Case Study

Impact of Land Use/Land Cover Change on Future

Flash Flood Vulnerability and Monsoon Runoff
in an Ungauged Himalayan Watershed
Dibyandu Roy 1; Bharath H. Aithal 2; Anirban Dhar 3; and
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Venkappayya R. Desai 4

Abstract: Hydrological extremes and land use are intimately connected, as land use alterations frequently exacerbate or mitigate the
consequences of extreme events. Therefore, assessment and prediction of land use dynamics retain great significance in projected scenarios
of hydrological extremes. This study aimed to quantify the impact of land use changes on future landslide-prone sites, flash flood vulnerable
zones (FFVZs), and monsoon runoff. In this context, a Markov chain–cellular automata (MC-CA) model integrated with a multilayer per-
ceptron–neural network (MLP-NN) model was designed to predict near-future land use/land cover changes (LULCC) using high-resolution
existing land use/land cover (LULC) information. Uncorrected finer-scale Coupled Model Intercomparison Project Phase 6 (CMIP6)
meteorological data sets and predicted LULC data were used to simulate daily runoff with the help of the Soil and Water Assessment Tool
(SWAT). Statistical evaluation matrices were used to assess the efficacy of using uncorrected CMIP6 data sets for hydrological modeling. The
methodology was evaluated for suitability in the ungauged high-altitude Ranikhola watershed. The prediction results showed that the water-
shed can be expected to be more susceptible to landslides and flash floods (9%). The most challenging discovery from this analysis was an
anticipated 32.50% increase in average monsoon runoff from 2021 to 2032 compared to the observed period, 2015–2020. In addition, the
runoff peak magnitude is projected to rise by 26.00%. Consequently, it is imperative to develop strategies and approaches for sustainable
watershed management to address future challenges. DOI: 10.1061/JHYEFF.HEENG-6173. © 2024 American Society of Civil Engineers.
Author keywords: Land use/land cover (LULC) prediction; Markov chain–cellular automata–multilayer perceptron–neural network
(MC-CA-MLP-NN); Soil and Water Assessment Tool (SWAT); Centro Euro-Mediterraneo per I Cambiamenti Climatici Earth System Model
Version 2 Coupled Model Intercomparison Project phase 6 (CMCC-ESM2-CMIP6); Future flash flood vulnerable zone (FFVZ); Monsoon
runoff.

Introduction provision of essential services and disrupt hydrological systems

(Bewket 2002; Gogoi et al. 2019). Most of these watersheds are
In the modern era, unsustainable growth in urbanization and forest susceptible to glacial lake outburst floods (GLOFs), flash floods,
degradation are triggering severe hydrological extremes in cities avalanches, and landslides (Prasad et al. 2016; Roy et al. 2023; Veh
and hill stations in India (Abdulkareem et al. 2018; Bhattacharjee et al. 2019). Furthermore, land use/land cover change (LULCC)
and Behera 2017; Din et al. 2023; Wang et al. 2021). The significantly impacts surface runoff by altering the hydrological
Himalayan watersheds are of particular significance with regard to cycle (Astuti et al. 2019; Dinka and Klik 2019; Tola and Shetty
sustainable development, because they provide several essential 2021). Therefore, the impact of future land use changes on regional
ecosystem services for safeguarding freshwater supply to nearby hydrological characteristics and extremes should be assessed in
lowland areas and hydrological systems (Kozak et al. 2017). More- order to develop mitigation and future water resource management
over, rapid urban growth and degradation of forest cover in the strategies.
mountainous watershed may cause problems with regard to the Spatiotemporal assessment of LULC provides information on
the distribution of natural resources, the extent of human impacts
1
Ph.D. Research Scholar, Dept. of Civil Engineering, Indian Institute of on the environment, and the effectiveness of conservation efforts
Technology, Kharagpur, West Bengal 721302, India (corresponding author). (Twisa and Buchroithner 2019). LULCC is regarded as a critical
ORCID: https://orcid.org/0000-0003-3283-7270. Email: dibyandu@iitkgp
aspect of climate change that decreases biodiversity and ecological
.ac.in
2
Associate Professor, Ranbir and Chitra Gupta School of Infrastructure status locally or globally (Dale 1997; Dale et al. 2011; Gogoi et al.
Design and Management, Indian Institute of Technology, Kharagpur, West 2019; Guo et al. 2022; Hertel et al. 2009; Kumar et al. 2022).
Bengal 721302, India. Email: [email protected] Therefore, land use/land cover (LULC) assessment and prediction
3
Professor, Dept. of Civil Engineering, Indian Institute of Technology, on a finer scale appears to be crucial in managing and monitoring
Kharagpur, West Bengal 721302, India. ORCID: https://orcid.org/0000 watershed hydrology, especially in the Himalayan region (Hassan
-0002-0287-3791. Email: [email protected] et al. 2016). Dynamically changing Land Use and Land Cover
4
Professor, Dept. of Civil Engineering, Indian Institute of Technology, (LULC) can be effectively visualized and accurately understood
Kharagpur, West Bengal 721302, India. Email: [email protected]
through the use of geoinformatics. Geoinformatics combines data
Note. This manuscript was submitted on September 11, 2023; approved
on December 20, 2023; published online on March 8, 2024. Discussion analytics with knowledge obtained from remote sensing data and
period open until August 8, 2024; separate discussions must be submitted geographic information systems (GIS) (Aithal et al. 2018; Saha
for individual papers. This paper is part of the Journal of Hydrologic En- et al. 2022) to develop sustainable water resource management
gineering, © ASCE, ISSN 1084-0699. strategies. There is a heightened need to improve land use

