Journal of the Civil Engineering Forum.
September 2025, 11.
:233-244 DOI 10.
22146/jcef.
Available Online at https://jurnal.
id/v3/jcef/issue/archive Development of a Numerical Model for the Formation of Complete and Incomplete Channel Blockages and Their Influences on River Flow Takashi Wada1* .
Hiroshi Miwa1 .
Naoto Aoki1 .
Yusei Fujii2 1 Department of Social Systems and Civil Engineering.
Tottori University.
Tottori.
JAPAN 2 Aisawa Construction Co.
Ltd.
Okayama.
JAPAN *Corresponding author: wada-t@tottori-u.
SUBMITTED 31 December 2024 REVISED 18 March 2025 ACCEPTED 20 March 2025 ABSTRACT Large landslides, triggered by torrential rain or earthquakes, can slide down mountainous slopes and block river channels at the lower end of the slopes.
In cases where the landslide volume is relatively small compared to the river discharge, or when the distance between the landslide slope and the river channel is long, incomplete channel blockages may occur due to an insufficient supply of landslide material to fully block the river flow.
Since the shape of the channel blockage is the final result obtained through the temporal changes in landslide material movement, river flow, and topography, considering their interactions, it is necessary to investigate the blockage shape by numerical analysis that accounts for these interactions.
Therefore, we developed a numerical model to predict the formation of various channel blockages by incorporating the combined conditions of topography, landslide volume, and river discharge.
The developed model is a two-dimensional .
-D) model, which can connect several one-dimensional calculation areas for mountainous streams at any selected point in the 2-D area.
In addition, the model can consider landslide material movements represented by cylindrical blocks.
To verify our model and identify appropriate values for the associated parameters, we investigated the MAE .
ean absolute erro.
for the deposit thickness distribution and the PWO .
ercentage of the area where the actual and calculated waterlogged areas overlappe.
between the actual and calculated results using our model for two previous channel blockages of different sizes.
Although our model and the associated parameters still need to be improved by considering the loss of landslide material, they are useful for estimating the magnitude and area of damage caused by large-scale landslides and the associated channel blockage and waterlogging in various river channels with steep side slopes.
The calculated results can be utilized in investigating disaster countermeasures for landslides in the area.
KEYWORDS Numerical modeling.
Two-dimensional calculation.
Channel blockage.
River flow.
Landslide material movement.
A The Author.
This article is distributed under a Creative Commons Attribution-ShareAlike 4.
0 International license.
1 INTRODUCTION
In East and Southeast Asia, including Japan, large-scale landslides triggered by torrential rains or earthquakes can occur on mountainous slopes, causing landslide materials to block river channels at the lower end of the slopes (Hung, 2000.
Sassa, 2005.
Li et al.
, 2011.
Fan et al.
, 2012.
Ishizuka et al.
, 2017.
Van Tien et al.
, 2.
Blocked river flow on the upstream side of the blockages can lead to flooding of buildings in waterlogged areas, while on the downstream side, debris flows caused by erosion due to overflow from the blockages can result in severe flood damage.
Thus, channel blockages .
andslide dam.
may lead to multimodal sediment disasters on both the upstream and downstream sides.
rainfall disaster on the Kii Peninsula in Japan.
Heavy rainfall from Typhoon Talas triggered multiple largescale landslides, and more than half of the channel blockages failed to fully block the river.
Consequently, these blockages collapsed within a day of formation (Inoue and Doshida, 2.
There were many cases where a complete blockage did not occur due to insufficient landslide material reaching the river channel.
According to the database by Peng and Zhang, which 239 landslide dams worldwide, 87% of the 204 recorded landslide dams with failure data collapsed within a year, 71% within a month, 51% within a week, and 34% within a day (Peng and Zhang, 2.
A complete blockage does not necessarily occur due to the inflow of landslide material into a river channel.
For example, the river channel blockage in Hsiaolin Village.
Taiwan, caused by heavy rainfall from Typhoon Morakot in 2008, was formed by the inflow of largescale landslide material.
However, the blockage quickly failed due to the erosion by overtopping floodwater resulting from the large river flood discharge (Li et al.
A similar case occurred during the 2011 heavy The formation of river blockages, whether complete or incomplete, depends on the relationship between the volume of landslide material flowing into the river and the riverAos channel size .
r flow discharg.
