J. Mt. Sci.
(2013) 10(2): 293–304
Comparison between FLO-2D and Debris-2D on the
Application of Assessment of Granular Debris Flow
Hazards with Case Study
WU Ying-Hsin1, LIU Ko-Fei1,2,*, CHEN Yi-Chin2
1 Department of Civil Engineering, National Taiwan University, 10617 Taipei, Chinese Taipei
2 Hydrotech Research Institute, National Taiwan University, 10617 Taipei, Chinese Taipei
*Corresponding author, e-mail: kfliu@ntu.edu.tw; First author, e-mail: wu.ahsin@gmail.com
© Science Press and Institute of Mountain Hazards and Environment, CAS and Springer-Verlag Berlin Heidelberg 2013
Abstract: Numerical simulation has been widely
applied to the assessment of debris flow hazards. In
East Asia and especially Taiwan, the most widely used
numerical programs are FLO-2D and Debris-2D.
Although these two programs are applied to the same
engineering tasks, they are different in many aspects.
These two programs were compared according to
their fundamental theories, input and output data,
computational algorithms and results. Using both
programs, the simulations of a real debris flow with
abundant granular material induced by landslides at
Xinfa village in southern Taiwan are performed for
comparison. The simulation results show that Debris-
2D gives better assessment in hazard area delineating
and flow depth predicting. Therefore, Debris-2D is
better for simulation of granular debris flows.
Keywords: Debris flow assessment; Program
comparison; FLO-2D; Debris-2D
Introduction
In debris flow hazard assessment, numerical
simulation is widely used for delineation of the
affected area. FLO-2D and Debris-2D (Liu and
Huang 2006) programs are the most popular one
in Taiwan. FLO-2D is commercially available
software for two-dimensional flood or mud-flood
routing, and is published by FLO-2D Software Inc.
since 1993. FLO-2D can simulate flood and mud
Received: 25 October 2012
Accepted: 8 January 2013 
DOI: 10.1007/s11629-013-2511-1
flows on different surfaces, and then produce the
temporal variation of flow depth, flow velocity and
affected area. These surfaces include plains, creeks,
alluvial fans, channels, or other artificial ones.
FLO-2D has been approved as a hydraulic program
for flood and debris flow simulation by Federal
Emergency Management Agency of the United
States (Flo-2D Software, Inc. 2006).
As the only commercially available tool, FLO-
2D has been widely applied for the assessment of
debris flow hazard since 1993. It was first applied
to simulate the mudflow event of the Rudd creek in
Utah, USA on urbanized area (O’Brien et al. 1993).
However, the simulation gave unreasonable
affected area compared with field measurements.
In most cases, to use FLO-2D, one needs to
calibrate several hydraulic parameters as well as
the discharge hydrograph. This calibration
procedure sometimes produces accepted results on
the spread but not the flow depth for debris flows.
Chen et al. (2004) simulated the affected zone of
the debris flow induced by breaching of a landslide
dam; Lin et al. (2005) assessed the debris flow
hazard of the Chuishe creek in the Chen-you-lan
watershed, Nantou, Taiwan. Bertolo and Wieczorek
(2005) analyzed the debris flows in the Yosemite
National Park, California, USA. Četina et al. (2006)
conducted the numerical simulation of debris flow
induced by landslides in the Stože Mountain in
northwest Slovenia. Hsu et al. (2010) assessed and
delineated the debris-flow hazard zone in Hualien,
293
J. Mt. Sci. (2013) 10(2): 293–304