© ASCE 05024006-1 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 information on the macro and micro watershed scale to address (FFVZ). Therefore, prediction of LULC is needed in order to
numerous hydroenvironmental challenges and promote sustainable identify and quantify future high FFVZ.
development practices. Improved land use data can assist in pre- Accurate, timely runoff estimations and predictions of future
dicting and mitigating natural disasters such as floods, landslides, scenarios on a watershed scale are crucial in order to assess and
and forest fires by enhancing understanding of land use patterns develop real-time monitoring systems and mitigation strategies
and identifying vulnerable areas. Therefore, selecting remote sens- against hydrological extremes. Estimating runoff with the help
ing data with sufficient resolution and image classification methods of rainfall-runoff modeling in high-altitude Himalayan areas is
challenging due to the harsh topography and a scarcity of data
with adequate accuracy is crucial when seeking to obtain proper
but inevitable due to the rapid and intense nature of extreme hydro-
LULC information.
Once the current change over a period of time is understood, logical events, especially on monsoon period (Chalise et al. 2003;
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Hannah et al. 2005; Sharma and Rai 2004; Tayal Senzeba et al.
it is imperative to understand and predict future land use changes.
2015). The Himalayan watershed receives significant rainfall over
In this study, the integration of geographic information system-
the monsoon period, leading to a rapid rise in water levels that can
remote sensing (GIS-RS) with statistical or machine learning meth-
contribute to high monsoon runoff (Harik et al. 2023; Merz et al.
ods was employed extensively in order to predict future LULC and
2006). Recently, the Soil and Water Assessment Tool (SWAT), a
changes in LULC (Dey et al. 2021; Mishra and Rai 2016). An in-
physically based semidistributed hydrological model, has been
tegrated stochastic Markov chain (MC) and cellular automata (CA)
widely used to simulate watershed hydrological processes at a
model can effortlessly simulate multidirectional variations and pre-
spatial resolution that is intermediate between lumped and fully
dict future scenarios (Cunha et al. 2021; Gharaibeh et al. 2020;
distributed models (Daide et al. 2022; Maher et al. 2022; Singh
Gidey et al. 2017; Guan et al. 2011; Hyandye and Martz 2017).
et al. 2013). The model considers the spatial variability of input
However, in the last decade, researchers have adopted several hy-
spatial and meteorological data sets and parameters within each
brid models of artificial neural network (ANN)—that is, multilayer
subcatchment, but the processes are assumed to be uniform.
perceptron–Markov chain (MLP-MC) (Dey et al. 2021), multilayer
One of the critical benefits of SWAT is its ability to integrate
perceptron–neural network–Markov chain model (MLP-NN-
high-resolution LULC data, which makes it well-suited for analyz-
MCM) (Saha et al. 2022), logistic regression (LR) (Hyandye ing the impact of land use changes on a finer scale. The SWAT
2015), and geographically weighted regression-based cellular model uses SWAT Calibration and Uncertainty Programs (CUP)
automata (GWRCA) (Zhao et al. 2021)—for simulating and pre- to process the input data, which can be evaluated by calibration
dicting LULC using software tools like Land Change Modeler and uncertainty analysis using several algorithms, such as sequen-
(LCM), Modules of Land Use Change Evaluation (MOLUSCE), tial uncertainty fitting version 2 (SUFI-2) (Hosseini and Khaleghi
and Conversion of Land Use and its Effects at Small regional extent 2020; Thavhana et al. 2018).
(CLUE-S) (Mas et al. 2014). The combination of MC, CA, and The impacts of climate change on regional water resources have
MLP-NN models provides a powerful framework for modeling been extensively studied using hydrological modeling for runoff
complex, dynamic systems with spatial and temporal dimensions, prediction and characteristics under future climate scenarios and
enabling a deeper understanding and prediction of system behavior with future LULC information (Geetha Raveendran Nair et al.
and patterns. Moreover, the hybridization of Markov chain–cellular 2023; Patil and Nataraja 2020; Sajikumar and Remya 2015; Shi
automata–multilayer perceptron–neural network (MC-CA-MLP- et al. 2022). In the process of runoff simulation in most hydrolog-
NN) is superior in terms of logic and accuracy for evaluating future ical models, potential evapotranspiration (PET) and rainfall are
LULC change detection (Mishra and Rai 2016; Rahman et al. essential inputs. A growing pool of climate forecast data is being
2017). Therefore, an integrated simulation-based approach is cru- delivered by global climate modeling centers utilizing global cir-
cial in identifying probable future LULC considering historical culation models (GCMs) (Teng et al. 2023). In order to provide
LULCC patterns, topography, elevation, slope, and so forth increasingly detailed representations of the global climate system
(Chandan et al. 2020; Kafy et al. 2021a). and simulate changes in precipitation and temperature for the lat-
Over the last decade, researchers observed that land urban- ter part of the 21st century, the Coupled Model Intercomparison
ization on topographically harsh terrain is one of the major causes Project Phase 6 (CMIP6) uses new scenarios of shared socioeco-
of increased risk of landslides and flash floods (Li et al. 2017; nomic pathways (SSPs) (Xue et al. 2022). Rainfall-runoff modeling
Mansour et al. 2020; Puente-Sotomayor et al. 2021; Rohan et al. has made substantial use of finer-scale GCM-CMIP6 and is reason-
2023). Deforestation on steep slopes has the potential to tempo- able to execute for watershed with sparse data (Abbaszadeh et al.
rarily decrease evapotranspiration and weaken root reinforcement, 2023; Maurya et al. 2023). In addition, meteorological data sets
increasing vulnerability to landslides. The quantification of the im- should be uniform with regard to monsoon rainfall for observed
pacts of deforestation and related disturbances on the occurrence of and predicted periods in order to analyze the effect of LULCC on
landslides is a significant difficulty, particularly when attempting to monsoon runoff.
make projections for faraway places where only a handful of data In accordance with the research questions discussed previously,
are available (Lehmann et al. 2019). Several studies have found that the prime objectives of this study were: (1) to assess the efficacy of
LUCC may influence the likelihood of landslides by affecting slope an optimized MC-CA-MLP-NN model for predicting future LULC
stability over time (Liu et al. 2021; Meneses et al. 2019; Pisano using high-resolution remote sensing data sets (5.8 m); (2) to ana-
et al. 2017; Quevedo et al. 2023). In addition, unplanned urbani- lyze the effect of LULCC on flash flood vulnerability in the near
zation in topographically harsh terrain is a primary factor in reduc- future in the topographically harsh high-altitude Ranikhola water-
ing infiltration rate and increasing the amount of runoff during shed and carry out spatial assessment of the impact of LULCC on
rainstorms. Excess water can saturate the soil and increase the weight landslide-prone areas; and (3) to evaluate the changes in average
on slopes, making them more prone to flash floods in high-altitude and peak monsoon runoff versus forest cover change and urbani-
regions (Anh et al. 2022; Feng et al. 2021). Recently, a novel flash zation. Runoff estimation and prediction were carried out using
flood vulnerability index (FFVI) was developed for a high-altitude finer-scale uncorrected GCM-CMIP6 meteorological data sets.
Himalayan watershed (Roy et al. 2023), and LULC was found to be LULC maps for 2020 and 2032 were used to simulate runoff using
the most influential factor for a high flash flood vulnerable zone SWAT for two periods, 2015–2020 (observed) and 2021–2032