Additionally, the volume of landslide material is affected by the topographic conditions between the landslide slope and the river channel, through sedimentation and other processes.
In the 2011 Kii Peninsula disaster, the incomplete blockages may have occurred be- Journal of the Civil Engineering Forum cause the channel width and river discharge were large enough that the supplied landslide volume was insufficient to fully block the river flow.
Chen et al.
also suggested that in-channel deposition of landslide material inflowing into the channel, which leads to landslide dam formation, is affected by the channel width and flow depth.
Therefore, it is essential to develop a method to predict the types of river channel blockages, such as complete or incomplete, under various conditions, considering the topography and river characteristics, to implement effective countermeasures for each type of blockage.
Previous studies have primarily focused on damage estimation downstream of landslide dams caused by floods resulting from dam failures.
Several statistical analyses have been conducted to predict the magnitude of dam-break floods .
, peak flow rate, hydrograph.
based on dam shape parameters .
uch as height, volume, and impounded water volum.
These studies proposed various relational equations to describe the relationship between dam shape and dam-break floods (Costa, 1985.
Walder and OAoConnor, 1997.
Chen et al.
Fan et al.
, 2.
Additionally, numerous studies have investigated the failure processes of landslide dams using numerical simulations .
Takahashi and Kuang .
Takahashi and Nakagawa .
Awal et al.
Akazawa et al.
Takayama et al.
Zhong et al.
), laboratory experiments .
Takahashi and Kuang .
Takahashi and Nakagawa .
Awal et al.
Zhou et al.
), and field experiments .
Akazawa et al.
Zhong et al.
Takayama et al.
) to identify the characteristics of floods caused by dam failures.
These studies have elucidated the process of landslide dam destruction by dividing it into several patterns.
The above research indicates that the flood hydrograph caused by landslide dam failure is strongly related to the dam shape.
Therefore, understanding the dam shape formation process is important.
However, few studies have focused on the formation process of channel blockages .
andslide dam formatio.
resulting from the inflow of landslide material into a river Swanson et al.
and Costa and Schuster .
proposed a geomorphological classification method based on the channel topography, which identifies where channel blockages may form.
However, this method is a statistical classification based on target topographic conditions and does not clarify how topography contributes to the formation process or final shape of the blockage.
Liao et al.
conducted flume experiments to examine the kinematic processes and emplacement of rockslides, and the factors that dictate landslide dam formation.
Their experimental results suggest that the height and shape of landslide dams, such as incomplete or complete blockages, are influenced by rockslide volume, rock fragmentation, and river flow depth.
Crosta et al.
proposed contin234 Vol.
11 No.
3 (September 2.
uum finite element models consisting of the conservation equations of mass, momentum, and energy for rock and debris avalanches caused by landslides.
However, this is a model for flow-like landslides in vertical two-dimensions, and it is difficult to account for the influence of planar topography.
In addition, their model did not consider the dam formation processes in conjunction with river flow.
Since the shape of a channel blockage is the final result obtained through temporal changes in landslide material movement, river flow, and topography, considering their interactions, it is necessary to investigate the blockage shape by numerical analysis that includes these interactions.
In this study, we aimed to establish a prediction method for the formation process of river channel blockages due to the inflow of landslide materials into a river channel, under various combinations of topography, landslide volume, and river discharge.
Our objective was achieved by developing a planar two-dimensional .
-D) numerical flow model with cylindrical blocks representing landslide material.
We validated the developed model and its associated parameters by comparing the simulated results with actual blockage formation and expanded flood areas.
2 DEVELOPED NUMERICAL MODEL
Figure 1 provides an overview of the developed model and its base model.
The developed model is based on the two-dimensional model AuNumerical Model Considering Multiple Inflows of Debris Flows and River FloodsAy by Wada et al.
, which predicts inundation volume and area caused by the simultaneous occurrence of debris flows and river floods.
In the model, the downstream ends of several 1-D calculation areas for mountain streams are connected to a 2-D calculation area for the floodplain at any selected point.
We incorporated into the model, the landslide material movements represented by cylindrical blocks, as developed by Satofuka and Takahashi .
