Early

warninig systems in landslide occurrences using modelling the landslide flow

paths

S. Nishanthan

Civil Engineering Department, University

Of Moratuwa, Katubedda

Prof. S. A. S. Kulathilaka

Civil Engineering Department, University Of Moratuwa,

Katubedda

Abstract: Landslide

is one

of the most frequent and catastrophic geologic hazards. Mostly landslide is

triggered due to extremely high rainfall events. This research is carried out mainly about

modelling the landslide flow paths during after triggering landslides. During these landslides, it caused

fatalities, loss of life and properties. So this research is significance to

avoid this kind of destructions and this research will be acted as an early

warning system to avoid these hazards & losses. The landslide flow path of

known landslide occurrence area is modelled based on using GIS software,

TRIGRS, and DAN 3D software to validate that flow path with original landslide

flow path. The landslide zones are found out from using the GIS software and

the initiation zone of landslide is obtained by using the TRIGRS software.

Thereafter DAN3D software is used to analyse the landslide flow path. It is necessary to

select a susceptibility model with high prediction capability which depends on

the methods used. Because of spatial prediction of landslides is considered to

be useful for land-use planning and the ?rst important step in landslide hazard

and risk assessment. Finally

early warning system will be proposed using these results with respect to rainfall

precipitation data.

Keywords: GIS

software, TRIGRS software, DAN3D, Debris flow, Landslide flow path,

Precipitation

1.

Introduction

Land slide is defined as the

movement of a mass of a rock, debris or mass of failed soil and other matter

down a slope which are affected by gravity. It is one of the most frequent and

catastrophic geologic hazards and it causes much destruction in its flow path

causing fatalities, loss of life, property and severe economic losses (Mahalingam

& Olsen, 2016). Due to extremely high

rainfall events, landslide is triggered. 80% landslides were triggered due to

the excessive rainfall

events. This has been witnessed in number of landslides such as Meeriyabedda

landslide at Badulla was caused by the extremely high rainfall events. Another

witnessed is preliminary landslide inventory from Badulla is prepared by

Badulla meterological station based on special reports. Such are 25 landslide

days between 1991 and 2015, 5 landslide days in 2014, and 17 landslides days in

2015. Research

is carried out mainly about modelling the landslide flow paths during after

triggering landslides. During these landslides, it caused fatalities, loss of

life and properties. So this research is significance to avoid this kind of

destructions and this research will be acted as an early warning system to

avoid these hazards & losses. There was no any landslide flow path

modelling with respect to rainfall factor. Scope of this study is build-up a

relationship between rainfall precipitation and the landslide flow path.

From this results an early warning

systems will be proposed.

Figure 1: Landslide analyse area

2.1 Application of GIS Software

GIS software is used to access landslide inventory

map and extract the particular landslide area from total area of study zone.

During GIS based landslide susceptibility modelling various sophisticated

machine learning techniques have been suggested by the previous researchers. The

performances of the model are differed compare with other models. The

prediction capability of a susceptibility model depends on those models. So

already a research was conducted for investigation and comparison on some

advanced machine learning methods such as Kernel logistic regression,

Naïve-Bayes tree model and Alternating decision tree model for landslide

susceptibility mapping. In this research 12 landslide conditioning factors were

considered which influence on models during the modelling such as distance to

roads, distance to river, distance to faults, altitude, land use, lithological

unit, mean annual precipitation, profile curvature, plan curvature, slope

angle, slope aspect and NDVI. Those data sets were collected from the National

Aeronautics and Space Administration (NASA) website (http://reverb.echo.nasa.gov/reverb/).

Slope aspect, slope angle, altitude,

pro?le curvature, and plan curvature were extracted using ArcGIS 10.0 software

from a digital elevation model (DEM) which model was extracted from the ASTER

GDEM data. Distance to rivers, distance to roads, and distance to faults were obtained

from topographic map (1:50,000), and geological map (1:500,000) respectively.

Mean annual precipitation map was produced by assembling data of the particular

area. Frequency ratios of each landslide conditioning factor were calculated

using the frequency ratio model and those ratios were normalized in the range

0f 0-1. Thereafter those normalized frequency ratios were used as inputs to

build a model and produce the landslide susceptibility maps. Subsequently, models

were explored that what contributions to landslide susceptibility models by different

condition factors. (Wei

Chen, GIS-based landslide susceptibility modelling: a comparative assessment

of kernel logistic regression, Na€ ?ve-Bayes tree, and alternating decision

tree models, 2017)

Figure 2: Contribution to landslide susceptibility models by different

condition factors

Distance to roads, distance to river, altitude and

land use have a highest importance to the three models and slope aspect, plan

curvature and profile curvature have low predictive capabilities. But from this

12 condition factors, mean annual precipitation is very important factor to

trigger the landslides. Because 80% landslide have been triggered by the

rainfall precipitation in Sri Lanka. The contribution of distance to roads occupies

the highest percentages of 17.667%, 19.846%, and 18.945% for KLR, NBTree, and

ADTree models, respectively. 80.4% of landslide locations occurred due to contribution

of distance to roads along the main valleys and road networks (less than 500 m).