© ASCE 05024006-2 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 (predicted). The watershed experiences several flash floods and watershed consists of temperate, deciduous forests (i.e., poplar,
landslides yearly due to high monsoon surface runoff. In addition, birch, oak, and elm) and evergreen, coniferous trees of the wet al-
the downstream foothill zones receive a high amount of water from pine zone. The topography of the Ranikhola watershed is highly
upland areas at the peak of the monsoon every year, causing several undulating, with a steep slope, and it is structurally discontinuous.
flood incidents. Therefore, LULC prediction and assessment of the The elevation ranges between 292 and 4,061 m above mean sea
impact of LULC on environmental and hydrological extremes helps level (AMSL). The watershed receives about 3,084 mm of annual
formulate mitigation strategies and proper managerial planning for rainfall, nearly two and a half times the average annual rainfall of
the environment and water resources. India (1,190 mm) (Yadav et al. 2017). Summer temperatures reach
17°C (April to June), and winter temperatures go below 0°C
(December to January). Due to excessive rainfall within the water-
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Study Area shed, landslides occur frequently each year during the monsoon
The study area was the Ranikhola watershed in the east district season. The entire watershed is rated as having high (10% of
of Sikkim, India. It lies between 27°14’15” N and 28°23’49” N the total area) and extremely high flood vulnerability (90% of the
(latitude) and 88°29’25” E and 88°43’19” E (longitude), as shown total area) by the National Flood Susceptibility Assessment System
in Fig. 1. The Ranikhola River is a tributary of the Teesta River (National Remote Sensing Center 2023). On October 4, 2023, a
and encompasses an area of 254 km2 . The capital of Sikkim, significant flash flood occurred throughout the watershed. This
Gangtok—amazingly seductive, delightfully raucous, and cloud- catastrophic, extreme hydrological event resulted in the destruction
wrapped—is situated in the central part of the Ranikhola water- of infrastructure, significant economic losses, and, most impor-
shed. Gangtok’s population density is 5,332 people per km2. The tantly, loss of life. Flash floods in hilly areas trigger landslides and

