J. Earth Syst. Sci.

(2020)129:30 Ó Indian Academy of Sciences

https://doi.org/10.1007/s12040-019-1277-4 (0123456789().,-volV)(0123456789(
).,-volV)

Susceptibility assessment of rainfall induced debris flow

zones in Ladakh–Nubra region, Indian Himalaya

H S NEGI*, ANANT KUMAR, N NARASIMHA RAO, N K THAKUR,

M S SHEKHAR and SNEHMANI
Snow and Avalanche Study Establishment, Him-Parisar, Sector-37A, Chandigarh 16036, India.
*Corresponding author. e-mail: [email protected]

MS received 17 April 2019; revised 29 July 2019; accepted 4 August 2019

In recent past, rainfall-induced debris flow events in Ladakh–Nubra region have caused loss of lives and
damages to civil infrastructures and army locations. Therefore, there is a need of high spatial and
temporal monitoring of precipitation, and further to assess susceptible rainfall-induced debris flow zones
in the area. We assessed the rainfall data collected at two gauge stations and observed a significant
increase in the rainfall amount over the study region during summer-monsoonal period 1997–2017.
Increasing trend was also observed from CRU gridded precipitation dataset. A GIS-based multi-criteria
evaluation (MCE) method was performed by combining topographical, environmental and hydrological
parameters for mapping of rainfall-induced susceptible zones. Suitability analysis of precipitation fore-
casts from WRF model at higher resolution (3 km) was also performed. A good agreement (r = 0.76) was
observed between 4-day model forecast and field observed rainfall. Further, the simulated precipitation
from WRF was incorporated into GIS model for assessment of debris flow susceptible zones for two cases
of heavy precipitation events. The modelled high, medium, low and very low risk susceptible zones
identified for the year 2015 events are validated with field survey and pre-post satellite imageries, and
found in good agreement (ROC = 76.6%). The model was able to identify affected areas during the Leh
cloud burst event in year 2010. In addition, a threshold value of rainfall for initiation of debris flow in the
region was also reported.
Keywords. Debris flow; Karakoram–Himalaya; rainfall; WRF; CRU; weighted overlay.

1. Introduction triggered various mass movement events in the

region and caused large scale destruction and
Debris flow and landslide triggered due to heavy several deaths (Juyal 2010). Similarly, during
or prolonged rainfall are amongst the most dev- 15–17 June 2013, torrential rainfall triggered
astating hazards over the Himalayan region as numerous landslides and debris flows in and
these are responsible for economic loss, injuries around Uttarkashi districts, claimed lives of more
and loss of lives during the monsoon season than 5700 people (Martha et al. 2015). In another
(Bookhagen and Burbank 2010; Bhan et al. 2015). event, heavy rainfall triggered debris flows at
In recent past during first week of August 2010, numerous places of Darjeeling Himalaya in
Leh district experienced 2 hrs of concentrated June–July month of 2015 and claimed lives of 38
rainfall reaching a peak intensity of *75 mm over people (Biswas and Pal 2016). More recently in
a 30-min period (Hobley et al. 2012), which July 2019, landslide events triggered by incessant
30 Page 2 of 20 J. Earth Syst. Sci. (2020)129:30

monsoonal rains across Nepal have claimed 78 extensive weather networks required to successfully
lives and displaced 17,500 others in the year 2019 observe these conditions. In recent years,
(‘Nepal floods’: Pradhan 2019). Above-mentioned space-borne platforms and computing technologies
studies and reports suggest dominant role played are effectively employed for hazard detection,
by rainfall in triggering debris flow and landslide monitoring and susceptibility assessment over
events. In addition, a few such activities may also complex terrain (Pardeshi et al. 2013). Remotely
get induced by tectonic activities (Adhikari and sensed gridded precipitation datasets such as Glo-
Koshimizu 2005) and lake/stream burst (Owen bal Precipitation Measurements (GPM) and
et al. 2008). Under on-going climate change, any Tropical Rainfall Measuring Mission (TRMM)
possible shift in the rainfall pattern over the offer high temporal and spatial scales to improve
Himalayas may lead to upsurge in frequency and the understanding of landsliding and debris flow
intensity of these hazardous events (Bhutiyani events (Nikolopoulos et al. 2014; Kirschbaum and
et al. 2010). Thus, it becomes imperative to Stanley 2018). Apart from rainfall measurements,
understand the threat posed by heavy or pro- optical remote sensing imageries such as Landsat,
longed rainfall, especially for the region under low Advanced Space-borne Thermal Emission and
rainfall zone, where level of preparedness may not Reflection Radiometer (ASTER), Sentinel-2, etc.,
be high by weather forecasters or disaster man- and digital elevation models (DEMs) such as
agement authorities (Ziegler et al. 2016). One Shuttle Radar Topographic Mission (SRTM),
such area situated in Nubra–Ladakh region, which ASTER GDEM, etc., are widely adopted to
is currently facing increased vulnerability to cli- retrieve aforementioned factors like land use/land
mate hazards and related risks (Barrett and cover (LU/LC), drainage, slope, curvature, vege-
Bosak 2018). After deadlier 2010 event in Leh, tation, etc. (Elkadiri et al. 2014).
another heavy rainfall event occurred during first Blais-Stevens and Behnia (2016) suggested the
week of August 2015 in Nubra region, where quantitative and qualitative techniques to develop
continuous rainfall caused series of debris move- debris flow susceptibility map. Quantitative tech-
ment at various places of the valley. Both of these niques such as logistic regression model (Chauhan
events have damaged various civil infrastructures et al. 2010), frequency-ratio model (Khan et al.
and army locations in the region. Despite threat- 2019), fuzzy logic (Kayastha et al. 2013), etc.,
ening debris flow events in the region, the sus- depend on probability of sliding by studying the
ceptibility maps are not available for most of the relationship between causative factors and past
area, which can be used for hazard assessment debris flow events. Qualitative techniques such as
and mitigation. analytical hierarchical process (AHP) and weigh-
Different studies suggest that debris flows ted overlay analysis (WOA) are mostly employed
susceptibility depends on several causative factors, for regional scale assessment (Kayastha et al.
which include environmental, hydrology, topogra- 2013). These techniques consider ranking and
phy, geology, etc. (Ventra and Clarke 2018; Crosta weighting of various causative factors based on
and Frattini 2003). The causative factors are usu- expert opinion about an area (Saaty 1990). Quali-
ally selected based on the analysis of debris flow tative techniques have been extensively studied for
types and the characteristics of the study area hazard susceptibility mapping over the Himalayan
(Wang et al. 2019). These factors may be further region (Basharat et al. 2016; Shit et al. 2016;
categorized into various levels (primary, sec- Kanwal et al. 2017). However, most of these studies
ondary, etc.) as per their potentiality in triggering have utilized static predictors to generate hazard
debris flow events (Dou et al. 2015). Prime trig- susceptibility maps. In order to reduce the impact
gering factor, i.e., rainfall facilitates the downward of rainfall-induced hazards, there is a need to
movement of the solids (boulders, rocks, etc.) or develop linking the above qualitative GIS model
soils on steep slopes by increasing pore water output with operational weather forecasts (Liao
pressure, seepage force and reducing effective et al. 2010). To quantify the spatial and temporal
stress of the soils (Lu et al. 2012). Consequently, variability in rainfall at regional scales, Numerical
real-time tracking of rainfall observations is critical Weather Prediction (NWP) models have the added
for assessment and early monitoring of rainfall-in- advantage of weather forecasting and thus can help
duced hazards (Huffman et al. 2007; Gebremichael in debris flow and landslide hazard predictions
and Hossain 2010). Yet, many parts in high-risk (Ochoa et al. 2014). Weather Research and Fore-
area of Himalayas lack resources to maintain the casting (WRF), a numerical mesoscale model
J. Earth Syst. Sci. (2020)129:30 Page 3 of 20 30