Therefore, we can come to conclude from this that landslide conditioning

factors tend to have different contributions depending on the types of models

used. The researcher has come to conclude that the KLR model is better model, considering

the overall performance and the model construction from this paper. Because of

KLR model has the highest degree of goodness-of-?ts on both training and

validation dataset. The KLR model shows 84.5% stable classi?cation ability for

the training dataset. But NBTree model has the highest goodness-of-?ts (91.4%)

for the validation dataset and not for training model. Therefore

KLR model was selected as a best model for the highest degree of

goodness-of-?ts on both training and validation process to produce the landslide

susceptibility maps for land-use planner to choose suitable construction sites.

(Wei

Chen, GIS-based landslide susceptibility modelling: a comparative assessment

of kernel logistic regression, Na€ ?ve-Bayes tree, and alternating decision

tree models, 2017)

KLR model is based on a kernel version of logistic

regression. It is carried out as follows:

Where “w” is

a landslide conditioning factors, (.) is a

nonlinear transformation to each input variable, and b is a vector of constant.

A logit function can be written as follows:

Then the

transformation equation was detrived.

Then they

defined the inner product between the images of vectors as the kernel function.

That is;

Several kinds

of kernel functions can be used. But in this case radial basis function kernel

is applied. It is the most commonly used in practical applications which is as

follows;

Where is the turning parameter that controls the

sensitivity of the Kernel. An optimal “w” that can be found by minimizing a cost function according to the

representer theorem (Kimeldorf & Wahba 1971; Scholkopf et al. 2001). These

are used equations in this research for KLR modelling. (Wei

Chen, GIS-based landslide susceptibility modelling: a comparative assessment

of kernel logistic regression, Na€ ?ve-Bayes tree, and alternating decision

tree models, 2017)

2.2 Application of TRIGRS software

The TRIGRS (Transient Rainfall Infiltration and Grid-Based

Regional Slope-Stability Model) software is used for

find-out the initiation zone of the landslide. The TRIGRS software is used for

conjunction with GIS software which is used to prepare inputs data for

inspecting different combinations of geotechnical parameters. TRIGRS has

simulated the dynamic hydraulic conditions within slopes triggered by rainfall

precipitations. In the

TRIGRS model for rainfall-induced landslides research (Viet,

2017)

they applied the TRIGRS model and Scoops3D for the 3D prediction of potential

landslides. Combination

of TRIGRS and Scoops3D software approach is an effective tool for 3D, spatially

distributed and time-dependent assessment of rainfall-induced landslides.

Further developments of mapping of the orthophotographs contribute to determine

the actual size and shape of each landslide and help

to provide a detailed evaluation of the sizes of landslide-triggering areas.

This approach was explored and results were compared with the landslide

initiation locations with respect to the rainfall events on landslide event in

July 2011 at Mount Umyeon, South Korea. In Three-dimensional, time-dependent modelling of

rainfall-induced landslides over a digital landscape research the

soil depth was analysed by using three approaches method such as stability

maps, obtained by the 1D (TRIGRS only) and 3D (TRIGRS and Scoops3D)

time-dependent approaches. (Viet,

2017)

Figure 3: Initiation area of landslide

Almost 1D

& 2D slope stability analyses provides results that suggest lower stability

than 3D analyses methods (Chakraborty

& Goswami, 2016). When modelling the actual mechanism of

landslides, if we use 1D or 2D approaches to that modelling then that model may

fail compare with original one (Bromhead

, 2002).

That is not all we can not consider the direction of the slip surface in 1D OR

2D modelling. That sliding surface is forced to move in assumed downward

direction (Bromhead

, 2002).

So 3D analyse methods are followed. In

this research they came to conclude that combined TRIGRS and Scoops3D approach

(3D model) is an effective tool for 3D, spatially distributed and

time-dependent assessment of rainfall induced landslides. Because of that most

of landslides are predicted by appropriate time of landslide triggering effect

due to the rainfall intensity. First pore water pressure, initial ground water

table and spatially distributed soil depth were calculated using 1D model

TRIGRS. Thereafter Scoops3D model was executed to measure the subsequent 3D

slope stability using previous calculated value of pore water pressure value.