Taiwan. But even with calibrated inputs, the eight flowing directions defined in FLO-2D
simulated results may not match with the field as program. They are east, west, south, north,
in Boniello et al. (2010) or Stolz and Huggel (2008). southeast, southwest, northeast and northwest
With the affected area obtained, many has applied directions. In each direction, the constant density
the result to evacuation planning and vulnerability fluid is assumed to be hydrostatic. The governing
and risk assessment (e.g., Chen et al. 2004; Lin et equations are continuity and dynamic wave
al. 2008; Chen et al. 2009; Lin et al. 2011b; Quan momentum equation, in each direction. For
Luna et al. 2011). But there is no record for FLO- example, the x-direction government equations are
2D to make any successful predictions of a hazard H   uH 
 i, (1)
before it occurs. t x
Unlike FLO-2D which can be applied for flood H u u 1 u
S f  S0    , (2)
routing, Debris-2D is only used for two- x g x g t
dimensional debris flow simulation and delineation. where x and t are spatial and temporal independent
Debris-2D program was developed in 2006. Liu variables respectively. Dependent variables H and
and Huang (2006) used this program to simulate u are flowing depth and depth-averaged velocity in
the debris flow hazard in Nantou County, Taiwan. x-direction. g is the gravitational acceleration. Sf
Compared with the field measurements, the and S0 are bottom friction and bed slope
simulation results had less than 5% error in respectively. i is excess rainfall intensity. If there is
affected area as well as the final height of no rainfall, i.e., i=0, Eqs. (1) and (2) become the
deposition. Then, Liu and Wu (2010) assessed the Saint-Venant equations.
debris flow event in Inje, northeast South Korea in Friction slope Sf is calculated using shear
2006, and the simulation result of flow depth and stresses. In modeling mud or debris flow, there are
distribution agreed with the field investigation very five components in shear stress
well. There have been several successful    c   mc   v   t   d , (3)
predictions by Debris-2D. For example, Debris-2D where τc is cohesive yield stress; τmc is Mohr-
was applied to assess the debris flow affected area Coulomb shear; τv is viscous shear stress; τt and τd
in Daniao tribe southeast Taiwan in 2006 and the are turbulent and dispersive shear stress
simulation agrees very well with the real event respectively. The constitutive law for shear stress
occurred in 2009 (Tsai et al. 2011). No calibration and strain-rate used is
is needed to use Debris-2D, all parameters must be 2
 u   u 
decided according to physical procedures. Debris-   0     C  y  , (4)

 y  
2D is also applied for vulnerability risk assessment
(Liu and Lee 2007; Tsai et al. 2011) and mitigation where τ0 is yield stress and τ0 = τc + τmc; η is
design evaluation (Liu et al. 2009). dynamic viscosity (poise); the turbulent- dispersive
In this study, we first introduce the governing coefficient C is
1
equations, input, computational algorithm and  6 3 2
  sin  I   s  1  en  C v ,
1
C  m l 2  2 3
(5)
input data for FLO-2D and Debris-2D separately. 12   
Then, a real case is simulated for comparison. where ρs is density of debris; ρm is density of
mixture; en is energy restitution coefficient; αI is
averaged impact angle of solid. Equation (4) is
1 Introduction to FLO-2D similar to the model proposed by Julien and Lan
(1991), but Julien and Lan model gives zero strain
The text written in this section comes from the rate for stress less than yield stress while Equation
user manual of FLO-2D (FLO-2D Software, Inc. (4) does not. From Equation (4) the friction slope
2006), but only the portion related to debris flow is Sf can be obtained as
mentioned. y K u n 2 u2
Sf    td 4
 m H 8 m H 2
 
  H 3 , 
(6)
Sy
1.1 Governing Equations Sv Std

where Sy and Sv are yield stress and viscous slope

In horizontal two-dimensional plane, there are respectively, and Sy + Sv represents the effects of