(Mourre et al. 2016) has the potential to simulate monsoon years, the southwest summer monsoon
heavy rainfall events. Over Himalayan region, penetrate into this region and mobilize large scale
Kumar et al. (2012) simulated 2010 ‘Leh cloud sediments in the form of debris flow from unpro-
burst event’ using WRF model and observed that tected mountain slopes (Juyal 2010). The variation
changing the microphysics schemes inside the in slope alongside road (between 22° and 50°), make
model could result into better simulation. Later on, this region highly sensitive and vulnerable to
Thayyen et al. (2013) studied this cloud burst various mass movements.
event using atmospheric modelling and hydrologi-
cal analysis to re-establish rainfall estimates during
the event. 2.1 Leh–Nubra valley road connectivity
The present study thus focuses on susceptibility The main road access to Nubra Valley is through
assessment of debris flow zones in Ladakh–Nubra Khardung-La pass (figure 1b) which is also
region. We studied summer precipitation pattern amongst one of the highest motorable roads in the
over this region using ground observatories as well world. The road connectivity has a huge impact for
as gridded precipitation data from Climate locals residing in the Nubra valley and has helped
Research Unit (CRU) for a common period in bringing changes in socio-economic condition.
1997–2017. GIS based multi-criteria evaluation The road to Nubra via Khardung-La does not
(MCE) technique has been performed based on provide all weather connectivity as it gets closed
static as well as dynamic predict or to identify during heavy snowfall in winter months and thus
debris flow susceptible zones. WRF forecasts have the region is completely cut off for some period.
been incorporated into GIS model for two cases of Also various rainfall-induced debris movement,
heavy precipitation events. Modeled high, medium, landslides and rock falls during summer causes
low and very low risk susceptible zones were eval- frequent road blockage.
uated and validated during field visit and satellite
imageries.
3. Data used

2. Study area 3.1 Precipitation dataset

The study focuses on Ladakh–Nubra region of 3.1.1 In-situ rain gauge observations
Indian part of Karakoram Himalaya (KH), having
Snow and Avalanche Study Establishment
the geographical extension between latitude
(SASE), India has a snow-meteorological observa-
34.00°–35.25°N and longitude 76.75°–78.00°E
tory in this region at Sasoma. Rainfall data is being
(figure 1) and covers an area of *4895 km2. The
measured using accumulation rain gauge at this
valley originates from south-western part of Tibetan
observatory since year 2000. The rainfall data
Plateau and runs over major thrust zone formed
collected on daily scale was analysed for the sum-
between Karakoram and Zanskar ranges, thus
mer months (June–September). In addition, rain-
stretching through Kargil, Zanskar, Leh and Nubra.
fall information available from Indian Air Force
Physiography of the terrain is harnessed with snow-
(IAF) station at Thoise was also utilised.
covered peaks, glaciers and non-glaciated valleys.
Lying on the rain shadow area of Himalayas, the
Nubra region combines the condition of both arctic 3.1.2 Gridded precipitation from CRU
and desert climate (Chundawat and Rawat 1994;
Negi 1995). There is a wide diurnal and seasonal The most recent version of the CRU-TS ver. 4.02
fluctuation in temperature with 40°C in winter to precipitation data available from University of
+35°C in summer with relative humidity ranging East-Anglia has been utilized in this study (https://
between 6% and 24% (Raj 2013). Precipitation crudata.uea.ac.uk/cru/data/hrg/). The data is
occurs mainly under the influence of westerly dis- entirely based on gauge information and consist of
turbances in winter and due to induced south-west monthly gridded fields of precipitation at 0.5° grid
monsoon during summer in lower reaches of valley spacing over land areas. The CRU dataset is pop-
(Bhutiyani et al. 2010). Being in the rain shadow ular because of its relatively long history and fine
region, devoid of vegetation and full of sediments spatial resolution of 0.5°. These datasets are con-
due to glaciated valley, during the abnormal tinuously updated through support from World
30 Page 4 of 20 J. Earth Syst. Sci. (2020)129:30

35°0’0’’N
Ka
ra
ko
ra
m
ra
ng
e Tibetian
Plateau

La
da
kh
ra

34°30’0’’N
Za ng
ns e
ka
rr
an
ge
Th
e
Gr
ea
tH ( )
(b)
im
al 77°30’0’’E
Pir ay
P an a
jal
ran
ge

(a)
Figure 1. (a) Location of the study area on Google Earth Imagery. (b) Ladakh–Nubra region shown on Landsat 8 imagery with
road connectivity from Leh to Nubra via Khardungla. Important places of this region are shown by red circles and star shows rain
gauge stations RG1 (Thoise) and RG2 (Sasoma).

Meteorological Organization (WMO) in collabora- Table 1. WRF model configurations used in the study.
tion with the US National Oceanographic and
Options WRF v3.9
Atmospheric Administration (NOAA) (Harris et al.
2014). The suitability of this dataset over North- Horizontal resolution 9 and 3 km
West Himalaya has recently been carried out by Microphysics WSM6
Long wave radiation RRTM
Negi and Kanda (2019). Thus, the dataset was
Shortwave radiation Dhudhia
extracted for Nubra region to study summer pre-
PBL physics YSU
cipitation climate variability. Land surface model Noah land-surface model
Cumulus parameterizations Kain–Fritsch
Vertical levels 40
3.1.3 WRF simulation
Model static fields USGS
The high resolution non-hydrostatic Advanced IC and BC NCEP GFS (0.25° 9 0.25°)
Research WRF (ARW; Skamarock et al. 2008)
mesoscale model (version 3.9) developed at National 3.2 Causes and associated factors
Centre for Atmospheric Research (NCAR) is used to
simulate the heavy rainfall cases. In the present Different causative factors for debris flow movement
study, double two-way nested domains with hori- can be divided into three groups consisting topo-
zontal resolution 09 and 03 km, keeping study area graphical, environmental and hydrological (figure 2).
and its adjoining areas as the center has been used. The final susceptibility map involves the generation
The configuration used to run the model is depicted of thematic maps of various causative factors.
in table 1. The initial and lateral boundary condi-
tions were obtained from final reanalysis data (FNL) 3.2.1 Topographical
at 1° 9 1° of National Center for Environmental
Prediction (NCEP) Global Forecast System (GFS; Topographical factors include slope, aspect and
https://doi.org/10.5065/D6M043C6). curvature derived using Advanced Space-borne
J. Earth Syst. Sci. (2020)129:30 Page 5 of 20 30
TERRAIN PARAMETERS

ENVIRONMENTAL
PARAMETERS

MEAN RAINFALL (1997-2017)

HYDROLOGICAL
PARAMETERS

Figure 2. Thematic layers of terrain, environmental and hydrological parameters considered for mapping debris flow zones.
Square black box in the figures represents the zoomed portion.

Thermal Emission and Reflection Radiometer north. Nine classes of aspect were prepared
(ASTER) digital elevation model (DEM). One of depending on the direction of slope. Over lower
the important factor, i.e., slope gradient was reaches of Nubra valley, induced monsoon rainfall
derived in degrees ranging from 0° to [55°. Aspect has more impact on south directing hill slopes
determines the direction of terrain slope w.r.t. compared to other directions which makes hill
30 Page 6 of 20 J. Earth Syst. Sci. (2020)129:30

