Landslide Studies Division

GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________

Landslide Susceptibility Analysis in GIS

What is Landslide Susceptibility?

Landslide susceptibility is a quantitative or qualitative assessment of the spatial distribution of landslides, which exist or
potentially may occur in an area. Although it is expected that landsliding will occur more frequently in the most
susceptible areas, in the susceptibility analysis, time frame is explicitly not taken into account.

Methods for Susceptibility Mapping

Landslide susceptibility is usually based on two assumptions:
A susceptibility map gives
First, that areas which have experienced landslides in the past are likely to
the qualitative or
experience them in the future, and
quantitative estimation
Second, that areas with a similar set of geo-environmental conditions, such as
that the mapping unit will
geology, slope, land use etc as that of the failed areas, are also likely to fail in
be affected by a future
future.
landslide source, but the
This means that the quantitative estimates of the spatial location of future landslide
“future” here is not
sources depend on the detailed information on the distribution of past landslides and a
connected to a particular
set of thematic variables such as slope angle, lithology etc. that has initiated these
time period, such as the
landslides. The second assumption facilitates in predicting the geographical location of
next day or the next 5
future landslides in passive areas (i.e., areas presently devoid of landslides) provided the
years or 50,000 years. It
geo-environmental conditions remain the same. The above assumptions have been
only indicates the
successfully used for statistically quantifying landslide susceptibility using landslide as
stability status of a slope
dependent variable.
irrespective of the time of
Landslide susceptibility can be estimated using a variety of statistical techniques. BIS has an event. This temporal
proposed Guidelines for a qualitative landslide susceptibility mapping (BIS-code; IS aspect is considered in
14496, Part 2), which is an indirect heuristic approach. The methods can be either landslide “Hazard”
qualitative or quantitative. The qualitative ones are generated using landslide inventory analysis.
analysis and knowledge driven methods, and quantitative ones are generated using data driven, deterministic, and
probabilistic methods (Fig 2.1).

Fig. 2.1 Methods for susceptibility and hazard assessment. Blue relates to landslide inventories. Red to susceptibility and
Green to Landslide hazard.
1
Landslide Studies Division
GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________
In this tutorial you will learn about landslide-dependent Bi-variate statistical method for susceptibility mapping.
In a bivariate statistical analysis (Fig. 2.2), each factor map (slope angle,
litholohy, land use etc.) is combined with the landslide distribution map,
and weight values, based on landslide densities, are calculated for each
parameter class (slope class, lithological unit, land use type etc). Several
statistical methods can be applied to calculate weight values, such as
the information value method, weights of evidence modeling, Bayesian
combination rules and fuzzy logic. The analysis is mostly carried out on a
grid (pixel) level.
There is a clear link between the scale of analysis and the type of
method that can be used, basically related to the possibility of obtaining
the required input data. Figure below gives a summary

Scale Qualitative methods Quantitative methods

Inventory Heuristic Statistical Deterministic Probabilistic
<1:10k Yes Yes No Yes Yes
1:25k – 50k Yes Yes Yes No Yes
>1:100k Yes Yes Yes/No No No

Fig. 2.2 Bi-variate statistical method for susceptibility mapping

(source: ITC)

The landslide susceptibility analysis should be carried out separately for debris type landslides and rock type landslides
because factors governing slope instability in both cases are different. It is strongly recommended that landslide zoning
be carried out in a GIS-based system so that the zoning can be readily be applied for land use planning and can be up-
dated as more information becomes available.

Validation of susceptibility maps

The most important question to be asked for each susceptibility mapping is related to the degree of accuracy. The
evaluation of the accuracy of a susceptibility map is generally very difficult. In reality a landslide prediction can only be
verified by observing if the event takes (or has taken) place in time ("wait and see"), but this is not a very useful method,
for obvious reasons.
For statistical and knowledge driven method the best way is to use success rate curves. The success rate is a statistical
method to determine how well the resulting susceptibility map has classified the areas of existing landslides as
susceptibility areas. It is merely a method that allows checking how many of the landslides occur in the high
susceptibility areas (well classified) and how many occur in the low susceptibility areas (wrongly classified).

Landslide Susceptibility Analysis in GIS

In this practical, we will demonstrate how to analyse landslide susceptibility on a macro-scale (1:50,000/25,000) using
landslide-dependent Bi-variate statistical method, being used in GSI under National Landslide Susceptibility Mapping
(NLSM) Project.
The method is demonstrated in a series of steps with example from Mandakini valley, Uttarakhand for an area covering
~140 sq. km. The mapping-unit considered for analysis is ‘pixel’, a square grid.