294

yield stress and viscosity; Std is turbulent-
dispersive slope and stands for the collision effects
between solids; K is resistance parameter; τy is
yield stress, η is dynamic viscosity and ntd is
turbulent flow resistance, and they can be
expressed as
 y   2 exp   2 C v  , (7)
   1 exp  1 C v  , (8)
ntd  nt b exp  mC v  ,
(9)
where b=0.0538 and m=6.0896; nt is turbulent n-
value. The values of K, αi and βi (i=1, 2) can be
found in the user manual (FLO-2D Software, Inc.
2006).
1.2 Program inputs
For simulating debris flow, the main inputs are
topography and friction coefficients. FLO-2D uses
digital elevation model (DEM) with square uniform
grids. In each gird, constant Manning’s n value is
used for bottom friction. No bottom erosion or
deposition is considered. Man-made structure can
be input into FLO-2D, the effect can be included in
the simulation.
The inputs for debris-flow are rainfall
hydrograph or debris-flow discharge hydrographs
with corresponding inflow locations and
rheological parameters. The rainfall hydrograph is
obtained through rainfall record during the debris-
flow event. The debris-flow discharge hydrograph
includes water runoff hydrograph and the temporal
variation of solid volume concentration Cv (%). The
volume concentration Cv ranges from 0.45 to 0.55
according to user manual (FLO-2D Software, Inc.
2006). The fixed inflow location corresponding to
each debris-flow inflow has to be prescribed. Every
grid element can be assigned to be the inflow
location. The needed rheological parameter inputs
include specific weight γm, yield stress τy, dynamic
viscosity η, and resistance parameter K. In the limit
of water, user can input Cv = 0. Then from
Equation (7), τy = α2. The value of α2 can be found
from user manual of FLO-2D, and the value is
between 0.1 and 0.01.
1.3 Computational algorithm for FLO-2D
FLO-2D uses central difference method and
Newton-Raphson method to solve the depth-
J. Mt. Sci. (2013) 10(2): 293–304
averaged velocity in each flowing direction. For a
gird element, the discharge Qi across the boundary
in each direction is computed by multiplying the
velocity and cross sectional flowing area. Then
summing the discharges in eight directions, the net
change in volume can be obtained as below:
8
Qit 1   Qit 1 , (10)
i 1
where superscript t+1 means the next time step;
the subscript i represents the eight different
flowing directions. The net change of flow depth at
next time step is
Qit  1 t
H it  1  , (11)
Asurf
where Δt and Aswf are time step and surface area of
the grid element respectively. With Equation (11) and
the value of flow depth from previous time step, flow
depth at next time step (i.e., t+1) can be obtained. The
averaging in Equation (10) can also be used to
smooth shock wave or hydraulic jump during the
computation (FLO-2D Software, Inc. 2006).
According to the variation of inflow
hydrograph, the time step Δt will be changed. For a
steep rising hydrograph, the time step is decreased
for the stability; otherwise, time step can be
increased to speed up the computation (FLO-2D
Software, Inc. 2006). The termination mechanism
for simulation is the maximum simulation time
inputted by user. There is no physical mechanism
or other conditions in program to terminate the
simulation.
1.4 Outputs
The default output time interval is hour.
Output in minutes can be done with the finest time
resolution. The output data includes temporal
variation of flow depth, flow velocity and impact
force. Maximum flow depth and velocities are also
recorded. FLO-2D has a very convenient and
friendly user interface. All the output can be
converted into ESRI shapefile automatically for
displaying in geographic information system (GIS).
2 Introduction to Debris-2D
2.1 Governing equations
The governing equations are mass and
295
J. Mt. Sci. (2013) 10(2): 293–304

momentum conservation with shallow water they will be involved in the future version. Since
assumption. The coordinate system is the Cartesian the bottom shear layer is ignored, the yield stress
coordinate with the average bed elevation as x axis. becomes the dominant bottom stress.
The used constitutive relation is the 3-D With nonlinear treatment, Debris-2D can
generalization of Julien and Lan (1991). derive the initiation criterion for any originally
stationary debris pile. This is the initiation
 0 
 ij    d  c  II   ij , for  II   0 , (12) condition
  II  2 2
 B H   B H 
 x  x  tan    y  y 
 II  0 , for  II   0 , (13)    
2
(17)
where  0 
  .
12 12   g cos  H 
1  1 
 II    ij  ij  ,  II    ij ij  The derivative of B and H represent pressure
2  2 
effect and tanθ is the gravitational effect. The right
1  ui u j  hand side is the resistance from yield stress. As is
and  ij    
2  x j xi  shown in Equation (17), debris flow can move only
and (i, j) = {x, y, z}; εij is strain-rate tensor; τ0 is if pressure and gravitational effects exceed the
yield stress; μd and μc are dynamic viscosity and yield stress effects.
turbulent-dispersive coefficient respectively.
Equation (12) represents the constitutive relation 2.2 Input data
in the region where the shear stress is greater than
yield stress; Equation (13) is for the region with The main inputs are topography and initial
shear stress less than yield stress. debris source distribution. Unlike FLO-2D, Debris-
From the analysis of field data, Liu and Huang 2D uses rectangular grids. There is no need to
(2006) find that the shear layer thickness is less input Manning’s n value, but an accurate value for
than 10% of total flow depth. So the shear layer can yield stress must be measured from samples. But
be ignored to the leading order. After simplification, for a rough estimation, yield stress value can be
the resulting governing equations in conservative estimated with grain size and composition. Man-
form are conservation of mass made structures can be included and modified into
H   uH    vH  the DEM for simulation.
  0, (14)
t x y The debris source distribution must be found
and conservation of momentum in x- and y- through field survey, aerial or satellite photos.
directions From the field survey or satellite photo analysis,
  uH 


 u2 H     uvH   gH sin one can find the dry debris volume Vd and its
corresponding triggering locations. From
t x y
(15) Takahashi (1991) the equilibrium solid volume
B H 1  0u
 gH cos  gH cos  , concentration Cv (%) of a flowing debris flow can be
x x  u 2  v2
expressed as
  vH 