slopes facing south-east or south-west more prone International Soil Reference Information Centre
to debris flows (Riaz et al. 2018). Curvature of (ISRIC) (https://www.isric.org/explore/soilgrids).
slope is responsible for controlling the flow of water The Soil Grids at 250 m spatial resolution are
during rainfall. It is one of the prime contributing outputs of a system for automated global soil
factors for slope failure during shallow landsliding mapping based on state-of-the-art spatial predic-
event. During analysis, curvature was classified as tion methods (Hengl et al. 2017). It provides global
concavity and convexity based on planar profiles. predictions for standard soil properties such as
Concave slopes with high negative values tend to organic carbon, bulk density, cation exchange
retain water for a longer period of time after rain- capacity (CEC), pH, soil texture fractions and
fall and this causes loosening of top layer soil coarse fragments at seven standard depths along
(Vijith et al. 2013). with predictions of depth to bedrock and distribu-
tion of soil classes based on World Reference Base
3.2.2 Environmental (WRB) and United States Department of Agri-
culture (USDA) classification systems. For the
Land cover and soil classes were considered under study region, soil type consists of five main classes:
environmental factors. Landsat-8 surface reflec- Luvisols, Leptosols, Regosols, Cambisols, and
tance imagery acquired on 30th August, 2015 with Cryosols. Properties of these soil classes were col-
cloud coverage less than 10% was downloaded from lected from various literatures and have been dis-
USGS Earth explorer (https://earthexplorer.usgs. cussed in table 2, which were also validated during
gov). The freely available Landsat-8 surface field survey in the study region. From figure 3, we
reflectance (SR) products are atmospherically and can see the soil class Leptosols and Regosols along
radiometrically corrected using Landsat Surface road connectivity to Sasoma. Compared to these
Reflectance Code (LaSRC) (Dwyer et al. 2018). two classes, other classes such as Luvisols, Cryosols
Employing unsupervised classification technique, and Cambisols are less susceptible to debris
major land cover zones were mapped in movement.
Ladakh–Nubra region and following categories
were generated: barren land, alluvial fan out area,
clean ice, glacier debris zone, vegetation and river. 3.2.3 Hydrological
Debris flow is an outcome of large quantities of ice-
Intensive as well as short duration continuous rain
cored glacial sediments which gets easily activated
has caused various debris flow events in Ladakh
from heavy precipitation, snowmelt and glacial
region during recent past. Therefore, long-term
lake outbursts (Haeberli et al. 1997). Since it
and short-term period thematic summer precipi-
originates mainly from the lower ablation region of
tation information was produced from CRU and
glaciers, debris zone of glaciers was considered as
WRF model respectively.
separate land cover class. In addition, alluvial fan
Stream networks are considered another
areas, which are triangle-shaped deposit of gravel,
important factor during debris flow events as they
sand/silt, and sediment are usually created by
contribute towards erosion of the slope and also
flowing water interaction with mountains, are
saturate the underwater section of the materials
considered as another class. Road connectivity to
(Miller and Burnett 2008). The stream network of
different places mainly passes through these fans
the study area was computed using ASTER DEM.
out areas which frequently get affected during
Four buffer zones with 500 m intervals were pro-
summer rainfall period (figure 3). Other classes
duced from the stream networks to see the impact
such as barren land, vegetation and water bodies
of distance. It was observed that the stream order
are well defined land cover classes (Congalton et al.
third and fourth contributes maximum to erosion
2014). Slope variation over barren land was found
on roadside slopes ranging 30°–45°.
varying between 36° and 55°, which makes it sus-
ceptible to various landsliding events as it cannot
hold the soil firmly. Vegetation areas are consid- 3.3 Field survey and validation
ered under less susceptible class as it reduces the
impact of landslide events by providing necessary Field survey was carried out in the study region
stability through their roots. during September 2015 (post-debris flow events)
Soil cover information was taken from global soil and locations were captured using Trimble Juno
information system called ‘‘Soil Grids’’ released by GPS. Further, Optical Land Imager (OLI) sensor
J. Earth Syst. Sci. (2020)129:30 Page 7 of 20 30

(a)
Screes accompanied by fracture valley

Palaeo river terraces

Debris covered valley

Ladakh hill slopes

Debris filled valley

River terrace River terrace

Talus fan

River channel Sandbars

(b)
Alluvial fan Flood plain

Cone of Talus

Cone of Talus
Mudflow along debris covered valley

(c)

Boulders deposited during peak flow in summers

(d)

Figure 3. (a) Debris field valley comprising mostly rock pieces mixed with clayey and sandy soil. (b) Older, well compacted,
naturally formed alluvial filled fan out area (Talus fan). (c) Mud flow in the lower part of the debris covered valley. (d) Boulders
deposition during period of peak discharge.

data from Landsat 8 was also utilized to observe study, the summer precipitation information
pre- and post-debris flow changes. available from two stations was analysed. Further,
gridded precipitation data from CRU was also
studied for the summer periods. The NetCDF data
4. Methods format of CRU was converted into Geo-TIFF for-
mat before analysis. The trend for cumulative
4.1 Precipitation analysis summer precipitation from CRU data in the study
region was analysed for the similar period
To assess the influence of summer precipitation on 1997–2017.
debris flows, it is important to study climatic pat- The early-warning for debris flow events can be
tern over the region. Field observational study on generated based on the temporal aspects of slope
precipitation for Nubra region is rather limited failures (Liao et al. 2010). Thus, an effort has been
(Chevuturi et al. 2015). Therefore in the present made in the present study to assess the suitability of
30 Page 8 of 20 J. Earth Syst. Sci. (2020)129:30

Table 2. ISRIC soil classes for Nubra region along with their properties.

Soil class Properties

Luvisols Porous and well aerated topsoil
High available water content
Water stagnation increases over the time period
When on slopes, prone to top soil erosion
Leptosols Are extremely gravelly ([80% gravel, stones, boulders by
volume)
Low water holding capacity due to limited depth and extreme
coarse structure
On steep slopes, prone to erosion
Regosols Undeveloped, medium textured soil
Consists of alluvial materials
Structureless due to lack of cohesion between soil particles
Often prone to erosion
Cambisols Good structural stability
Not much prone to erosion
Good internal drainage
Cryosols Occurrence of variable amount of ice in the sub soil
Thawing phenomenon lead to irregular land surface
Soils with evidence of freezing or thawing

WRF simulations for rainfall triggered debris flow the observations through indicators such as mean
during heavy rainfall events. The following two cases absolute error (MAE) and root mean square error
were taken to simulate heavy rainfall events: (RMSE).
(a) In 2015, this region experienced destructive
storm and continuous rainfall in first week of 4.2 Mapping of debris flow susceptible zones
August and was subsequently hit by series of
debris flows and flash floods. Amount of the For mapping susceptible zones, multi-criteria
daily rainfall intensity was simulated from evaluation (MCE) method was used. Weighted
WRF and compared with field data. WRF overlay analysis (WOA) was implemented in Arc-
forecasted 5-day cumulative precipitation GIS environment as it is a simple and reliable
layer was generated for aforementioned heavy technique (Riad et al. 2011). This technique
rainfall event and later incorporated into GIS involves overlaying of several raster files (causative
model. factors) to produce single output by assigning
(b) In 2010, intense rainfall occurred in various weight of each raster according to their importance
places of Ladakh during the period 1–5th (Saaty 1980). Overall sum of the weight of causa-
August which triggered deadly flash floods tive factors must be equal to 100 after comparison.
and debris flows. Destruction was concen- The thematic layers of various topographical,
trated in areas located proximal to the moun- environmental and hydrological factors were
tain slopes and dried channels (Juyal 2010). reclassified and rating value between 1 and 9 (1
The event caused at least 234 deaths in the being least susceptible and 9 most susceptible to
Ladakh region (Ziegler et al. 2016). Another landslides) was assigned to each class based on
800 were reported missing post-landsliding their potential to trigger debris flows (Calligaris
event as per army hospital records (Gupta et al. 2013). Pair-wise comparison matrix was cre-
et al. 2012). ated to find weight values and their results are
shown in figure 4. Weights were calculated for each
The sensitivity of various WRF configurations in causative factors after inter- and intra-comparison
simulating heavy rainfall events is assessed using (table 3). Assigning the weights and rank values
the Noah land surface model (LSM) (table 1) for each classes were based on expert opinions, field
(Tewari et al. 2004). The ability of WRF model observations and previous published literatures
configuration to simulate heavy rainfall event was (Lepore et al. 2011; Calligaris et al. 2013; Long and
evaluated by comparing the simulated rainfall with De Smedt 2019).
 (a)
RAINFALL High (>200 mm) Medium (100-200 mm) Low (50-100 mm) Very Low (<50 mm)
DRAINAGE BUFFER 500 1000 1500 2000
High (>200 mm) 1.00 3.00 5.00 7.00
Medium (100-200 mm) 0.33 1.00 3.00 5.00 500 1.00 3.00 5.00 7.00
Low (50-100 mm) 0.20 0.33 1.00 3.00 1000 0.33 1.00 3.00 5.00
Very Low (<50 mm) 0.14 0.20 0.33 1.00 1500 0.20 0.33 1.00 3.00
2000 0.14 0.20 0.33 1.00
SLOPE 30-45 15-30 45-55 >55 0-15
30-45 1.00 4.00 6.00 7.00 9.00
CURVATURE <-4 -4 to -1 >3 1 to 3 -1 to 0 0 to 1
15-30 0.25 1.00 3.00 5.00 7.00
J. Earth Syst. Sci. (2020)129:30