2
Landslide Studies Division
GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________

Folder: Training_data
It contains File Geodatabase under name: Data.mdb (projection WGS84_UTM_44N)
It contains Feature dataset [a boundary file (boundary), mapped landslide polygons (landslide), a thematic map of
Landuse and Landcover (lulc) and Geomorphology (Geom)]
It contains Raster layers [a DEM (dem50), a Slope map (slpcls50) and a susceptibility score map (susc_all)]

Step 1: Creating landslide raster layer

The first step is to create landslide raster layer for carrying out landslide dependent analysis. It requires selecting
landslide polygon (landslide scars) from the landslide database. Here we consider only landslide scar for susceptibility
analysis as we are interested to know the environment of the failed slope. Landslide runout and deposition may lead to
misleading interpretation as these are deposits over the stable/on failed earth mass.

Selecting depletion area of specific landslide types: Open ArcGIS (ArcMap).

In Arc Catalog, explore to Geodatabase: Data.mdb / maps / landslide. Select landslide and drag to TOC of ArcMap.
The landslide polygon map will be displayed.
Right click on landslide and select Open Attribute Table. A Table box will open. From dropdown click on Select by
Attribute / and select Name_of_sl/ Get Unique Values.
Double click on Name_of_sl and select same expression / [Name_of_sl] = 'Debris slide' OR [Name_of_sl] = 'Debris
subsidence' OR [Name_of_sl] = 'Rock slide' / click Apply

3
Landslide Studies Division
GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________
Landslide scar polygon smaller than one-pixel size, if rasterised will lead to deletion of the polygon. In this exercise we
intend to carry out pixel-based analysis using pixel of size 50 m × 50 m. In the following steps you will understand how
to select smaller landslide polygons from the landslide database for making a composite landslide layer for analysis.
Selecting smaller landslides (area <2500 sq. m i.e., one pixel size)

Right click on landslide in Table of Contents (TOC) / selection / Create layer from selected feature. A new layer of the
selected file will be created and opens up
Right click on landslide selection in TOC / Data / Export Data / select a destination folder (or working geodatabase)
and name it Ls / Click OK
Right click on Ls in TOC / Open Attribute table
Select by Attribute box will open (as in figure) and select [Shape_Area] and select expression [Shape_Area] <=2500 /
click Apply [all landslide polygons smaller than or equal to 2500 will be selected.
Right click on Ls in TOC/ Data / Export Data /select destination folder and name it ls_l25 / OK [this gives all
landslide scar with area <=2500 sq. m]

In the following steps you will understand how to select landslide polygons larger than one-pixel size for ratserisation.

4
Landslide Studies Division
GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________

Right click on Ls in TOC / Open Attribute table/select switch selection [If the selection is removed then step above
to select landslide with area >2500 sq. m]
Right click on Ls in TOC / Data / Export Data /select destination folder and name it ls_m25 / select OK [this gives
all landslide scar with area >2500 sq. m]
[Now you will see how to convert polygon layer into point layer].
In search window type “feature to point” and select feature to point tool (Data Management)
In Feature to Point box, Select Input as ls_l25 and in output select the destination folder and name as ls_l25_pt.
Check Inside (optional). [this will generate point file with area less than 2500 sq.m].
Right click onls_m25 in TOC / Open Attribute table/ click Add Field /Give Name as Slide, Type as Text and Length
as 10 / click OK

In attribute table go to Slide column and right click and press Field Calculator and in field calculator write “LS” / click
OK [use “” to write the text attribute]
Now you will see how to convert polygon layer into raster layer. In search window type “polygon to raster” and select
polygon to raster tool.
A Polygon to Raster box opens. Select ls_m25 as input and output name as ls_m25_r/ Give Value as Slide / set cell
size at 50/ click OK. [this is the Raster landslide file with area more than 2500 sq.m]
What you have done in the above steps, can you guess?
[If you convert the landslide polygon <2500 sq. m to a raster layer of 50 m pixel then more likely you loose many of
the smaller landslides as these may not get converted into a raster map. Therefore, we take two-way approach: First
we convert smaller landslides to point to get at least one point per slide polygon and then converting it to raster. This

5
Landslide Studies Division
GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________
also results in miss representation of actual landslide size (Can you guess how?). Second we directly convert larger
landslide to raster and add both together to get a composite landslide raster map]
In search window type “raster to point” and select raster to point tool (Data Management)
Select Input as ls_m25_rand give output name as ls_m25_pt/ Select Field as Slide / click OK. [this is the point
landslide file with more than 2500 sq.m].

[Now you will see how to join two landslide maps to get a composite map].