  uvH 


 v2 H  Cv 
 tan
t x y
(16)
    tan   tan  , (18)
B H 1 0v where ρ is water density; σ is the density of dry
 gH cos  gH cos  ,
y y  u 2  v2 debris (around 2.65 g/cm3); φ is internal friction
where H(x, y, t) is flow depth; B(x, y) is the fixed angle (about 37◦) ; θ is average bottom slope angle
bed topography; u(x, y, t) and v(x, y, t) are depth- in the field. The maximum value of Cv cannot
averaged velocities in x- and y-direction exceed 0.603. With Cv obtained from Equation (18),
respectively; tanθ is the bottom bed slope; τ0 and ρ the volume of the flow Vd can be calculated as
are debris-flow yield stress and density, which are Vd/Cv. These source volumes and their locations
all assumed to be constant; g is the gravitational can be inputted through graphical user interface.
acceleration. The effects of bottom erosion and The only rheological parameter needed for
deposition are not considered in this version, but input is the yield stress τ0. Its value varies

296

according to the solid size and composition in the
field, and usually ranges from 102 to 104 Pa for
debris flows with large granular materials. The
larger the averaged grain diameter, the higher the
yield stress. For mud flows, the yield stress is
usually less than 100 Pa. The value of yield stress
can be calibrated using the method proposed by
Liu and Huang (2006).
2.3 Computational algorithm
Finite difference method is applied to
discretize the governing equations, i.e., Eqs. (14),
(15) and (16). In spatial discretization, the 1st-order
Upwind method is applied to discretize convective
term and 2nd-order central difference method is
used for the remaining terms. The explicit 3rd-order
Adams-Bashforth method is used for time
advancing.
To start the computation, Debris-2D will
determine where debris flow can be initialized by
Equation (17). If Equation (17) is not satisfied, the
mass stays stationary and velocities are zero with
the flow depth unchanged. Time step Δt is fixed in
Debris-2D. During computation, if the maximum
velocity in the whole computational domain is less
than numerical error, the computation terminates
automatically.
2.4 Output data
The output time interval of Debris-2D can be
in seconds. Similar to FLO-2D, Debis-2D can
output temporal variation of flow depth, depth-
averaged velocities for the whole domain. Debris-
2D can also calculate impact force (Liu and Lee
1997; Liu et al. 1997) at specified locations, and
output the final affected area and the maximum
flow depth and velocities variation in time.
3 Comparisons between FLO-2D and
Debris-2D
An overall comparison between FLO-2D and
Debris-2D is listed in Table 1. FLO-2D treats debris
flows as a flow sediment routing problem and not a
real debris flow with big granular material inside.
As a result, although the constitutive relations of
the two comes from the same literature, Debris-2D
J. Mt. Sci. (2013) 10(2): 293–304
treats constitutive relation as a nonlinear
discontinuous relation while FLO-2D treats it only
as part of the bottom stress and a continuous
concept. This can be seen from comparing
Equation (4) with Equations (12) and (13).
Physically, Derbis-2D can simulate the starting and
termination of debris flows. Flows in FLO-2D
simulation cannot stop unless user terminates the
computation manually. The simulation termination
method can have dramatic effect on the final
outcome. Since FLO-2D is terminated through user
defined maximum calculation time, the final
spread of debris flow can be different by an order of
magnitude for different termination time. But
Debris-2D terminates simulation by physical
criteria, no termination time is necessary and the
final spreading of debris flow will not be affected by
simulation time.
The most significant difference of input data is
the debris flow volume. FLO-2D uses the user
defined debris flow discharge hydrographs and
treats debris flow (or mud flow) as flood routing.
So the total volume relates to the artificially
determined hydrograph may not be the same as the
real sediment existed in the field. Debris-2D
simulates the flowing motion with real source
distributions determined through field survey or
satellite photo analysis. Therefore, a landslide
induced debris-flow event should be better
simulated by Debris-2D.
Numerically, Debris-2D treats shocks better
than FLO-2D due to the use of upwind method and
conservative form of equations. But in computation
efficiency, FLO-2D uses variable time steps but
Debris-2D uses fixed time step. Therefore, FLO-2D
is more efficient in computation time of Central
Processing Unit (CPU).
The outputs from both programs are similar,
with minor difference on output time interval. But
FLO-2D has a friendly interface and well developed
link to graphical user interfaces (GUI) and GIS
programs. Although Debris-2D has the GUI
program for user, the interface still needs
improvement.
4 Case Study
4.1 Introduction of the field case
In order to validate and compare the two
297
J. Mt. Sci. (2013) 10(2): 293–304