Thus, the model can simultaneously predict multiple flows, flood propagation processes, and landslide material movement.
This indicates that the model can predict how landslide material inflowing into flood flows changes the flow direction, resulting in river blockage formation on a two-dimensional plane.
1 1-D and 2-D flow calculation parts Figure 2 shows the outlines of the 1-D and 2-D calculation parts of the developed model.
Several 1-D calculation areas are integrated into a unified temporary 1-D calculation area, separated by designated calculation points at the upstream ends .
st_n , n: number of 1-D calculation area.
and downstream ends .
e_n ), for each 1-D area.
These 1-D areas operate indepen- Vol.
11 No.
3 (September 2.
Figure 1 Outlines of the developed model .
and its base model .
Journal of the Civil Engineering Forum Figure 3 Outline of landslide material models represented by cylindrical blocks.
(Quoted from Satofuka and Takahashi .
2 Landslide cylindrical block model Figure 3 illustrates the landslide material model represented by cylindrical blocks.
The landslide material is considered an assembly of vertically standing circular cylinders arranged in a hexagonal close-packed structure.
To account for porosity between the cylinders, when a hexagonal column is converted into a cylindrical block, the cylinder diameter is increased so that the cross-sectional area matches that of the original hexagonal column.
Figure 2 Outline of base model for our developed model.
(Partially revised the figure by Wada et al.
The model applies several 1-D and 2-D models at each time step.
In addition, the discharge (Qn ) and the sediment discharge (Qbn ) from the debris flow at the downstream ends of the 1-D areas are added to the inflow point in the 2-D calculation area.
The average bed elevation .
ave_n ) and flow depth .
ave_n ) at the inflow points are assigned to the bed elevation and flow depth at the downstream ends of the 1-D areas.
The transfer of momentum from the 1-D areas to the 2-D area is also considered in the model.
The model enables continuous calculation of debris flows and river floods on multiple streams, including their inflows into stream and river confluences, and the resulting deposition and flood propagation.
For calculation point placement, staggered grids are used in the Specifically, vector quantity calculation points are placed 1/2OIx or 1/2OIy downstream .
ositive directions of the xOeaxis or yOeaxis direction.
from the scalar calculation points, where OIx and OIy represent the grid spacing along xOe and yOeaxes in the 2-D area.
For further details on the connection between the 1-D and 2-D areas in the model, please refer to Wada et al.
The block shape is cylindrical to allow consistent calculation of distances between block surfaces in all directions.
However, a disadvantage is that converting the landslide material into cylindrical blocks creates gaps between blocks that do not exist in reality.
This disadvantage can be mitigated by optimally setting the parameters related to the soil properties of the block.
A cylindrical block consists of an upper unsaturated layer and a lower saturated layer.
only the saturated layer can be eroded by bottom friction.
At the bottom of the cylinder, the eroded saturated layer is transformed into a debris flow with a sediment concentration of 0.
while the remaining landslide material is carried on the surface of the block.
The water content required for debris flow formation is supplied by the remaining cylindrical blocks.
When the movement velocity of a cylindrical block is sufficiently low .
, below the threshold velocity for stopping block movement.
Ublim ), the landslide material is considered deposited, and the volume of the remaining block is transferred into a rising riverbed volume at the stopping point.
The cylindrical blocks move due to surface topographic irregularities and interactions with other blocks.
Each block is affected by gravity and frictional forces acting on its bottom along the slope direction, based on surrounding topographic irregularities.
Additionally, the block is influenced by soil cohesion and repulsive forces that prevent overlap with neighboring blocks.
These forces depend on the distance between blocks and their relative movement direction.
When a flow layer is present beneath a block, shear forces based Journal of the Civil Engineering Forum Vol.
11 No.
3 (September 2.
on velocity differences between the block and the flow layer also affect the block.
These forces are introduced for modeling convenience, and their physical validity should be verified in future studies.
3 Governing equations 1 Equations for a cylindrical block and its motion A cylindrical block is composed of a lower saturated layer .
aturation ratio: Sb ) and an upper unsaturated layer, with thicknesses of hs and hsb , respectively, as illustrated on the left side of Figure 2.
The mass of a block (Mb ) is given by:
Mb = AT S0 .
hsb hS ) .
where S0 is the bottom area of a block.