Right click on ls_l25pt in TOC / Open Attribute table / select Select All option / go to “Edit” and select Copy
[copying all ls<2500 sq. m for merging it with ls>2500 sq. m].
Right click on ls_m25_pt in TOC and select Edit Features from main window of ArcGIS / Start Editing.
Now again select Edit Features from main window of ArcGIS / Paste (select Target as ls_m25_pt, if this layer is
not appeared make sure this layer is on in TOC) / click OK
Now go to editor / Save Edit / Stop Edit

Right click on ls_m25_pt on TOC / Open Attribute table /Select grid_code / right click grid_code and select “Field
Calculator” / write 1 in calculator field.
In search window search for “point to raster” and select “point to raster” tool (Conversion)
Give Input feature as ls_m25_pt / Value field as grid_code / Cell size set as 50/Output Raster as ls_all/ click OK. [This
is your final raster landslide with 50 m pixel used for further calculation].

6
Landslide Studies Division
GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________
NOTE: Here you can have two raster landslide dataset; 1. For calibration of
the model (using 75% of the landslide points selected randomly or event-
wise) and 2. For Validation (rest of the data).

Step 2: Calculating Ratings and Weights of thematic maps

Thematic data preparation: Here we learn how to convert a polygon feature

dataset to a raster for facilitating calculation of ratings and weights

Open Geodatabase: Data.mdb / maps / lulc

Right click on lulc on TOC and select open attribute table[see the attributes, there
are 11 lulc classes and all polygons are dissolved]
In search window type “polygon to raster” and select “polygon to raster tool”
(Conversion)
In a polygon to raster tool box, select Input Feature as lulc, Value field as class,
Cellsize as 50, and give filename as lulc_raster and output location under Output
Raster Dataset. Leave all other as default. [this will create a raster layer of lulc
map]
Right click on lulc_raster map on TOC and select open attribute table [see the
attributes, there are 11 lulc classes as shown under VALUE and COUNT is the
number of pixel. To know the area multiply pixel value with 2500 (50 m × 50 m
pixel size]
Open an excel sheet and name it as Analysis. Name one sheet as LULC and make
column as ID, Name etc as shown in the training material. Under Name put all
the class name of lulc as given in lulc_raster. NpixC is the total pixel count of a given class (read the COUNT field in
lulc_raster).

Calculation of Ratings and Weights: Here we learn how to calculate ratings and weights with landslide as dependent
variable

Open lulc_raster and ls_all in TOC.

In search window type “combine” and select tool „combine (Spatial
Analyst)‟ [this tool is used to superimpose landslide with thematic map]
Use Input rasters as lulc_raster and ls_all; create output raster as
lulcXls_all / OK
Right click on lulcXls_all in TOC / Open Attribute table
Select by Attribute lulc_raster and sort it as ascending. [the numbers
under lulc_raster denote the 11 lulc classes and Count denotes the number of pixel of
landslide falling within each lulc class]
Open an excel sheet Analysis. Enter „Count‟ values under field T11 [T11 is the npix
that intersected the lulc of a given class]. Now calculate all other values in the excel.
LOFS is the rating of the factor class computed based on the spatial distribution of
landslide considering also the spatial area coverage of the class.

[Similarly calculate LOFS for all categorical factor class such as Geom provided in the
training dta] Example of slope is shown below
Generating SLOPE map from DEM.
Open „dem50‟ in TOC

7
Landslide Studies Division
GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________
In search window type “slope” and select tool „Slope‟ (3D Analyst) [this tool is used to calculate slope angle map from
DEM]
In Slope tool, give Input Raster as dem50, Output Raster as Slope and rest leave as default / OK
In search window type “Reclassify” and select tool „Reclassify‟ (3D Analyst) [this tool is used to reclassify slope angle
into discrete class]
In Reclassify tool, put Input Raster as
Slope, Reclass field as „value‟ / Classify / in
Classification window, select Classes as „10‟
and give Break Values as shown below / OK
/ Output Raster as Slp_cls and rest leave as
default / OK
Right click on Slp_cls in TOC / Open
Attribute table /[see the attributes, there
are 10 slope classes as shown under VALUE
and COUNT is the number of pixel]
Open Slop_cls and ls_all in TOC
In search window type “combine” and
select tool „combine‟ (Spatial Analyst) [this
tool is used to superimpose landslide with thematic map]
Use Input rasters as Slp_cls and ls_all; create output raster as SlpclsXls_all / OK
Right click on SlpclsXls_all in TOC / Open Attribute table
Select by Attribute Slop_cls and sort it as ascending. [the numbers under Slp_cls denotes the 10 slope classes and
Count denotes the number of pixel of landslide falling within each slope class]

[SIMILAR to LULC above fill the excel Analysis sheet for SLOPE]
[FOR LINEAR FEATURE USE BUFFER/MULTIPLE RING BUFFER TOOL TO GET POLYGON and proceed]
In ‘Analysis’ table calculate Ratings and Weights as documented.