Table 1 Comparison between FLO-2D and Debris-2D

Attribute
FLO-2D Debris-2D
Category Item
Mass and momentum equations in Mass and momentum
Governing equation
8 directions equations in conservative form
Fundamental
Constitutive relation Julien and Lan (1991)
Theory
Initiation criterion No Derived
Shock treatment Direction average Damping scheme
Uniform and square grids in x, y
DEM Grids in x, y directions
directions
Manning’s n value Required Not required
Rainfall hydrograph From rainfall data Not required
Mass distribution from field
Debris flow mass Discharge hydrograph (calibrated)
survey
Debris flow
Predefined as initial input Not required
Input Data Inflow location
Initial mass Determined from field survey
Not required
distribution or aerial photo
Landslide location Through hydrograph and debris Mass distribution input
effect flow inflow location according to landslide location
Volume concentration Cv, yield
Input parameters stress τ0, viscosity η, Manning’s n Yield stress τ0
value, Resistance parameter K, etc.
Numerical scheme in 1st-order Upwind method for
Central difference method
space convective term
Numerical scheme in Forward difference 3rd-order Adam-Bashforth
(Time step adjusted through CFL method in time differencing
Computational Time ( t )
condition) (Time step fixed)
Algorithm
Flow stops automatically
User defined maximum simulation
Stopping mechanism during calculation (yield stress
time
effect)
Volume conservation Conserved
Temporal variation of Flow depth; depth-averaged velocities in whole
Output data
Output domain; maximum flow depth and impact force of time variation
Time resolution Hour Second
Program interface and functionality Complete and friendly Complete but not yet friendly
Note: DEM=digital elevation model; CFL condition=Courant-Friedrichs-Lewy condition

programs, a real case occurred during typhoon
Morakot in Xinfa village, southern Taiwan (Figure
1) was studied. Typhoon Morakot hit Taiwan on
August 5-10, 2009 with world record heavy rainfall.
There were severe floods, countless landslides, and
debris flow disasters associated with Typhoon
Morakot. The catastrophic landslide in Xiaolin
village (Tsou et al. 2011) is the one among many
disasters. There were 673 deaths, 26 people
missing, and about 70 million US dollars of
agricultural and property loss induced by Typhoon
Morakot and is the worst typhoon disaster in 20
years in Taiwan (Soil and Water Conservation
Bureau 2010).
Xinfa village landslide and debris flow is one
of the disasters. The rainfall hydrograph is shown
in Figure 2 and was recorded at the Xinfa
meteorological station. This station is 2.2 km away
from the disaster location. The accumulative
rainfall in 5 days exceeds the average local annual
accumulative rainfall 2,398 mm. At 6:00 AM on
August 9, landslides in eastern part of Xinfa village
occurred and induced debris flows. There were 5
people dead, 12 people injured, 6 houses buried,
and about 15 hectare agriculture land destroyed.
The orthorectified aerial photo, which was taken
six months after the disaster, of landslide and the
final deposition area are shown in Figure 1. The
landslide areas are, from right to left, 21,183; 3,562
and 285 m2 respectively, as is circled by red lines in
Figure 1. According to the field survey (Soil and
Water Conservation Bureau 2010), the total
volume of debris deposition is about 130,000 m3.
From field examination, debris flow traces can be
clearly identified in local buildings, one of the trace
is shown in Figure 3. The height of the trace

298
 J. Mt. Sci. (2013) 10(2): 293–304

Figure 1 Aerial photo of Xinfa Village (shooting time: March 2010). The region circled by red line is landslide area,
and orange for deposition zone. The solid blue circle is the inflow location for FLO-2D simulation. The marks A and
B are the locations of buried houses (see the zoom-in photo in Figure 3 for more details).

Figure 3 Photo of the buried houses. The photo was

Figure 2 The hourly and accumulative rainfall
hydrograph recorded by Xinfa meteorological station
(No. C1V240, Central Weather Bureau of Taiwan). The
triggering time of the landslide was at 6 AM on August
9, 2009.
taken toward northwest. The marks A and B
correspond to the locations marked A and B in Figure
1 respectively. At the location A, the flow depth of
debris flow is about 6 m at least according to the trace
on the wall. The deposition depth of the debris flow
front is about 2-3 m, as shown at the location B.

indicated that debris flow passed that particular
building at a height about 6 m and the location of
this house is marked A in Figure 1.
The major lithological formation in Xinfa
village is the Chaochoiu formation with weakly
deformed rocks of argillite and slate which are
clipped by quartzite sheets. In the field, the
weathered rocks form onion-skin-like weathering
shape or pencil structure (Lin et al. 2011a), and
fractures usually occur along the foliation.

