Peru.
Glob. Plan. Change, 47: 122-134.
C
endeRelli
d.a. & w
oHl
e.e. (2001) - Peak discharge estimates of glacial lake outburst floods and “normal” climatic floods in
Mount Everest region, Nepal. Geomorphology, 40: 57-90.
C
HRisten
C., b
aRtelt
P. & k
owalski
J. (2010a) - Back calculation of the In den Arelen avalanche with RAMMS: interpretation
of model results. Ann. Glac., 51 (54): 161-168.
C
HRisten
m., k
owalski
J. & b
aRtelt
, b. (2010b) - RAMMS: Numerical simulation of dense snow avalanches in three-dimensional
terrain. Cold Reg. Sci. Technol., 63:1-14.
CIAT (i
nteRnational
C
entRe
foR
t
RoPiCal
a
GRiCultuRe
) (2004) - Void-filled seamless SRTM data V1, available from the CGIAR-
CSI SRTM 90 m Database. http://srtm.csi.cgiar.org
C
laRke
G.K.C. (1982) - Glacier outburst floods from ‘Hazard Lake,’ Yukon Territory, and the problem of flood magnitude
prediction. J. G
laC
, 28 (98): 3-21.
C
oRominas
J., C
oPons
R., v
ilaPlana
J.m., a
ltamiR
J. & a
miGó
J. (2003) - Integrated Landslide Susceptibility Analysis and
Hazard Assessment in the Principality of Andorra. Nat. Haz., 30: 421-435.
C
osta
J.e. & s
CHusteR
R.l. (1988) - The formation and failure of natural dams. Geol. Soc. Am. Bull., 7: 1054-1068.
C
osta
J.E. (1988) - Floods from dam failures. In: b
akeR
V.R. (1988, ed.) - Flood Geomorphology: 439-463, Wiley, New York.
d
ussaillant
a., b
enito
G., b
uytaeRt
w., C
aRlinG
P., m
eieR
C. & e
sPinoza
f. (2009) - Repeated glacial-lake outburst floods in
Patagonia: an in-creasing hazard? Nat. Haz., DOI 10.1007/s11069-009-9479-8.
e
vans
s.G. & C
laGue
J.J. (1994) - Recent climatic change and catastrophic geomorphic processes in mountain environments.
Geomorphology 10 (1-4): 107-128.
e
vans
s.G. (1986) - The maximum discharge of outburst floods caused by the breaching of man-made and natural dams. Can.
Geotech. J., 23: 385-387.
f
Rey
H., H
aebeRli
w., l
insbaueR
a., H
uGGel
C. & P
aul
f. (2010) - A multi-level strategy for anticipating future glacier lake
formation and asso-ciated hazard potentials. Nat. Haz. Earth Syst. Sci., 10: 339-352.
G
amma
P. (2000) - dfwalk - Ein Murgang-Simulationsprogramm zur Gefahrenzonierung. Geographica Bernensia G66. 144 pp.
[in German].
H
aebeRli
W. (1983) - Frequency and characteristics of glacier floods in the Swiss Alps. Ann. Glac., 4: 85-90.
H
ewitt
K. (1982) - Natural dams and outburst floods in the karakorum Himalaya. In: G
len
J.W. (1982, ed.) - Hydrological
aspects of alpine and high-mountain areas. IAHS Publication 138: 259-269.
H
uebl
J & s
teinwendtneR
H. (2001) - Two-dimensional simulation of two viscous debris flows in Austria. Phys. Chem. Earth
(C), 26: 639-644.
H
uGGel
C., H
aebeRli
w., k
aeaeb
a., b
ieRi
d. & R
iCHaRdson
s. (2004a) - An assessment procedure for glacial hazards in the
Swiss Alps. Can. Geotech. J., 41(6): 1068-1083.
H
uGGel
C., k
aeaeb
a. & s
alzmann
n. (2004b) - GIS-based modeling of glacial hazards and their interactions using Landsat-
TM and IkONOS imagery. Norsk Geografisk tidsskrift / Norwegian J. Geogr., 58: 761-773.
H
uGGel
C., k
aeaeb
a., H
aebeRli
w. & k
RummenaCHeR
b. (2003) - Regional-scale GIS-models for assessment of hazards from
glacier lake out-bursts: evaluation and application in the Swiss Alps. Nat. Haz. Earth Syst. Sci., 3: 647-662.
background image
M. MERGILI, D. SCHNEIDER, R. wORNI & J. F. SCHNEIDER
982
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
H
unGR
o., C
oRominas
J. & e
beRHaRdt
e. (2005) - State of the Art paper: Estimating landslide motion mechanism, travel distance
and velocity. In: H
unGR
o., f
ell
R., C
outuRe
R., e
beRHaRdt
E. (2005, eds.) - Landslide Risk Management. Proceedings of
the International Conference on Landslide Risk Management, Vancouver, Canada, 31 May - 3 June 2005, 129-158.