AT is the apparent density of the saturated layer, and 1 is the difference in density between the saturated and unsaturated layers, which is expressed as:
AT b ECOT A.
Oe COT )Sb AT ECOT A.
Oe COT ) .
where AT b is the apparent density of the unsaturated layer, and COT is the volume concentration of sediment in the static mass.
The xOe and yOedirection equations for the migration velocities of a block .
b , vb ) are as follows:
OCUb Esx S0 = g sinhx x fsx ) OCt OCvb Esy S0 = g sinhy .
y fsy ) OCt where t is the time.
g is the gravitational acceleration.
hx and hy are the xOe and yOe direction gradients of the flow layerAos surface .
f no flow layer exists, they represent the gradients of the bed surfac.
Esx and Esy are the xOe and yOe direction boundary shear stresses, respectively, fx and fy are the xOe and yOedirection cohesion or repulsive forces between blocks around it, respectively, fsx and fsy are the xOe and yOe direction shear stress fractions between blocks around it, and is the summation of these forces.
The soil cohesion force .
b ) determines the values of fx , fy , fsx , and fsy .
The overlapping area ratio for the bottom of the k-th block and the calculation mesh at point (I, .
, denoted as (Pk(I,.
) is used to determine the forces acting on each block.
For further details on the calculation of these forces, please refer to Satofuka and Takahashi .
and Takahashi .
2 Governing equations for debris and flood flows including the flow layer below a block bottom The governing equations for debris and flood flows, erosion/deposition, and riverbed shear stress were partially modified from the equations used in the previous debris flow simulation by Nakagawa et al.
The following are the governing equations for the 2-D area.
The equations for the 1-D area are derived by excluding the y-direction terms.
In the 1-D area, longitudinal changes in stream width can be considered using averaged hydraulic quantities in the cross-stream direction.
Note, however, that energy loss in debris flows due to rapid width changes cannot be strictly accounted for in a 1-D area.
The continuity equation for the total volume of debris flow is:
OCh OCM
OCN
= i ib OCt OCx OCy where h is the flow depth.
M and N are the momentum fluxes in the xOe and yOedirections (M = uh .
N = vh ), respectively, u and v are the flow velocities in the xOe and yOedirections, respectively.
I is the erosion/deposition rate of the riverbed, and ib is the erosion rate of a cylindrical block supplied from its bottom.
Ib is based on the physical properties of landslide material, which are insufficiently understood.
For convenience, the following formula was used to determine the ib value of each cylindrical block:
Oe ub )2 .
Oe vb )2 OIxOIy .
where is a coefficient related to the erosion rate at the bottom of the cylindrical block.
Although was 012 in the previous simulation for flow and deposition caused by a past large-scale landslide (Satofuka, 2.
, the value was identified through trial-and-error in this study.
The continuity equation for the volume of sediment in a debris flow is:
OCCh OCCM
OCCN
= iCO ib COT OCt OCx OCy .
where C is the sediment concentration in the debris flow, and CO is the sediment concentration in the initial mobile layer of the riverbed.
The momentum equation in the xOedirection is:
OCM
OCuM
OCvM
OCH
Ebx Esx = Oegh Oe 0 Oe 0 OCt OCx OCy OCx where H is the flow surface level (H = z .
, z is the riverbed level.
Ebx is the riverbed shear stress in the xOedirection, and AAo is the apparent mass density of the Vol.
11 No.
3 (September 2.
Journal of the Civil Engineering Forum debris flow (= E y C Am y .
AeC).
E is the mass density of the sediment, and Am is the mass density of the interstitial flui.
1 Target river blockages by inflowing landslides The momentum equation in the yOedirection is:
OCN
OCuN
OCvN
OCH
Eby Esy = Oegh Oe 0 Oe 0 OCt OCx OCy OCy .
where Eby is the riverbed shear stress of the yOedirection.
The continuity equation for the riverbed level is:
OCz
i=0 OCt
3 REPRODUCTION CALCULATION FOR VALIDATION OF
DEVELOPED MODEL
The bed shear stresses (Ebx .
Eby ) are calculated using three different flow resistance theories based on sediment-transport modes: debris flow, sediment sheet flow, and ordinary turbulent water flow including bed material load.