299
J. Mt. Sci. (2013) 10(2): 293–304

According to field investigation (Soil and upstream watershed is 3,500 m2. With rainfall
Water Conservation Bureau 2010), the debris size hydrograph as in Figure 2, the corresponding water
of deposition is about 20 cm in average. The discharge hydrograph is obtained using HEC-HMS
maximum diameter of debris is larger than one (2010). Using the resulting discharge hydrograph,
meter. Besides, by sieve analysis, more than 60% of as shown in Figure 4, we propose two scenarios for
deposit material is gravel with the diameter larger
than 2 mm (Chang 2011). With the deposit material
brought from the field, we used the method
proposed by Liu and Huang (2006) to calibrate the
parameter of yield stress in the laboratory. The
obtained value of yield stress is 1,250 Pa and is
used in all simulations.

4.2 Input data

Table 2 shows the complete listing of inputs Figure 4 The water discharge during the typhoon
for both programs. There are common inputs. The Morakot at the inflow location for FLO-2D simulation
topography input is DEM in resolution of 5 m × 5
m. For both simulations the value of yield stress is
1,250 Pa, and the solid volume concentration Cv is
0.603 from Equation (18). With Cv and the total
debris 130,000 m3, we obtain the total volume of
debris flow is about 215,600 m3. In what follows we
describe the input for FLO-2D and Debris-2D
separately.
FLO-2D considers infiltration. But debris flow
occurred on the third day of the typhoon event, the
accumulation rainfall is already 2,754 mm,
therefore we assume that all the debris and soil are
saturated at that time. Thus the infiltration effect is Figure 5 The two scenarios, i.e., Scenario I and II, of
negligible. The discharge hydrograph and inflow water discharge hydrographs for the input of FLO-2D
simulation. Scenario I and II set the event duration to
locations are needed for FLO-2D. The inflow be one hour and 58 hours respectively. The total debris
location at the downstream of landslide is marked flow volumes of Scenario I and II are 19,144 and
as solid blue circle in Figure 1. The total area of 215,600 m3 respectively.

Table 2 Input data of FLO-2D and Debris-2D for Xinfa village case study
Attribute
FLO-2D Debris-2D
Category Item
DEM resolution 5m × 5m
n=0.03 for riverbed
Topography
Manning’s n value† n=0.1 for farm land Not required
n=0.04 for road & artificial surfaces
Cv 0.603
α2=0.0723 and β2=20 for Equation
Yield stress (Pa) τ0=1,250
(7)
Parameters
Viscosity α1=0.05 and β1=20 for Equation (8) Not required
input
Resistance parameter K 2,000 Not required
Specific weight (kg/m3) 2,650
Simulation time 40 hours Not required
Debris flow Initial mass distribution Not required Figure 6a
input Water discharge hydrograph see Scenario I and II in Figure 5 Not required
Note: DEM=digital elevation model; Cv=Volume concentration (%); †reference: FLO-2D Software, Inc. 2006

300

debris flow simulations with FLO-2D, as is shown
in Figure 5. For Scenario I, we adopt a typical
application in Taiwan. Lin et al. (2008) assumed
the event starts at the peak of runoff and the
duration of debris-flow event is one hour, i.e.,
from 7:40 PM to 8:40 PM on August 8. Afterward,
there is no debris flow inflow but simulation keeps
going. With the given inflow hydrograph, the total
debris-flow volume input in Scenario I is only
19,144 m3. For Scenario II, the debris-flow event
is assumed to last 38 hours, i.e., from 7:40 PM on
August 8 to 9:50 AM on August 10. With the given
debris flow input hydrograph, the total debris-
flow volume 215,600 m3 in Scenario II. The
simulation time for input is equal to the field
Figure 6 The simulation results using Debris-2D at different time. The black line represents the front of the
in-situ deposition zone.
J. Mt. Sci. (2013) 10(2): 293–304
measurement, i.e., FLO-2D is 40 hours for both
Scenarios.
As for mass input in Debris-2D, we distribute
the debris-flow mass, of total volume 215,600 m3,
to the three landslide areas by constant but
different depth. The initial mass distribution is
shown in Figure 6a.
4.3 Simulation results and discussion
Table 3 summarizes the comparison of the
simulation results of FLO-2D and Debris-2D. The
simulation results of FLO-2D for Scenario I and II
are displayed at 1, 10, 20 and 30 hour as shown in
301
J. Mt. Sci. (2013) 10(2): 293–304

Table 3 Comparison of the simulation results

FLO-2D
Attribute Debris-2D
Scenario I Scenario II
Maximum Flow depth 1.03 m 1.34 m 6.53 m (see Figure 9).
at location A in Figure 1 (see Figure 8) (see Figure 8) Matched the in-situ condition (see Figure 3)
Termination time of 40 hours (Simulation time set by user). 163 seconds.
simulation Disagreed with the witness’ description Matched the witness’ description
Simulated distribution Disagreed. Matched. (see Figure 6)
compared to in-situ The distributions get wider and wider Maximum error of the front position is 20
affected zone over time (see Figure 7). m. (<12% w.r.t. the final transverse spread).