<-4 1.00 3.00 5.00 7.00 8.00 9.00

45-55 0.17 0.33 1.00 4.00 5.00
-4 to -1 0.33 1.00 3.00 4.00 5.00 7.00
>55 0.14 0.20 0.25 1.00 3.00
0-15 0.11 0.14 0.20 0.33 1.00
>3 0.20 0.33 1.00 4.00 4.00 6.00
1 to 3 0.14 0.25 0.25 1.00 3.00 5.00
SOIL CLASS Regosols Leptosols Luvisols Cambisols Cryosols -1 to 0 0.13 0.20 0.25 0.33 1.00 3.00
Regosols 1.00 3.00 4.00 6.00 8.00
0 to 1 0.11 0.14 0.17 0.20 0.33 1.00
Leptosols 0.33 1.00 3.00 5.00 7.00
Luvisols 0.25 0.33 1.00 5.00 7.00 South- North- South- North-
Cambisols 0.17 0.20 0.20 1.00 4.00 ASPECT West West South West East East North East Flat
Cryosols 0.17 0.14 0.14 0.25 1.00 South-West 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00
West 0.50 1.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00
Alluvial fan out Glacier Debris Barren Clean Vegetation River South 0.33 0.33 1.00 3.00 4.00 5.00 6.00 7.00 8.00
LAND COVER area Zone Land Ice
North-West 0.25 0.25 0.33 1.00 2.00 4.00 5.00 6.00 8.00
Alluvial fan out area 1.00 3.00 5.00 7.00 8.00 9.00
South-East 0.20 0.20 0.25 0.50 1.00 4.00 5.00 6.00 8.00
Glacier Debris Zone 0.33 1.00 3.00 5.00 6.00 9.00
East 0.17 0.17 0.20 0.25 0.25 1.00 3.00 5.00 7.00
Barren Land 0.20 0.33 1.00 5.00 5.00 7.00
Clean Ice 0.14 0.20 0.20 1.00 5.00 7.00
North 0.14 0.14 0.17 0.20 0.20 0.33 1.00 4.00 7.00
Vegetatation 0.13 0.17 0.20 0.20 1.00 5.00 North-East 0.13 0.13 0.14 0.17 0.17 0.20 0.25 1.00 7.00
River 0.11 0.11 0.14 0.14 0.20 1.00 Flat 0.11 0.11 0.13 0.13 0.13 0.14 0.14 0.14 1.00

(b)
Rainfall Land cover Soil Slope Drainage Curvature Aspect
Rainfall 1.00 3.00 3.00 4.00 5.00 7.00 7.00
Land cover 0.33 1.00 3.00 4.00 7.00 8.00 8.00
Soil 0.33 0.33 1.00 3.00 6.00 7.00 8.00
Slope 0.25 0.25 0.33 1.00 3.00 5.00 6.00
Drainage 0.20 0.14 0.17 0.33 1.00 4.00 2.00
Curvature 0.14 0.13 0.14 0.20 0.25 1.00 2.00
Aspect 0.14 0.13 0.13 0.17 0.50 0.50 1.00

Figure 4. Pair-wise comparison matrix for (a) each causative factors and (b) thematic layers.
Page 9 of 20 30
30 Page 10 of 20 J. Earth Syst. Sci. (2020)129:30

Table 3. Weights of main criteria and degree of preference for reclassified sub-criteria.

Main criteria Subcriteria

Group Criteria (Sij) Weight (Wi) Reclassification Weight (%) Importance
Topographical Slope 20% 0–15 3.3 1
15–30 23.4 6
30–45 52.6 9
45–55 14.3 4
[55 6.4 2
Aspect 2% Flat 1.0 1
North 4.3 2
North-east 4.0 2
East 6.1 3
South-east 9.7 5
South 16.2 7
South-west 22.3 9
West 20.4 8
North-west 16.0 6
Curvature 5% \4 45.2 9
4 to 1 32.2 7
1 to 0 2.1 3
0 to 1 1.3 2
1 to 3 7.5 4
[3 11.7 5
Environmental Land cover 15% Vegetation 3.6 2
River 1.3 1
Glacier debris 27.4 8
Clean ice 12.4 4
Alluvial fan 33.0 9
Barren land 22.3 7
ISRIC soil classification 23% Luvisols 17.4 6
Leptosols 32.0 8
Regosols 40.3 9
Cambisols 6.8 4
Cryosols 3.5 1
Hydrological Rainfall 28% High 55.8 9
Medium 26.3 6
Low 12.2 3
Very low 5.7 1
Drainage buffer 7% Unclassified 1.2 1
500 m 43.5 9
1000 m 30.1 7
1500 m 19.6 5
2000 m 5.6 2

Having determined the average summer weighted factor (28%). It was then followed by the
precipitation for the region during 1997–2017, the soil class (23%) and slope (20%) since after heavy
amount was further divided into four ranges. The precipitation event mostly rock pieces mixed with
area receiving highest amount of summer precipi- clayey and sandy soils were observed over fan out
tation ([175 mm) was likely to be more susceptible area. Maximum coverage was observed by the soil
towards debris flow hazard. Earlier Chevuturi et al. class Regosols (medium textured, eroded soil con-
(2015) also reported that Leh has the maximum sisting alluvial materials) and Leptosols ([80%
precipitation during the monsoon period (JAS) gravels, stones, and boulders by volume). Both
which also comprises extreme precipitating events. these soil classes often get eroded during summer
Thus, rainfall was observed as the heaviest either due to snow/glacier melting or period of
J. Earth Syst. Sci. (2020)129:30 Page 11 of 20 30

heavy rainfall. Slope gradient influences soil water

content, formation and potential for the erosion.
Generally, slopes with high gradient are expected
to have more influence on debris movements. On
the gentle slopes where shear stress is also low,
debris movements are not expected (Basharat et al.
2016). Curvature and aspect were seen amongst
the least weighted factor. Once the above compu-
tation was achieved, final landslide susceptibility
index map was produced using the following
equation (equation 1).
P
Wi Sij
WOA ¼ P ð1Þ
Wi
where Wi is the raster weight of importance given
to each factor, Sij is the reclassified raster in com-
mon criteria scale and WOA represents landslide
susceptibility index map.