Fig. 1. Ranikhola watershed, east Sikkim, India.

© ASCE 05024006-3 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 debris flows; these often accompany heavy rainfall and rapidly downloaded from the Alaska Satellite Facility (ASF) Distributed
moving water. In the last decade, 55 significant and several minor Active Archive Center (DAAC). High-resolution visible-infrared
landslides occurred within the Ranikhola watershed. Landslide spectra satellite data from Linear Imaging Self-Scanning Sensor-
inventory points (LIPs) indicate landslide-vulnerable zones within IV (LISS-IV) was acquired from the National Remote Sensing
the watershed. Due to landslides, slope wash, and linear erosion, Centre (NRSC) for 2008, 2013, 2016, and 2020 to understand land
large volumes of silt—called aggradations—are deposited yearly use dynamics. Three toposheets (1∶50,000 scale) were obtained
in the Ranikhola river channels (especially near Ranipool and from the open data archive of the Survey of India (SoI). Toposheet
Singtam). No. 78 A/12 (1932) (1∶63,360 scale) was collected from the historic
open data archive of South Asia, because it was unavailable in the
SoI open data archive. The area’s road network vector data was put
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Materials and Method together using Open Street Map (OSM). The LIPs were obtained
from the Bhukosh web page of the Geological Survey of India (GSI
The overall methodology used to assess the influence of LULCC 2023). Soil properties information for Sikkim was collected from
on hydrological extremes in the future is illustrated in Fig. 2. The the Centre for Interdisciplinary Studies of Mountain and Hill Envi-
methodological framework is divided into three sections: (1) LULC ronment, University of Delhi, India.
prediction by the optimized MC-CA-MLP-NN model; (2) estima- In this study, daily gridded rainfall (0.25° × 0.25°) and maxi-
tion of future flash flood vulnerability and landslide susceptibility; mum and minimum temperatures (0.50° × 0.50°) for the period
and (3) optimized SWAT-based runoff estimation and prediction from 1991 to 2020 were obtained from the India Meteorological
using the CMIP6 model. Department (IMD) (Pai et al. 2014). This data set served as the
observed meteorological data for this study (Dey et al. 2023).
Furthermore, finer daily gridded data for the CMIP6 models with
Data
future scenarios (i.e., SSP1-2.6, SSP2-4.5, and SSP5-8.5) were ob-
Table 1 presents the data sets, their resolutions, and acquisition tained from the Copernicus Climate Change Service (Copernicus
dates or periods. A high-resolution terrain-corrected digital eleva- C3S) website. Daily observed discharge data (2015–2020) from
tion model (DEM), advanced land observing satellite (ALOS), the Central Water Commission (CWC), Sikkim Investigation Divi-
phased array type L-band synthetic aperture radar (PALSAR) ra- sion, government of India, was collected for the Singtam gauge site
diometrically terrain corrected (RTC) (12.5-m resolution) was (i.e., outlet).

Fig. 2. Schematic representation of the methodology adopted for this study.