Figure 7 The simulation results of Scenario I and II using FLO-2D at different time. The black line represents the
front of the in-situ deposition zone.

302

Figure 7. The results of Scenario I show that the
debris flow is essentially a thin layer after 1 hr.
Thus, neither the affected zone nor the total
debris-flow volume matches the field measurement.
In Scenario II, the simulated distribution does not
match the real affected zone in the field. Also, from
the results of two scenarios, we find that the
simulated distribution grows wider and wider over
time. This result proves that FLO-2D will not
terminate the simulation automatically until the
simulation time reaches. Figure 8 illustrates the
simulated depths of the two scenarios at the
location A in Figure 1. The maximum flow depths
of Scenario I and II are 1.03 m and 1.34 m
respectively. According to the in-situ condition (see
Figure 3), the maximum flow depth at the location
reached two floors high, i.e., higher than 6 m. Both
simulated depths are much less than the field
measurement. So FLO-2D failed to simulate the
motion of debris flow properly.
The results of Debris-2D are shown in Figure 6,
and the whole debris flow event is finished in 163
seconds. The depth distributions at time 0(initial),
Figure 8 The simulating flow depths of the Scenario I
and II using FLO-2D at the buried house, as is marked
A in Figure 1.
Figure 9 The simulating flow depth of time variation
using Debris-2D at the buried house, as is marked A in
Figure 1. At the 16.3 second, the maximum depth is 6.53
m, and it matches the in-situ condition (see Figure 3).
J. Mt. Sci. (2013) 10(2): 293–304
10, 20, 30, 60 and 163 second are shown in Figure
6, respectively. The final deposition area matches
well with the affected area (black boundary). From
the aerial photo, the maximum error is 20 m and
the error is less than 12% with respect to the final
transverse spread. The temporal variation of flow
depth at the buried house (marked A in Figure 1) is
shown in Figure 9. The maximum depth is about
6.53 m at 16.3 second, and this depth agrees well
with the in-situ measurement in Figure 3. The
simulated final deposition depth at the front (line B
in Figure 3) is about 2-3 m, and agrees with field
measurement well. The duration of the whole event
is close to 3 minutes and also fits the local
residents’ description. The simulation results show
that Debris-2D gives better assessments for this
debris flow event.
5 Concluding Remarks
This paper compares FLO-2D and Debris-2D
for the applicability of debris flows with a lot of
granular material. As for the fundamental theories
of the two models, the major divergence is the
methods in treating rheological relations. As a
result, very different simulation termination
mechanisms are used by two programs. FLO-2D is
a flow routing program, so a termination condition
is usually not required. However, in the case of
debris flow modeling, one has to manually
terminate the simulation. In contrast, Debris-2D
does display the characteristic that debris flows do
stop.
From the results of a real case study, the mass
distribution simulated by FLO-2D does not agree
with the field or the depth distribution. But the
final distribution and maximum depth simulated
from Debris-2D agree with the field condition very
well. Since the case in this paper has a lot of
granular material, so the comparison can only
stand for debris flows with a lot of granular
material. Therefore, Debris-2D is recommended
for the application on assessment of landslide-
triggered debris-flow hazards.
Acknowledgements
The authors would like to thank the financial
support from National Science Council of Chinese
303
J. Mt. Sci. (2013) 10(2): 293–304
Taipei (Grant No. NSC 96-2625-Z-002-006-MY3).
Lots of useful and valuable data provided by Dr.
References
Boniello MA, Calligaris C, Lapasin R, Zini L (2010) Rheological
investigation and simulation of a debris-flow event in the
Fella watershed. Natural Hazards and Earth System Sciences
10:989-997.
Bertolo P, Wieczorek GF (2005) Calibration of numerical
programs for small debris flows in Yosemite Valley, California,
USA. Natural Hazards and Earth System Sciences 5:993-1001
Chang JJ (2011) On Deposition Length of Debris-flow at Hsia-