4.3 Precipitation threshold estimation for debris

flow event

Precipitation thresholds for a region may indicate

the value which when reached is likely to trigger
debris flow events. These thresholds can be estab-
lished using historical data on landslide causing
rainfall events and corresponding landslide records.
Figure 5. Rainfall trend during summer-monsoon period
The precipitation threshold value can be estimated
(JJAS) from 1997 onwards. (a) Rainfall data collected at
using following categorized parameters: Thoise (RG1) and Sasoma (RG2); and (b) CRU gridded data
(i) intensity and duration (I–D) averaged for Ladakh–Nubra region.
(ii) total rainfall
(iii) rainfall event-duration and rainfall amount is observed at both stations post
(iv) rainfall event-intensity (Mathew et al. 2014). year 2010. The maximum rainfall amount is found
during the years 2014 and 2015, which reached up to
Our approach to estimate threshold value during
heavy rainfall event was based on 5-day total rainfall 90 mm/season at Sasoma and 45 mm/season at
Thoise. These identified heavy precipitation years
information estimated using WRF and ground sta-
may be considered exceptional as there is a general
tion. Cumulative rainfall (mm/day) was calculated
opinion that monsoon flow does not reach to upper
till event occurrence day from WRF and station data.
part of Ladakh region (Bhan et al. 2015). Rising
trend (+3.6 mm/season) in summer precipitation is
5. Results and discussion also observed from CRU dataset for the same period
with some bias with respect to field observation
5.1 Climatological analysis of rainfall (figure 5b). Again, the maximum summer precipi-
tation is observed for the year 2015 which led num-
The climatological analysis of observed summer ber of rainfall-induced debris flow events in this year.
rainfall data showed rise in monsoonal rainfall dur- The weather activity of Ladakh is affected mainly
ing study period (figure 5a). The average annual by western disturbances and less by monsoonal
rainfall is found to be 28 and 19 mm/season at rainfall. As upper part of Ladakh comes under a very
Sasoma and Thoise station respectively. For both low rainfall zone, it is generally believed that the
the stations increasing rainfall amount is observed monsoon current does not reach this region. However,
with slope values +2.10 mm/season at Sasoma and our results show the presence of induced monsoonal
+0.45 mm/season at Thoise. Sudden increase in rainfall with rising trend over valley region of Ladakh
30 Page 12 of 20 J. Earth Syst. Sci. (2020)129:30

and Nubra. Houze et al. (2007) and Medina et al.

(2010) have examined the rainfall over north-west
Himalaya and reported that the region is prone to
deep convection during the period of monsoon. Later
on, Ray et al. (2001) also reported that the interaction
of the monsoon trough with western disturbances
may occasionally cause dense clouding and heavy
precipitation in the months of July and August, which
might be the reason for unusual heavy rainfall in 2015.
Bhan et al. (2015) analysed the Leh cloud burst event
in 2010 and discussed the increasing impact of south-
west monsoon in this region. More importantly, these
studies also highlighted the orography factor in cre-
ating heavy rainfall events over a small and compact
region surrounded by mountains. Thus, increased
precipitation amount in this region has led to hazard
vulnerability, mainly in terms of transportation of
debris along with the rain water running down-
stream. This has been further discussed in succeeding
sections.

5.2 Debris flow susceptible zones

A weighted overlay hazard rating methodology was Figure 6. Map showing debris flow susceptible zones derived
employed integrating remote sensing, GIS and using weighted overlay technique.
NWP technique. All raster datasets of terrain
parameters such as slope, aspect, curvature, etc.,
were resampled using bilinear interpolation tech- Table 4. Statistics of debris flow susceptibility zones based on
nique. Nearest neighbour technique was employed Phase-1.
for categorical data such as soil and land cover Class Area (km2) Area coverage (%)
raster before WOA. As dynamic rainfall intensity
Very Low 2845 41.22
was considered as a triggering parameter for vari-
Low 2087 30.18
ous debris flow events in the region, the model was
Medium 1322 19.14
run in following two phases: High 653 9.45

5.2.1 Phase 1: Susceptible zones identified based

on mean summer precipitation
while about 19.14% is classified under medium risk
In first phase, mean summer precipitation for the hazard zone (table 4). Most of the slope failures
period 1997–2017 was provided in the model along over high vulnerability zones take place on the
with other causative factors. The susceptible map road and blockage to roadside drains and car-
of study area is classified into four hazard risk riageway. With total road length of *92 km from
categories: very low risk, low risk, medium risk and Diskit to Turtuk and *106 km from Diskit to
high risk. The road connectivity to various places Siachen glacier, the affected road length in hazard
and glaciers moves through vulnerable slopes map is found to be *45 km and *37 km respec-
where various debris movements takes place during tively, and thus more number of high risk zones is
summer rainfall, thus a risk map is prepared for found along Diskit and Turtuk.
non-glaciated terrain as shown in figure 6. Figure 6 The susceptible zones identified by Phase 1 are
depicts hotspots of the high debris flow susceptible overall hotspots in the region. Their class of vul-
zones from Leh to Siachen glacier and Leh to nerability may change depending upon the main
Turtuk (via Diskit). The high risk zones accounted causative factor, i.e., rainfall being dynamic in
for 9.45% of land areas in the susceptibility map, nature. For this, two different cases of heavy
J. Earth Syst. Sci. (2020)129:30 Page 13 of 20 30

90
80
OBS WRF

Acc. Precipitation
70
60

(mm/day)
50
40
30
20
10
0
1 2 3 4 5
Aug 1 to 5, 2015

Figure 7. Comparison of WRF 5-day precipitation forecast and gauge-based rainfall observations.

rainfall events simulated by WRF were incorporated heavy rainfall events in the region during 1–5
into GIS model as discussed in the next section. August, 2015 (Ziegler et al. 2016). It was also
observed from field data that during 19–23 July
5.2.2 Phase 2: Based on WRF simulation 2015, a small duration of precipitation maintained
for heavy rainfall events high soil moisture in this region (figure 9). The
WRF simulated precipitation showed more than
Second phase involves prototype for early predic- 200 mm in 5 days during event period at multiple
tion of debris flow zones based on two heavy rain- pockets of study region and thus induced several
fall events simulated using WRF model. One of the debris flow events.
critical points for proper early prediction of rainfall Figure 8(b) represents the scenario during Leh
induced debris flow events is an accurate prediction cloud burst event period (1–5 August, 2010) along
of temporal intensity of rainfall, for which we with the susceptibility map. This event was
compared WRF simulated precipitation with rain somehow different from above mentioned 2015
gauge observations. The comparison is carried out event as a diurnal heating of TP region triggered
for one of the heavy rainfall events in 2015 at convective initiation of environment of this event
Thoise. The model simulated 4-days precipitation towards the Leh region (Rasmussen and Houze
is found to be in good agreement with gauge 2012). Conversely, a monsoon trough also inter-
observation (r = 0.76) having RMSE and MAE of acted with MCS which formed intense convective
13.23 and 8.75 mm/day respectively (figure 7). group of clouds over Leh region (Kumar et al.
However, the complete 5-day accumulated precip- 2012; Thayyen et al. 2013). Our results showed
itation forecasts is found relatively in lesser that over the Leh region the WRF simulated
agreement (r = 0.50) compared to 4-day accumu- precipitation for 5-Day has accumulated above
lated forecasts. This may be due to insufficient 200 mm, which led high risk prone zones to Leh
parameterizations of NWP beyond 3–4 days fore- region.
casts over high mountainous regions (Norris et al. From figure 8(a–b), it can be observed that
2017). The WRF simulated quantitative precipi- different hotspots of debris flow were identified
tation was further incorporated into the GIS model during these events. While regions from Khalsar
to map probable hazard zones during heavy pre- to Sumur, Sumur to Sasoma, and Thoise to Tur-
cipitation events. tuk were identified as high risk zones during 2015
Figure 8(a) shows spatial distribution of 5-day event, these were found medium to low risk zones
accumulated precipitation by WRF as well as during 2010 cloud burst event. Similarly during
susceptibility map of debris flow zones during the 2010 event, apart from Leh, Khardung-La pass
period 1–5 August 2015. During this event, a and regions of sloped terrains near Diskit was
mesoscale convective system (MCS) was developed found highly susceptible to debris flow events. The
over the Tibetian Plateau (TP) along with the rain overall comparison of identified hotspots in both
storm fronts arriving towards the Nubra valley. In the above discussed cases is shown in table 5. It
the meantime, a westerly upper air advection also was found in good spatial agreement with field
contributed in sustaining this system which led to observations.
30 Page 14 of 20 J. Earth Syst. Sci. (2020)129:30

Figure 8. Two cases studies of debris flow events: (a) WRF simulation for period 1st to 5th August 2015 and corresponding
susceptibility debris flow zonation; (b) WRF simulation for period 1–5th August 2010 and corresponding susceptibility debris
flow zonation.