IPCC (2007) - Climate Change 2007: Synthesis Report, Intergovernmental Panel on Climate Change, 26-73.
i
tuRRizaGa
L. (2005a) - Historical Glacier-dammed Lakes and Outburst Floods in the karambar Valley (Hindukush-karakoram).
GeoJournal, 62 (3-4): 1-47.
i
tuRRizaGa
L. (2005b) - New observations on present and prehistorical glacier-dammed lakes in the Shimshal valley (karakoram
Mountains). J. Asian Earth Sci., 25 (4): 545-555.
i
veRson
R.m., s
CHillinG
s.P. & v
allanCe
J.w. (1998) - Objective delineation of lahar-inundation hazard zones. GSA Bull., 110
(8): 972-984.
k
aeaeb
a., H
uGGel
C., f
isCHeR
l., G
uex
s., P
aul
f., R
oeR
i., s
alzmann
n., s
CHlaefli
s., s
CHmutz
k., s
CHneideR
d., s
tRozzi
t.
& w
eidmann
y. (2005) - Remote sensing of glacier- and permafrost-related hazards in high mountains: an overview. Nat.
Haz. Earth Syst. Sci., 5: 527-554.
m
anville
V. (2001) - Techniques for evaluating the size of potential dam-break floods from natural dams. Institute of Geological
and Nuclear Sciences Science Report 2001/28, 72 pp.
m
eRGili
m., f
ellin
w., m
oReiRas
s.m. & s
toetteR
J. (2008) - Integrated modelling of debris flows in the Central Andes based
on Open Source GIS. Geophys. Res. Abstr., 10. [abstract]
n
aef
d., R
iCkenmann
d., R
utsCHmann
P. & m
C
a
Rdell
b.w. (2006) - Comparison of flow resistance relations for debris flows
using a one-dimensional finite element simulation model. Nat. Haz. Earth Syst. Sci., 6: 155-165.
n
aRama
C., d
uisHonakunov
m., k
aeaeb
a., d
aiRov
m. & a
bdRakHmatov
k. (2009) - The 24 July 2008 outburst flood from
western Zyndan glacier lake, Tien Shan, kyrgyzstan. Nat. Haz. Earth Syst. Sci., doi:10.5194/nhess-10-647-2010.
o’b
Rien
J.S. (2001) - FLO-2D Users Manual. Nutrioso, Arizona. 83 pp. and appendices.
o’b
Rien
J.s., J
ulien
, P.y. & f
ulleRton
w.t. (1993) - Two-dimensional water flood and mudflow simulation, J. Hydraul. Eng.,
119: 244-261.
P
eRla
R., C
HenG
t.t. & m
C
C
lunG
d.m. (1980) - A Two-Parameter Model of Snow Avalanche Motion. Journal of Glaciology,
26: 197-207.
P
ReutH
t., b
aRtelt
P., k
oRuP
o. & m
C
a
Rdell
b.w. (in press) - Fragmentation and frictional relaxation in rock avalanches. J.
Geophys. Res.
Q
uinCey
d.J., R
iCHaRdson
s.d., l
uCkman
a., l
uCas
R.m., R
eynolds
J.m., H
ambRey
m.J. & G
lasseR
n.f. (2007) - Early
recognition of glacial lake hazards in the Himalaya using remote sensing datasets. Glob. Plan. Change, 56 (1-2): 137-152.
R
iCHaRdson
s.d. & R
eynolds
J.m. (2000) - An overview of glacial hazards in the Himalayas. Quatern. Int., 65/66: 31-47.
R
iCkenmann
D. (1999) - Empirical Relationships for Debris Flows. Nat. Haz., 19: 47-77.
R
iCkenmann
d., l
aiGle
d., m
C
a
Rdell
b.w. & H
uebl
J. (2006) - Comparison of 2D debris-flow simulation models with field
events. Computational Geosciences, 10: 241-264.
s
CHneideR
d., b
aRtelt
P., C
aPlan
-a
ueRbaCH
J., C
HRisten
m., H
uGGel
C., & m
C
a
Rdell
b. w. (accepted) - Insights into rock-ice
avalanche dynamics by combined analysis of seismic recordings and a numerical avalanche model. J. Geophys. Res.
s
CHneideR
J. f., G
meindl
m. & t
RaxleR
k. (2004) - Risk Assessment of Remote Geohazards in Central and Southern Pamir/GBAO,
Tajikistan. Report to the Ministry of Emergency, Tajikistan and the Swiss Agency for Development and Cooperation (SDC).
t
inti
s., m
aRamai
a. & C
eRutti
a.v. (1999) - The Miage Glacier in the Valley of Aosta (western Alps, Italy) and the extraordinary
detachment which occurred on August 9, 1996. Phys. Chem. Earth (A), 24 (2): 157-161.
v
ilímek
v., z
aPata
m.l., k
limes
J., P
atzelt
z., & s
antillán
n. (2005) - Influence of glacial retreat on natural hazards of the
Palcacocha Lake area, Peru. Landslides, 2 (2): 107-115.
v
oellmy
A. (1955) - Ueber die Zerstoerungskraft von Lawinen. Schweizerische Bauzeitung, 73: 159-162, 212-217, 246-249,
280-285. [in German].
w
aldeR
J.s. & C
osta
J.e. (1996) - Outburst floods from glacierdammed lakes: the effect of mode of lake drainage on flood
magnitude. Earth Surf. Proc. Landforms, 21: 701-723.
w
atanabe
t. & R
otHaCHeR
d. (1996) - The 1994 Lugge Tsho glacial lake outburst flood, Bhutan Himalaya. Mt. Res. Dev., 16:
77-81.
Statistics