These flow modes are classified according to the sediment concentration of the debris flow The erosion/deposition rate of the riverbed .
depends on the large/small relationship between C and the equilibrium sediment concentration (CO ).
If C is smaller than CO , the erosion rate is positive value .
b > .
, meaning the flow erodes the riverbed.
Conversely, if C is larger than CO , the deposition occurs .
b < .
is calculated by using the equations proposed by Takahashi .
For further details on the above equations, please refer to Takahashi .
The targets for reproduction calculations using the developed model were two landslides of different scales:
the landslide in the Obara district.
Hyogo Prefecture, in 2018, with a volume of 30,000 m3 .
uoted in the investigation report of The Japan Landslide Society, 2.
, and the landslide in the Ichinose district.
Tottori Prefecture, in 2004, with a volume of 330,000 m3 .
uoted in The Japan Landslide Society, 2.
Figures 4 and 5 show the outlines of the target landslides, respectively.
The topographic and geologic conditions in the two districts are different, as shown in Table 1.
In particular, the distances and inclinations between the landslide slopes and river channels are remarkably different.
In addition, the dimensionless net resistance coefficients proposed by Iverson .
L/H, in both districts are also different, with L/H > 2 in the Obara district and L/H < 2 in the Ichinose district.
Based on his experimental results with well-sorted gravel, the landslide material in Obara could have been unsaturated, whereas the material in Ichinose may have been Despite these differences, both landslides The finite-difference method was used to discretize these equations in the model.
Forward, upwind, and centered finite-difference methods were used for time discretization, the discretization of advective terms, and spatial discretization, respectively.
Table 1.
Topographic and geologic conditions in both landslide Contents Obara
30,000 m3
Ichinose 330,000 m3
Landslide volume Distance between the landslides O45 m
O150 m
slope and river channel
Landslide-slope O24A
O30A
Inclination between the landslides O8A
O22A
slope and river channel Dimensionless net resistance coefficient, 1/R Estimated river discharges 74 m s * 245 m3 s-1 ** Geologic properties Weathered granite Pelitic schist * The coefficient was equal to L/H, where L is the horizontal distance from landslide source to deposit area, and H is the vertical elevation of the source above the area, proposed by Iverson .
** The value was estimated by the rational formula using the actual hourly rainfall intensity at near observation point.
*** Value corresponds to the actual daily average river discharge at a neighboring observation station.
Figure 4 Landslide disaster and short-term river blockage in the Obara district.
Hyogo Prefecture Figure 5 Area for validation of calculation result and identification of optimal values in Ichinose district Journal of the Civil Engineering Forum Vol.
11 No.
3 (September 2.
Figure 6 Landslide disaster and river blockage in the Ichinose district.
Tottori Prefecture.
uoted from The Japan Landslide Society, 2.
Figure 7 Area for validation of calculation results and identification of optimal values in the Obara district blocked the river channels at the lower end of the slope.
3 Verification and Identification for optimal parameter In the former case (Obar.
, as illustrated in Figure 4, the blockage was not confirmed in subsequent field investigations.
We suggest that the landslide material supplied to the river was insufficient to maintain the blockage, leading to its early disappearance.
As shown in Figure 4, the temporary blockage caused a change in flood flow direction, diverting the main flow behind the revetment on the opposite bank, which led to erosion at the rear of the revetment.
In contrast, the magnitude of the latter landslide (Ichinos.
was large enough to completely block river flow at the base of the On the upstream side, the river overflowed due to retained floodwater, resulting in inundation of several riverside buildings.
After the landslide occurred, a spillway tunnel was constructed through the opposite side of the mountain to release the accumulated floodwater caused by the river channel blockage.
2 Calculation condition The calculations involved multiple cases with different combinations of values for three parameters: the coefficient of block bottom erosion, , the cohesion of the landslide cylinder block, cb , and the threshold velocity for stopping block movement.
Ublim .
These parameters remain physically unclear, and their values were varied within ranges close to those proposed by Satofuka .
Table 2 lists the parameters used in the reproduction calculations.
For the Obara landslide, appropriate parameter values were identified to best reproduce the actual deposit thickness distribution.
The method for validating and identifying optimal parameter values is described in the next section.