Hsinkai tribe, Kaohsiung, Taiwan, in 2009 Morakot Typhoon.
Master Thesis, Department of Civil engineering, National
Kaoshiung University of Applied Sciences, Chinese Taipei.
Chen CY, Chen TC, Yu FC, Hung FY (2004) A landslide dam
breach induced debris flow-a case study on downstream
hazard areas delineation. Environmental Geology 47:91-101.
Chen SC, Wu CY, Wu TY (2009) Resilient capacity assessment
for geological failure areas: examples from communities
affected by debris flow disaster. Environmental Geology
56:1523-1532.
Chen SC, Wu CY, Huang BT (2007) Risk assessment of debris
flow disaster in Songhe village. Journal of Chinese Soil and
Water Conservation 38(3):287-298. (In Chinese)
Četina M, Rajar R, Hojnik T, et al. (2006) Case study: numerical
simulations of debris flow below Stože, Slovenia. Journal of
Hydraulic Engineering ASCE 132(2):121-129.
FLO-2D Software, Inc. (2006) FLO-2D Users manual (Version
2006.01), Arizona, USA
Hsu SM, Chiou LB, Lin GF, et al. (2010) Applications of
simulation technique on debris-flow hazard zone delineation:
a case study in Hualien County, Taiwan. Natural Hazards and
Earth System Sciences 10:535-545.
HEC-HMS (2010) Hydrological Engineering Center-
Hydrological Modeling System, User’s manual (Version 3.4),
US Army Corps of Engineers.
Julien PY, Lan Y (1991) Rheology of hyperconcentrations.
Journal of Hydraulic Engineering ASCE 117(3):346-353.
Lin CW, Lin WH, Kao MC (2011a) Geological map of Taiwan
scale 1:50,000. Central Geological Survey, Ministry of
Economic Affairs, Chinese Taipei. (In Chinese)
Lin DG, Hsu SY, Chao CH, et al. (2008) Applications of
simulation technique on hazard zone delineation and damage
assessment of debris flow. Journal of Chinese Soil and Water
Conservation 39(4):311-319. (In Chinese)
Lin JY, Yang MD, Lin BR, Lin PS (2011b) Risk assessment of
debris flows in Songhe Stream, Taiwan. Environmental
Geology 123:100-112.
Lin ML, Wang KL, Huang JJ (2005) Debris flow run off
304
SHEN Che-Wei is also acknowledged.
simulation and verification - case study of Chen-You-Lan
watershed, Taiwan. Natural Hazards and Earth System
Sciences 5:439-445.
Liu KF, Lee FC (1997) Experimental analysis on impact
mechanism of granular flow. Chinese Journal of Mechanics
13(1):87-100. (In Chinese)
Liu KF, Lee FC, Tsai SP (1997) The flow field and impact force
on a debris dam. Proceedings of the First International
conference on Debris-flow Hazards Mitigation: Mechanics,
Prediction, and Assessment. Taipei, Taiwan. pp 737-746.
Liu KF, Li HC, Hsu YC (2009) Debris flow hazard defense
magnitude assessment with numerical simulation. Natural
hazards 49(1):137-161.
Liu KF, Li HC (2007) The assessment of debris flow emergency
measures. City and Planning 34(1):57-73. (in Chinese)
Liu KF, Wu YH (2010) The Assessment of Debris Flow Hazard
in Korea Using Debris-2D. INTERPRAEVENT 2010-
International Symposium in Pacific Rim, Taipei, Taiwan. pp
820-827.
Liu KF, Huang MC (2006) Numerical simulation of debris flow
with application on hazard area mapping. Computational
Geosciences 10:221-240.
O’Brien JS, Julien PY, Fullerton WT (1993) Two-dimensional
water flood and mudflood simulation. Journal of Hydraulic
Engineering ASCE 119(2):244-260.
Quan Luna Q, Blahut J, van Westen CJ, et al. (2011) The
application of numerical debris flow modelling for the
generation of physical vulnerability curves. Natural Hazards
and Earth System Sciences 11:2047- 2060.
Soil and Water Conservation Bureau (2010) The Investigation of
vulnerability factors and risk analysis of debris flow potential
creeks. Soil and Water Conservation Bureau, Chinese Taipei.
(In Chinese)
Stolz A, Huggel C (2008) Debris flows in the Swiss National
Park: the influence of different flow programs and varying
DEM grid size on programing results. Landslides 5:311-319.
Tsai MP, Hsu YC, Li HC, et al. (2011) Applications of simulation
technique on debris flow hazard zone delineation: a case
study in Daniao tribe, Eastern Taiwan. Natural Hazards and
Earth System Sciences 11:3053-3062.
Takahashi T (1981) Debris flow. Annual Review of Fluid
Mechanics 13:57-77.
Tsou CY, Feng ZY, Chigira M (2011) Catastrophic landslide
induced by Typhoon Morakot, Shiaolin, Taiwan.
Geomorphology 127:166-178.