5.3 Model validation and the comparison was made. Besides, Landsat
image of period 10th May and 14th October, 2015
Field survey was conducted in the September was also analysed. For better detection of pre and
month in Nubra after the debris flow events of post changes, image pan sharpening technique was
2015. Field identified hazard locations were col- employed using OLI multispectral (30 m) and
lected using GPS. Later these hazard locations panchromatic band (15 m), and debris flow zones
were overlaid on modelled susceptible classes, i.e., were identified. It was observed that most of the
very low risk, low risk, medium risk and high risk identified field events occurred over the sloped
J. Earth Syst. Sci. (2020)129:30 Page 15 of 20 30
80 0.5
RAIN (mm) SOIL MOISTURE (m3/m3)
70
0.4

SOIL MOSITURE
60
50 0.3

RAIN
40
30 0.2

20
0.1
10
0 0

03-Aug-15

06-Aug-15

09-Aug-15
01-Jul-15

04-Jul-15

10-Jul-15
07-Jul-15

13-Jul-15

16-Jul-15

19-Jul-15

22-Jul-15

25-Jul-15

28-Jul-15

31-Jul-15
DATE

Figure 9. Observational data from Thoise station for July-Aug period. Short duration of rain was observed between 19th July
and 23rd July 2015 which maintained high soil moisture (source: SMAP L-Band radiometer data; https://www.mosdac.gov.in/
soil-moisture-0) in the region.

Table 5. An overall comparison of identified hotspots in two cases studied.

Class Area (km2) Area coverage (%) High risk zones

Case I (2015 event)
Very Low 2839.22 41.13 Khalsar–Sumur, Sumur–Sasoma, Thoise–Turtuk
Low 2407.78 34.88
Medium 1212.17 17.56
High 443.86 6.43
Case II (2010 event)
Very Low 3152.61 45.67 Leh, Khardung-La, Nearby Diskit region
Low 2790.89 40.43
Medium 728.96 10.56
High 301.66 4.37

terrains alongside the road which subsequently the accuracy of the analysis under this predictive
hampered the connectivity over Ladakh–Nubra curve is the area under curve (AUC). The most
region. For instance, we can see from figure 10 that ideal model will have the largest area under the
the debris (mostly boulders and alluvial fanned curve. An AUC value close to 1.0 presents an ideal
materials) sliding have occurred near Diskit town model whereas AUC value close to 0.5 indicates
and regions between Sasoma and Siachen glacier inaccuracy of the model. The ROC was thus pro-
after heavy rainfall events of 2015. These are duced by plotting landslide occurrences (TPR) at
mainly the habitat areas. The observed debris flow y-axis and landslide susceptibility index (FPR).
locations were compared with WRF based hazard With limited field identified landslide evidences,
risk zones for the 2015 event (figure 10a). Landsat-8 the collected points were compared with 4 classes
imageries were also utilized to infer changes caused of susceptibility map and the curve generated
due to debris flow at Thoise. The post event image provides rate of success in the model. The com-
clearly showed changes like disconnection of roads puted area under the curve is 0.76, which showed
and flow of debris materials. an accuracy of 76.6% for the final susceptibility
After visual based validation, the model map (figure 10b).
performance was checked for 2015 events using
statistical predictive Receiver Operating Charac-
teristics (ROC) curve. ROC is widely popular 5.4 Critical rainfall thresholds
technique for evaluating the accuracy and effec-
tiveness of the susceptibility model (Lee 2005). An analysis on the critical threshold value to induce
ROC curve shows model performance based on the debris flows was carried out for the study region
true positive rate (TPR) and false positive rate using ground rainfall observations and WRF esti-
(FPR) (Pradhan 2010). An important measure of mates. This was evaluated over Thoise for 2015
30 Page 16 of 20 J. Earth Syst. Sci. (2020)129:30

(a)

PRE POST

ROC Curve
(b) 1

0.8
Landslide Occurrences

0.6

0.4

0.2

AUC: 76.6%
0
0 0.2 0.4 0.6 0.8 1
Landslide susceptibility

Figure 10. (a) Validation of retrieved susceptibility zones of 2015 debris flow events in Nubra valley using field survey and
satellite data and (b) Assessment of susceptibility map performance for 2015 events based on ROC (normalized over both the
axes).

event. Cumulative precipitation (mm/day) was model simulation was able to capture the quantitative
calculated from WRF and ground station till the accumulated precipitation (225 mm) in 4 days
event occurrence day, i.e., 4th August 2015 (figure 11). and thus was found in agreement with the ground
The cumulative precipitation for 4 days based on observation. Kirschbaum et al. (2012) studied and
ground observation was 204 mm, which on daily tested various rainfall triggered debris flow events
basis estimated threshold value as 51 mm. The based on TRMM estimates over the Himalayan arc
J. Earth Syst. Sci. (2020)129:30 Page 17 of 20 30
DAILY PRECIPITATION AMOUNT temporal WRF information with other causative
CUMULATIVE PRECIPITATION (STATION)
factors. Two cases of heavy precipitation event
CUMULATIVE PRECIPITATION (WRF)
occurred in the year 2010 and 2015 were modeled
350 Debris flow event 80
using WRF for assessing debris flow susceptible
zones. The investigated area was again divided into

Daily Precipitaon amount(mm)

300
Cumulave Precipitaon (mm)

60 four classes of susceptibility which showed that the

250
225 mm debris zone has been exposed on alluvial fans of sev-
200 204 mm
eral mountain streams in the area.
40 The approach used in this study is simple and
150
effective, provides quick results and can be ratio-
100 nalised depending on various factors. This would
20
help in disaster preparedness for the region since this
50 region is facing climate change impacts since last
0 0 decades. Satellite based estimates based on TRMM
have already reported increased heavy rainfall and
29-Aug-15