Preparation of LOFS maps: Here we learn how to combine excel sheet containing ratings and weights with thematic
maps in GIS to generate attribute (LOFS) maps

Open lulc_raster and excel worksheet „LULC‟ in TOC. Right click on

lulc_raster and select open attribute table / Table Options / Join and Relates
/ Join [This opens Join Data box]
In Join Data box, under 1 put CLASS (Class field in LULC will act as a key for
joining), in 2 select LULC sheet from excel, in 3 put Name (Name field in
excel LULC sheet will act as a key for joining). Keep all as default / OK.
Right click on lulc_raster map and select open attribute table
[You can now see the two tables to be joined. Note this joining is temporary
and will be removed if ArcGIS is closed OR you can export the table and
save it]

8
Landslide Studies Division
GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________

In search window type “Lookup” and select „Lookup‟ (3D Analyst) [This tool will help to generate attribute map of
LOFS Lulc]
In „Lookup‟ box, put input raster as lulc_raster, lookup field as „LOFS’ and Output raster as LOFS_lulc / OK.

[SIMILAR to LOFS_LULC, generate LOFS maps for all factors].

Preparation of SUSCEPTIBILITY SCORE map: Here we learn how to integrate all LOFS maps to generate a susceptibility
score map

Open all LOFS raster maps in TOC.

In search window type “Raster Calculator” and select „Raster
Calculator‟ (Spatial Analyst) [This tool will help to add all raster
layers
In „Raster Calculator box, put the expression as indicated below and
Output raster as susc_all / OK. [add all LOFS raster map by double
clicking on it and multiply it by weight and finally divide all by
summation of weights]
Open susc_all in TOC. This a continuous raster map showing range of
cumulative LOFS value per pixel.
This map needs to be classified into different landslide susceptibility
zones (High, Moderate and Low).

Preparation of success rate curve and classified susceptibility map: Here we learn how to generate success rate curve
for getting boundary condition for different susceptibility zone.

Open susc_al and ls_all raster maps in TOC

9
Landslide Studies Division
GHRM Centre, Geological Survey of India Practical Tutorial #2 Updated May 2018
__________________________________________________________________________________________________
In search window type “Reclassify” and select tool „Reclassify‟ (3D Analyst) [this tool is used to reclassify susceptibility
score map into discrete classes]
In Reclassify tool, put Input Raster as susc_all, Reclassfield as „value‟ / Classify / in Classification window, select
Method as Quantile and Classes as „20‟ and (Note the Break Values and write down in excel) / OK / Output Raster as
susc_relas20 and rest leave as default / OK
Open an excel sheet Success rate Calculation. Under field Class-break enter break value in reverse order, starting with
highest value on top and lowest break value on bottom.

Right click susc_reclas20 in TOC / open attribute table / in Table box, right click on Value field and sort descending /
copy pixel from Count and paste in excel (Success rate Calculation) under field NpixC.
In search window type “combine” and select tool „combine‟ (Spatial Analyst) [this tool is used to superimpose ls_all
with susc_reclas20 map]
Use Input raster as susc_reclas20 and ls_all; create output raster as susccls20Xls_all / OK
Right click susccls20Xls_all in TOC / open attribute table / in Table box, right click on sucs_reclas field and sort
descending / copy pixel from Count and paste in excel (Success rate Calculation) under field Sld.

In excel (Success rate Calculation) note the Class-break value corresponds to 90% and 70% of CumSld% (this is
cumulative percentage of landslide) and use these value to reclassify the susceptibility map into different susceptibility
zones.

Open susc_al lraster map in TOC

In search window type “Reclassify” and select tool „Reclassify‟ (3D Analyst) [this tool is used to reclassify susceptibility
score map into discrete classes]
In Reclassify tool, put Input Raster as susc_all, Reclassfield as „value‟ / Classify / in Classification window, select
Method as Manual and Classes as „3‟ / under Break Values put first value of 70% of CumSld% as notes above and
second 90% of CumSld%/ OK / Output Raster as susc_final and rest leave as default / OK [70% break value will be
the boundary limit of HIGH and MODERATE and 90% break value will be boundary limit of MODERATE and LOW
susceptibility class]

10