Figure 4: Time dependence of

the percent of

unstable area predicted

by the 1D

and 3D

2.3 Application

of DAN3D Software

The areas of landslide run-out is estimated by DAN3D

which is suitable software for dynamic analysis of rapid flow slides, debris

flows and avalanches carry through numerical model developed by McDougall and

Hungr (Hungr,

2010).

This run out analysis approach is based

on continuum-based dynamic simulation models which are based on equations of conservation

of mass and momentum.

3.

Conclusion

Some details & parameters of recent landslide area will

be selected relative to that landslide before modelling the flow path. After

collecting that particular data and information, modelling will be commenced. Those

results should have to validate and check whether that landslide flow paths

modelling is correlated with original landslide. First of all we should focus

on processing of inventory map. So the particular landslide area will be

extracted from total area of study zone using the GIS software based on KLR

model. Then using the combination of TRIGRS and Scoops3D software, the

initiation zone of the landslide will be found out. Finally this result will be

applied in DAN 3D software to get a run out areas with the run out distance,

maximum depth of transport zone and velocity of flowing masses in transport

zone for the dynamic analysis of rapid flow slides and debris flows in the

landslides. This run out analysis approach is based on continuum-based dynamic

simulation models which are based on equations of conservation of mass and

momentum. Then the results will be validated and checked whether with original

landslide flow paths. If it will be validated with the original landslide flow

path then another landslide occurrence area due to the excessive rainfall will

be randomly selected and landslide flow path will be obtained with respect to

the different amount of the precipitation. From this critical value of

precipitation amount will be obtained and that will be checked with original

precipitation value. If this result will be validated with original

precipitation value then we can come to conclude that landslide will be

triggered due to excessive rainfall than critical value of precipitation. This

is act as an early warning system. Because, if we can predict the landslide

flow path during excessive rainfall than critical value of precipitation. Hence

we can avoid the landslide hazards, risks and vulnerability during the

landslides.

Acknowledgements

I would like to convey my gratitude to all who

helped me to successfully conduct my research project.

First of all I would like to thank to Eng. Prof. S. A. S. Kulathilaka professor of civil

engineering department, University of Moratuwa for support to my research

project by giving proper guidance and data as a research supervisor. Next I

like to thank to Dr. Ashani Ranthunga to encourage and guide to write

the research proposal and research literacy view in proper way. Then I like to

convey my gratitude to my friends who are help to my project.

Bibliography

Olsen , M., & Mahalingam , R. (2016). Evaluation

of the in?uence of source and spatial resolution of DEMs on derivative

products used in landslide mapping. Geomat Nat Haz Risk, 7:1835–1855.

National Symposium On Landslides In Sri Lanka.

(1994). Srilanka: National Building Research Organization.

Bromhead , E. (2002). Three-dimensional stability analysis

of a coastal landslide at Hanover Point, Isle ofWight. Quart J Eng Geol

Hydrogeol , https://doi.org/10.1144/qjegh.35.1.79.

Chakraborty, A., & Goswami, D. (2016). Three

dimensional (3D)slope-stability analysis. International Journal of

Geotechnical Engineering, https://doi.org/10.1080/19386362.2016.1172807.

Corominas, J., Bonnard, C., Cascini , L., Leroi , E.,

Savage, W., & R, F. (2008). Guidelines for landslide susceptibility,

hazard and risk zoning for land-use planning.

Dikau, R., Denys Brunsden, & Lothar Schrott. (n.d.). Landslide

Recognition, Identification, Movwment and Causes. Geomorphology

Geomorphologie.

Hungr, O. (2010). Dynamic Analysis of Landslides in

Three Dimensions. West Vancouver, B.C., Canada: Geotechnical Engineering,

Inc.

Mahalingam, R., & Olsen, M. (2016). Evaluation of the

influence of source and spatial resolution of DEMs on derivattive products

used in landslide mapping. Geomat Nat Haz Risk.

Viet, T. T. (2017). The TRIGRS model for rainfall-induced

landslides. Research gate.

Wei Chen. (2017). GIS-based landslide susceptibility

modelling: a comparative assessment of kernel logistic regression, Na€

?ve-Bayes tree, and alternating decision tree models. Geomatics, Natural

Hazards and Risk.

Wei Chen, Xiaoshen Xie, Jianbing Peng, Jiale Wang, Zhao

Duan, & Haoyuan. (2017). GIS-based landslide susceptibility modelling:. GEOMATICS,

NATURAL HAZARDS AND RISK, 2017,

http://www.tandfonline.com/doi/pdf/10.1080/19475705.2017.1289250.

Wikipedia. (n.d.). Retrieved from

https://en.wikipedia.org/wiki/Causes_of_landslides:

https://en.wikipedia.org/wiki/Causes_of_landslides