30-Aug-15

31-Aug-15

01-Sep-15

02-Sep-15

03-Sep-15

04-Sep-15

05-Sep-15

06-Sep-15

07-Sep-15

08-Sep-15
extreme events over Leh–Ladakh region (Bharti
et al. 2016). Our study based on field observation
Date
supports these on-going changes in rainfall pattern.
Figure 11. Daily and cumulative rainfall from gauge station The reappearance of heavy rainfall causing flash
and WRF during heavy rainfall period at Thoise. floods and debris flows in the summer-monsoon
suggests upsetting situation in this high mountain
desert. As per the future climate change scenario the
and reported a threshold amount of 79 mm/day for most of the study suggests there will be an increase in
potential triggering. This showed the threshold precipitation and temperature over the region
value for Nubra region is less compared to other (Chaturvedi et al. 2012). Thus apart from increasing
region of Himalaya, which is evident as the Nubra precipitation, increase in temperature over these
region has glacier deposits, loose landforms and permafrost regions due to global warming will
devoid of vegetation. enhance the probability of larger mass movement
events (Huggel 2009). Though the studied region is
scanty populated, but because of strategic reason it
6. Conclusions is generally fully occupied. Due to increasing
requirements of locals for different resources such as
The intense and prolonged rainfall induced debris water, agriculture land, transportation etc., the area
flow events can cause serious damages to people and mostly remains habitat along the river bed or drai-
social economic progress. To avoid this in nage (alluvial areas). Therefore, there is an urgent
Ladakh–Nubra region, there is a need of better need to safeguard the people and infrastructure in
understanding of the increased vulnerability from this region from increasing extreme events in climate
debris flows which gets trigger during prolonged change scenarios. This can be done by either passive
heavy rainfall; such events were in years 2010 and method i.e. issuing forecast or by active methods,
2015. Analysis based on field information of rainfall i.e., control methods. For high risk zones as identi-
intensity for summer-monsoonal period indicated ed from the present study (table 5), control struc-
significant rise during the last 20 years, which has ture measures, such as, diversion/protection wall or
caused upsurge in several water induced hazards over water channel, along the eroded nallahs can be
this region. The debris flow susceptibility map was constructed to divert or channelized the water/de-
produced using WOA method. Seven important bris flow. Therefore, the present study may be useful
causative factors of debris flow occurrence were con- for the decision/policy makers as well as locals to
sidered for construction of thematic data layers using minimise the rainfall induces hazards in the region.
GIS. The susceptibility map classified the study area
into very low risk, low risk, medium risk and high risk.
The map revealed that *28% area falls in high and Acknowledgements
medium risk susceptibility zones. The present study
also attempts to identify debris flow susceptible zones The authors are thankful to Director SASE for
in Ladakh–Nubra region by integrating high spatio- motivation and support to this work. We thank the
30 Page 18 of 20 J. Earth Syst. Sci. (2020)129:30
technical persons of SASE Roshan Tamang, G Chundawat R S and Rawat G S 1994 In: Proceedings of the
Arun, Ritesh Mujawdiya, and Bhupinder Kumar Seventh International Snow Leopard Symposium; Interna-
for field data collection. We are grateful to tional Snow Leopard Trust, Seattle, pp. 127–132.
Congalton R G, Gu J, Yadav K, Thenkabail P and Ozdogan M
Air Force Station, Thoise for providing the 2014 Global land cover mapping: A review and uncertainty
meteorological data. We would like to acknowledge analysis; Remote Sens. 6 12,070–12,093.
various institutions for providing gridded CRU Crosta G B and Frattini P 2003 Distributed modelling of
precipitation products, ASTER DEM and shallow landslides triggered by intense rainfall; Nat.
Landsat-8 data. This work was carried out under Hazards Earth Syst. Sci. 3 81–93.
Dou J, Bui D T, Yunus A P, Jia K, Song X, Revhaug I, Xia H
DRDO project ‘Him-Parivartan’.
and Zhu Z 2015 Optimization of Causative Factors for
Landslide Susceptibility Evaluation Using Remote Sensing
References and GIS Data in Parts of Niigata, Japan; PLOS ONE 10(7)
e0133262.
Dwyer J, Roy D, Sauer B, Jenkerson C, Zhang H and
Adhikari D P and Koshimizu S 2005 Debris flow disaster at
Lymburner L 2018 Analysis ready data: Enabling analysis
Larcha, upper Bhotekoshi Valley, central Nepal; Island
of the Landsat archive; Remote Sens. 10(9) 1363.
Arc. 14 410–423.
Elkadiri R, Sultan M, Youssef A M, Elbayoumi T, Chase R,
Barrett K and Bosak K 2018 The role of place in adapting to
Bulkhi A B and Al-Katheeri M M 2014 A remote sensing-
climate change: A case study from Ladakh, western
based approach for debris-flow susceptibility assessment
Himalayas; Sustainability 10 898.
using artificial neural networks and logistic regression
Basharat M, Shah H R and Hameed N 2016 Landslide
modeling; IEEE J. Sel. Top. Appl. Earth Observ Rem.
susceptibility mapping using GIS and weighted overlay
Sens. 7(12) 4818–4835.
method: A case study from NW Himalayas, Pakistan; Arab.
J. Geosci. 9 292. Gebremichael M and Hossain F 2010 Satellite rainfall appli-
Bhan S C, Devrani A K and Sinha V 2015 An analysis of cations for surface hydrology; Dordrecht: Springer Nether-
monthly rainfall and the meteorological conditions associ- lands 3 22.
ated with cloudburst over the dry region of Leh (Ladakh), Gupta P, Khanna A and Majumdar S 2012 Disaster manage-
India; Mausam 66(1) 107–122. ment in flash floods in Leh (Ladakh): A case study; Indian
Bharti V, Singh C, Ettema J and Turkington T A R 2016 J. Comm. Med. 37(3) 185–190.
Spatiotemporal characteristics of extreme rainfall events Haeberli W, Wegmann M and MUhll D V 1997 Slope stability
over the Northwest Himalaya using satellite data; Int. problems related to glacier shrinkage and permafrost
J. Climatol. 36 3949–3962. degradation in the Alps; Eclogae Geol. Helv. 90 407–414.
Bhutiyani M R, Kale V S and Pawar N J 2010 Climate change Harris I, Jones P D, Osborn T J and Lister D H 2014 Updated
and the precipitation variations in the north-western high-resolution grids of monthly climatic observations – the
Himalaya: 1866–2006; Int. J. Climatol. 30(4) 535–548. CRU TS3.10 Dataset; Int. J. Climatol. 34 623–642.
Biswas S S and Pal R 2016 Causes of landslides in Darjeeling Hengl T, Mendes J J, Heuvelink G B M, Ruiperez G M and
Himalayas during June–July, 2015; J. Geogr. Nat. Disast. 6 Kilibarda M et al. 2017 Soil Grids 250 m: Global gridded
173. soil information based on machine learning; PLOS ONE
Blais-Stevens A and Behnia P 2016 Debris flow susceptibility 12(2) e0169748.
mapping using a qualitative heuristic method and flow-R Hobley D E J, Sinclair H D and Mudd S M 2012 Reconstruc-
along the Yukon Alaska Highway Corridor, Canada; Nat. tion of a major storm event from its geomorphic signature:
Hazards Earth Syst. Sci. 16 449–462. The Ladakh floods, 6 August 2010; Geology 40 483–486.
Bookhagen B and Burbank D W 2010 Toward a complete Houze R A Jr, Darren C W and Smull B F 2007 Monsoon
Himalayan hydrological budget: Spatiotemporal distribu- convection in the Himalayan region as seen by the TRMM
tion of snowmelt and rainfall and their impact on river Precipitation Radar; Quart. J. Roy. Meteorol. Soc. 133
discharge; J. Geophys. Res. 115 F03019. 1389–1411.
Calligaris C, Poretti G, Tariq S and Melis M T 2013 First steps Huffman G J, Adler R F, Bolvin D T, Gu G, Nelkin E J and
towards a landslide inventory map of the Central Karakoram Bowman K P 2007 The TRMM multi-satellite precipitation
National Park; Eur. J. Remote Sens. 46 272–287. analysis: Quasi-global, multi-year, combined-sensor precip-
Chaturvedi R K, Joshi J and Jayaraman M 2012 Multi- itation estimates at fine scale; J. Hydrometeorol. 8(1) 38–55.
model climate change projections for India under repre- Huggel C 2009 Recent extreme slope failures in glacial