These identified values were then used as a reference to define parameter ranges for the reproduction calculation of the larger Ichinose landslide.
To verify the calculation results and identify optimal parameter values in the Obara case, we evaluated the Mean Absolute Error (MAE) between actual and calculated deposit thicknesses in each case.
The parameter set that yielded the lowest MAE was considered optimal.
MAE is calculated using the following equation:
M AE =
1 XN
zical Oe dzisur | .
where N is the total number of 2-D calculation points in the validation area shown in Figure 6, i is the number of the calculation point, dzical is the calculated deposit thickness at the i th-point, and dzisur is the corresponding surveyed value.
In the Ichinose case, validation and parameter optimization were conducted by calculating the Percentage of Waterlogged Overlap (PWO), defined as the proportion of the actual waterlogged area correctly predicted by the simulation.
The validation area is shown in Figure 7.
PWO =
nwlap nwact_only nwcal_only .
where nwlap is the number of the calculation points whose calculated flood depths are larger than 0.
within the actual waterlogged area shown in Figure 5, nwact_only is the number of the points whose calculated flood depths are smaller than 0.
05 m within the actual waterlogged area, and nwcal_only is the number of the points whose calculated flood depths are larger than 05 m without the actual waterlogged area.
Vol.
11 No.
3 (September 2.
Journal of the Civil Engineering Forum Table 2.
Parameters incorporated in reproduction calculations Parameters/Variables Total simulation time (Tmax ) Time step (OI.
Minimum flow depth of flux .
fmin ) Minimum flow depth .
min ) ManningAos roughness coefficient .
m ) Gravity acceleration .
Diameter of sediment .
O) Volume density of sediment .
* Parameters of sediment Volume density of interstitial fluid .
* Sediment concentration by volume in the movable bed layer (C*)* Internal friction angle of sediment .
anI)* Coefficient of erosion rate .
* Coefficient of deposition rate .
* Diameter of sediment .
* River flow discharge (Constant suppl.
Total landslide volume (Including internal voi.
Parameters of landslide cylindrical model Parameters of 2-D area Value of landslide Obara Ichinose m-1/3 s m s-2 kg m-3 m3 s-1 Landslide cylinder model size A regular hexagon with 2 m on each side Number of cylinder model Time of Starting landslide movement 1 hour after starting of calculation Cohesion .
b ) ** 05, 0.
10, 0.
20, 0.
05, 0.
10, 0.
Coefficient of bottom erosion .
** 010, 0.
050, 0.
010, 0.
Threshold velocity for stopping movement (Ublim )** 05, 0.
Saturation in lower unsaturated layer Coefficient of dynamic bottom friction.
Generating debris flow at the bottom No debris flow at the bottom Number of 2-D calculation points .
e2 y je2 ) Interval of 2-D calculation points (Dx2 yDy2 ) Unit kN m-2 156 y 183 906 y 464 2y2 * Values referenced from the debris flow simulation by Nakagawa et al.
** These parameters were varied across multiple calculation cases.
4 CALCULATION RESULTS AND DISCUSSION
1 Identification of appropriate coefficient of bottom erosion () The appropriate value was investigated by comparing the actual and calculated landslide deposit thickness distributions in the Obara district.
Figure 8 presents the actual and calculated results for the deposit thickness distribution in cases where only the value differs, while cb and Ulim values had common values of 100 N m-2 and 0.
50 m s-1 , respectively.
The figure shows that when is 0.
05 or 0.
10, the calculated results closely align with the actual landslide thickness If is significantly less than 0.
05, the calculated deposition area becomes larger than the actual area, which is not considered < 0.
05 was appropriate.
The MAEs for the deposit thickness distribution in cases where were 0.
01, 0.
05, and 0.
10 were 1.
075, and 0.
839, respectively.
These suggested that the appropriate value to predict the landslide sediment movement accurately was the value range from 0.
05 to 10, and closer to 0.
This value is larger than the = 0.
012 used in the previous simulation by Satofuka .
, possibly reflecting the fact that the landslide material in this case was more easily decomposed due to heavy weathering.
This may also be influenced by the unsaturated condition of the material, as suggested by IversonAos coefficient L/H, indicating that landslides involving unsaturated materials may have higher erosion coefficients.
2 Identification of appropriate cohesion of cylindrical block .
b ) The appropriate cb value was investigated using the same approach as in Section 4.
Figure 9 shows the Journal of the Civil Engineering Forum actual and calculated results for the deposit thickness distribution in cases where only cb differs, while and Ulim were fixed at 0.
05 and 0.
50 m s-1 , respectively.
The calculated results under cb = 50 or 100 N m-2 agree relatively well with the actual thickness distribution.
However, when cb was increased to 200 N m-2 , the calculated deposition area was smaller than the actual one, resulting in poorer reproducibility.
The MAEs in cases where cb were 50, 100, and 200 N m-2 were 1.
026, 1.
075, and 302, respectively.
This suggested that the appropriate cb value was the value range from 50 to 100 N m-2 .
The appropriate cb value was also investigated in terms of the calculated river-flow distributions in these cases.
Figure 10 depicts the calculated flow-depth distribution two hours after the landslide.
The results under cb = 50 or 100 N m-2 well predicted the partial river blockage by inflowing the landslide material.
Additionally, these results also predicted the changes of the main flow direction, passing behind the opposite-side revetment.
This also suggested that the appropriate cb value was the value range from 50 to 100 N m-2 .
Note that these calculated river blockages remained two hours af- Vol.
11 No.
3 (September 2.
ter the landslide.
Further investigation is needed to determine whether this blockage could be cleared by overtopping flow using extended calculation periods.
The estimated appropriate cb value was slightly smaller than the value of 179 N m-2 set in the previous simulation by Satofuka .
This may suggest that the landslide material was easily decomposed due to the great weathering.
This value is an order of magnitude .
-1 ) less than that of general physical soil properties.
This discrepancy might be attributed to the consolidation of finer landslide material particles into virtual larger blocks in the model.
The difference between actual and modeled cb values should be further explored in future studies.
3 Identification of appropriate threshold velocity for stopping movement (Ublim ) The appropriate Ublim value was also investigated in the same manner of 4.
1 and 4.
Figure 11 presents the actual and calculated results for the landslide deposit thickness distribution in cases where only Ublim varies, with and cb fixed at 0.
05 and 100 N m-2 .
The figure illustrates that the differences between the results for Figure 8 Calculated results for deposit thickness distribution two hours after the landslide in the Obara district, showing cases where only the coefficient of bottom erosion () differs Figure 10 Calculated results for river flow depth distribution 2 h after the landslide in the Obara district, showing cases where only cohesion of landslide cylindrical block .
b ) differs Figure 9 Calculated results for deposit thickness distribution 2 h after the landslide in the Obara district, showing cases where only cohesion of landslide cylindrical block .
b ) differs Figure 11 Calculated results for deposit thickness distribution 2 h after the landslide in the Obara district, showing cases where only threshold velocity for stopping block movement (Ublim ) differs Vol.
11 No.
3 (September 2.
Journal of the Civil Engineering Forum Table 3.
MAEs of actual and calculated results for the deposit thickness distribution in each case Coefficient of bottom erosion () Parameter values Cohesion of Threshold velocity for cylinder block stopping movement .
b ) (N m-2 ) (Ublim ) .
s-1 ) MAE Figure 12 Calculated results for distributions of deposit thickness and river flow depth 1 h after the landslide in Ichinose district for = 0.
01 or 0.
05, in cases where cb and Ublim values had common Table 4.
PWOs of actual and calculated results for the deposit thickness distribution in each case Coefficient of bottom erosion () Parameter values Cohesion of Threshold velocity for cylinder block stopping movement .
b ) (N m-2 ) (Ublim ) .
s-1 ) PWO cases where only Ublim differs are slight.
The MAEs in cases where Ublim were 0.
05 and 0.
50 m s-1 were almost identical at 1.
072 and 1.
075, respectively.
Thus.
Ublim exhibits a minimal effect on the deposit distribution.
This suggested that the decomposition of the front-part material was completed before the cylindrical blocks stopped moving.
4 Identification for optimal parameter values of ObaraAos We investigated all MAEs of the actual and calculated results under all combinations for , cb .
Ublim as shown in Table 3.
The optimal parameter values, that is the case with the smallest MAE is under the combination for = 0.
10, cb = 50 N m-2 .