sentative concentration pathways; Curr. Sci. 103(7) environments: Effects of thermal perturbation; Quat. Sci.
791–802. Rev. 28 1119–1130.
Chauhan S, Sharma M, Arora M K and Gupta N K 2010 Juyal N 2010 Cloud-burst triggered debris flows around Leh;
Landslide susceptibility zonation through ratings derived Curr. Sci. 99(9) 1166–1167.
from artificial neural network; Int. J. Appl. Earth Obs. 12 Kanwal S, Atif S and Shafiq M 2017 GIS based landslide
340–350. susceptibility mapping of northern areas of Pakistan, a case
Chevuturi A, Dimri A P, Das S, Kumar A and Niyogi D study of Shigar and Shyok Basins; Geomat. Nat. Haz. Risk
2015 Numerical simulation of an intense precipitation 8(2) 348–366.
event over Rudraprayag in the central Himalayas during Kayastha P, Bijukchhen S M, Dhital M R and De Smedt F
13–14 September 2012; J. Earth Syst. Sci. 124(7) 2013 GIS based landslide susceptibility mapping using a
1545–1561. fuzzy logic approach: A case study from Ghurmi-Dhad
J. Earth Syst. Sci. (2020)129:30 Page 19 of 20 30
Khola area, Eastern Nepal; J. Geol. Soc. India 82(3) Nikolopoulos E I, Crema S, Marchi L, Marra F, Guzzetti
249–261. F and Borga M 2014 Impact of uncertainty in rainfall
Khan H, Shafique M, Khan M A, Bacha M A, Shah S U and estimation on the identification of rainfall thresholds for
Calligaris C 2019 Landslide susceptibility assessment using debris flow occurrence; Geomorphology 221 286–297.
Frequency Ratio, a case study of northern Pakistan; Egypt. Norris J, Carvalho LMV and Co-authors 2017 The spatiotempo-
J. Remote Sens. Space Sci. 22 11–24. ral variability of precipitation over the Himalaya: Evaluation
Kirschbaum D and Stanley T 2018 Satellite-based assessment of one-year WRF model simulation; Clim. Dyn. 49 2179.
of rainfall-triggered landslide hazard for situational aware- Ochoa A, Pineda L, Crespo P and Willems P 2014 Evaluation
ness; Earth’s Future 6 505–523. of TRMM 3B42 precipitation estimates and WRF retro-
Kirschbaum D, Adler R, Adler D, Peters-Lidard C and spective precipitation simulation over the Pacific–Andean
Huffman G 2012 Global Distribution of Extreme Precipi- region of Ecuador and Peru; Hydrol. Earth Syst. Sci. 18
tation and High-Impact Landslides in 2010 Relative to 3179–3193.
Previous Years; J. Hydrometeor. 13 1536–1551. Owen L A, Kamp U, Khattak G A, Harp E, Keefer D K and
Kumar M S, Shekhar M S, Rama Krishna S S V S, Bhutiyani Bauer M 2008 Landslides triggered by the 8 October 2005
M R and Ganju A 2012 Numerical simulation of cloud burst Kashmir earthquake; Geomorphology 94(1–2) 1–9.
event on August 05, 2010, over Leh using WRF mesoscale Pardeshi S D, Autade S E and Pardeshi S S 2013 Landslide
model; Nat. Hazards 62(3) 1261–1271. hazard assessment: Recent trends and techniques; Spring-
Lee S 2005 Application of logistic regression model and its erPlus 2 523.
validation for landslide susceptibility mapping using GIS and Pradhan B 2010 Application of an advanced fuzzy logic model
remote sensing data; Int. J. Remote Sensing 26(7) 1477–1491. for landslide susceptibility analysis; Int. J. Comput. Int.
Lepore C, Kamal S A, Shanahan P and Bras R L 2011 Syst. 3(3) 370–381.
Rainfall-induced landslide susceptibility zonation of Puerto Pradhan S B 2019 July 16 Nepal floods Death toll touches 78
Rico; Environ. Earth Sci. 66(6) 1667–1681. over 17 500 displaced; The Wire, https://www.theweek.in/
Liao Z, Hong Y, Wang J, Fukuoka H, Sassa K, Karnawati D wire-updates/international/2019/07/16/fgn24-nepal-floods-
and Fathani F 2010 Prototyping an experimental early toll.html.
warning system for rainfall-induced landslides in Indonesia Raj A 2013 Is Ladakh a ‘cold desert’? Curr. Sci. 104(6) 687–688.
using satellite remote sensing and geospatial datasets; Rasmussen K L and Houze R A Jr 2012 A flash-flooding storm
Landslides 7(3) 317–324. at the steep edge of high terrain: Disaster in the Himalayas;
Long N T and De Smedt F 2019 Analysis and mapping of Bull. Am. Meteorol. Soc. 93 1713–1724.
rainfall-induced landslide susceptibility in a Luoi district, Ray T K, Sinha D and Guhathakurta P 2001 Landslides in
Thua Thien Hue province, Vietnam; Water 11(1) 51. India; India Meteorological Department Pune, pre-pub-
Lu N, Sener-Kaya B, Wayllace A and Godt J W 2012 Analysis lished scientific report No. 1/2001 20.
of rainfall induced slope instability using a field of local Riad PH, Billib M, Hassan A A, Salam M A and El Din
factor of safety; Water Resour. Res. 48(9) W09524. MN 2011 Application of the overlay weighted model and
Martha T R, Roy P, Govindharaj K B, Kumar V K, Diwakar Boolean logic to determine the best locations for artificial
P G and Dadhwal V K 2015 Landslides triggered by the recharge of groundwater; J. Urban Environ. Eng. 5(2) 57–66.
June 2013 extreme rainfall event in parts of Uttarakhand Riaz M T, Basharat M, Hameed, N, Shafique M and Luo J 2018
state, India; Landslides 12 135–136. A data-driven approach to landslide-susceptibility mapping in
Mathew J, Babu D G, Kundu S, Kumar V K and Pant C C mountainous terrain: Case study from the northwest Hima-
2014 Integrating intensity-duration-based rainfall thresh- layas, Pakistan; Nat. Hazards Rev. 19 05018007.
old and antecedent rainfall-based probability estimate Saaty TL 1980 The Analytical Hierarchy Process: Planning,
towards generating early warning for rainfall-induced Priority Setting, Resource Allocation; 1st edn, McGraw-Hill
landslides in parts of the Garhwal Himalaya, India; New York, ISBN: 0070543712.
Landslides 11(4) 575–588. Saaty T L 1990 An exposition of the AHP in reply to the paper
Medina S, Houze R A Jr, Kumar A and Niyogi D 2010 ‘‘remarks on the analytic hierarchy process; Manag. Sci.
Summer monsoon convection in the Himalayan region: 36(3) 259–268.
Terrain and land cover effects; Quart. J. Roy. Meteor. Soc. Shit P K, Bhunia G S and Maiti R 2016 Potential landslide
136 593–616. susceptibility mapping using weighted overlay model
Miller D J and Burnett K M 2008 A probabilistic model of (WOM); Model. Earth Syst. Environ. 2 21.
debris-flow delivery to stream channels, demonstrated for the Skamarock W C and Co-authors 2008 A description of the
Coast Range of Oregon, USA; Geomorphology 94 184–205. Advanced Research WRF version 3; NCAR Tech. Note
Mourre L, Condom T, Junquas C, Lebel T, Sicart J E, NCAR/TN-475 + STR 113.
Figueroa R and Cochachin A 2016 Spatio-temporal assess- Tewari M, Chen F, Wang W, Dudhia J, LeMone M A,
ment of WRF, TRMM and in situ precipitation data in Mitchell K, Ek M, Gayno G, Wegiel J and Cuenca R H 2004
a tropical mountain environment (Cordillera Blanca, - Implementation and verification of the unified NOAH land
Peru); Hydrol. Earth Syst. Sci. 20 125–141. surface model in the WRF model (Formerly Paper Number
Negi S S 1995 Cold deserts of India; Indus Publishing 17.5), In: 20th Conference on Weather Analysis and
Company, New Delhi. Forecasting/16th Conference on Numerical Weather Pre-
Negi H S and Kanda N 2019 An appraisal of spatio-temporal diction; Am. Meteor. Soc., pp. 11–15.
characteristics of temperature and precipitation using Thayyen R J, Dimri A P, Kumar P and Agnihotri G 2013
gridded datasets over NW Himalaya; Climate Change and Study of cloudburst and flash floods around Leh, India
the White World; Springer (in press). during August 4–6, 2010; Nat. Hazards 65 2175–2204.
30 Page 20 of 20 J. Earth Syst. Sci. (2020)129:30

Ventra D and Clarke L E (eds) 2018 Geology and Ghats, India; Georisk: Ass. Manag. Risk Eng. Syst.
Geomorphology of Alluvial and Fluvial Fans: Terrestrial Geohazards 8 48–62.
and Planetary Perspective; Geological Society, London Wang Y Q, Ma C B and Wang Z F 2019 Prediction of
440. Landslide Position of Loose Rock Mass at Mountain Tunnel
Vijith H, Krishnakumar K N, Pradeep G S, Ninu Krishnan Exit; Adv. Civ. Eng. 2019 3535606.
M V and Madhu G 2013 Shallow landslide initiation Ziegler A D, Cantarero S I, Wasson R J, Srivastava P,
susceptibility mapping by GIS-based weights-of-evi- Spalzin S, Chow W T L and Gillen J 2016 A clear and
dence analysis of multi-class spatial data-sets: A case present danger: Ladakh’s increasing vulnerability to flash
study from the natural sloping terrain of Western floods and debris flows; Hydrol. Process. 30 4214–4223.

Corresponding editor: NAVIN JUYAL