Ublim = 0.
50 m s-1 .
Figure 13 Temporal variation of block-averaged thicknesses and movement velocities for both landslide calculations under the combination for = 0.
01, cb = 50 N m-2 .
Ublim = 0.
50 m s-1 5 Identification for optimal parameter values of IchinoseAos landslide We investigated all PWOs by comparing the actual and calculated waterlogged areas caused by the Ichinose landslide under all combinations of , cb .
Ublim as shown in Table 4.
The table shows that the appropriate and cb values for accurately predicting the Ichinose landslide movement were approximately 0.
01, and 50 N m-2 , respectively.
In all cases where was greater than or equal to 0.
05, the PWO was less than 0.
PWO
< 0.
45 indicates that only river flow occurred, without waterlogging caused by a river channel blockage.
The appropriate value for the Ichinose landslide was much smaller than that for the Obara landslide.
This difference may reflect that the degree of weathering affected block bottom erosion during movement.
Considering that the Ichinose landslide material could have been saturated, as suggested by IversonAos coefficient.
L/H, the saturation of the landslide material may have also contributed to lower erosion coefficients.
Investigating the loss mechanisms of landslide materials during movement is necessary, particularly by considering their physical properties, such as particle size distribution, weathering, and water content.
Journal of the Civil Engineering Forum Figure 12 displays the distributions of landslide deposit thickness and river flow depth one hour after the landslide in the Ichinose district for cases where only the value differs, while cb and Ublim were fixed at 50 N m-2 50 m s-1 , respectively.
In the case where = 0.
the landslide material did not reach the river channel, and consequently, waterlogging did not occur.
The case with = 0.
01 produced results closer to the actual inundation damage.
However, the calculated results only partially matched the observed areas of landslide deposition and waterlogging.
This discrepancy is attributed to a decrease in the volume of cylindrical blocks caused solely by bottom erosion during their movement.
According to Figure 13, which shows the temporal variation of block-averaged thicknesses and movement velocities for both landslide cases under the combination of = 0.
01, cb = 50 N m-2 , and Ublim = 0.
50 m s-1 , the velocities of the Obara landslide were lower than those of the Ichinose landslide due to its milder slope inclination.
However, the erosion rates of the blocks in both cases were similar, with erosion being completed in approximately the same amount of time.
This suggests that the developed model needs to incorporate additional processes for block decomposition, such as soil
5 CONCLUSION
We developed a numerical model to predict the formation of various channel blockages by incorporating combined conditions of topography, landslide volume, and river discharge.
The model successfully reproduced two landslides, of different scales and their impacts on river flows by optimizing the soil-related parameters of the cylindrical block model.
Although the simulated river blockage in Obara persisted for two hours, postdisaster surveys showed it had already disappeared.
Further investigation is needed to determine whether the blockage could have been cleared by overtopping flows during an extended simulation period.
The appropriate cb and Ublim values for both landslides were the same, which were 50 N m-2 and 0.
50 m s-1 .
The cb value is an order of magnitude .
-1 ) less than that of general physical soil properties.
The difference between actual and modeled cb values should be further explored in future studies.
The appropriate value for the IchinoseAos landslide was much smaller than that for the ObaraAos landslide.
The results suggest that the difference for weathering progress and saturation of both landslide materials effect on the loss of the block during their movement.
These findings indicate that the loss of landslide material is influenced not only by bottom erosion and material fluidization but also by additional mechanisms related to their physical property, such as particle size distribution, weathering, and water content, such as soil dispersion.
To enhance the accuracy of the model in predict242 Vol.
11 No.
3 (September 2.
ing various channel-blockage formations, future studies should focus on investigating the transport mechanisms of the landslide materials, incorporating a wider range of values for these physical factors.
Although the developed model still needs to be improved in identification of optimal value and consideration of the loss of landslide material, it is useful for estimating the magnitude and area of damage caused by a large-scale landslide and the associated channel blockage and waterlogging in various river channels with their side steep slopes.
The calculated results can be utilized in investigating disaster countermeasures for the landslide in the area.
DISCLAIMER
The authors declare no conflict of interest.
ACKNOWLEDGMENTS
This study was supported by a Research Grant by Sabo and Landslide Technical Center (STC).
REFERENCES