© ASCE 05024006-4 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 Table 1. Remote sensing, meteorological, and hydrological data sets used in this study
Data sets Resolution/scale Date of acquisition/period
ALOS-PALSAR-DEM 12.5 m June 12, 2009
LISS-IV (Path-107, Row-052) 5.8 m January 9, 2008
January 1, 2013
November 25, 2016
November 25, 2020
Topographic map: 78 A/7, 78 A/8, and 78 A/11 1∶50,000 2010
Topographic map: 78 A/12 1∶63,360 1932
Road map of Sikkim: Open Street Maps Vector 2020
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Landslide inventory points Vector 2020

Soil properties maps of Sikkim 1∶100000 2017
Daily rainfall: IMD 0.25° × 0.25° 1991–2020
Daily maximum and minimum temperature: IMD 0.50° × 0.50° 1991–2020
Daily meteorological data: CMIP6 BCC-CSM2-MR (China) 1.125° × 1.125° 1991–2032
Daily rainfall: CMIP6 MRI-ESM2-0 (Japan) 1.125° × 1.125° 1991–2032
Daily rainfall: CMIP6 NorESM2-MM (Norway) 0.94° × 1.25° 1991–2032
Daily meteorological data: CMCC-ESM2-CMIP6 (Italy) 0.9° × 1.25° 1991–2032
Daily discharge Point 2015–2020

LULC Prediction by Optimized MC-CA-MLP-NN Model networks were extracted from the flow accumulation using Strahler’s
formula (Strahler 1964). Ranikhola (the highest-order stream) and
Preparation of Existing LULC Information and Spatial its major tributaries were considered in order to determine the dis-
Variables for LULC Prediction tance from the stream’s raster. The vector data acquired from the
The surface features of the Ranikhola watershed for 2008, 2013, OSM was used to delineate the distance from the road raster. These
2016, and 2020 were derived using radiometrically and geometri- spatial variables express the impact of nature on the dynamic
cally corrected LISS-IV imagery. Land use/land cover classification mobility of LULC classes (Aburas et al. 2016).
was performed using the maximum likelihood classification (MLC)
Markov Chain–Cellular Automata Model
algorithm of the supervised classification technique for its reliabil-
MC-CA analysis was executed as the dependent variable, and the
ity and higher accuracy (Faisal Koko et al. 2021; Sisodia et al.
independent variable was considered the input variable when deter-
2014; Soheili Majd et al. 2012). More than 100 training sites were
mining predictions of the future. The existing change pattern of the
delineated using a high-resolution virtual earth database based on
LULC maps was regarded as the dependent variable, and the in-
field visits (i.e., ground truth data). We ensured that training sites
dependent variables were the physical constraints considered in
were distributed throughout the study area in order to ensure that
this study. The initial and final LULC areas were computed by their
heterogeneous landforms were well-considered. In this study,
class statistics table. The transition probability matrix (TPM) was
six major LULC classes were recognized: (1) forest (deciduous
calculated to evaluate the ratio of pixels that shifted between
and semievergreen), (2) scrub (alpine scrub and grasslands),
classes. MC, a stochastic model, used historical LULC change
(3) agricultural land, (4) urban area, (5) other (barren land, sand,
trends at various spatiotemporal scales to estimate the likelihood
and snow), and (6) water bodies. The classified output is com-
that LULC would shift from one state to another. In other words,
pared with observed data to evaluate the accuracy of the classifier.
MC is an assortment of random values whose probabilities are
Classwise producer accuracy and accuracy for each classified im-
determined by the weight of the corresponding at a prior time
age were also computed (Foody 2020). The overall accuracy of the
(Clarke et al. 1997; Gidey et al. 2017). The general equation for
classified LULC maps was determined by the ratio of the sum of
MC analysis is expressed as follows:
correctly classified pixels to the total number of pixels. The pro-
ducer accuracy of the classified LULC maps was determined by the Qðt;tþ1Þ ¼ ρij × Qt ð1Þ
ratio of the number of correctly classified pixels belonging to a spe-
cific class to the sum of correctly classified pixels and the number where QðtÞ = system status at initial time t; Qðt þ 1Þ = system
of pixels that belong to that class but were incorrectly classified as status at time t þ 1; and ρij = TPM in a time-state that is com-
something else or not classified at all. puted using the following formula (Mondal et al. 2016; Subedi
After gaining the desired accuracy for the classified images and et al. 2013):
checking the independence of the physical constraints by evaluat-
0 1
ing Pearson’s correlation, three LULCC maps (2008–2013, 2013– ρ11 ρ12 : : : ρ1n
2016, and 2016–2020) were obtained (Table S1). However, there is B C
B ρ21 ρ22 : : : ρ2n C
no rationale for the attributes chosen to evaluate changes in land use B C
ρij ¼ B . .. .. .. C
and land cover, and it may be tailored to each study area (Aburas B .. . . . C
@ A
et al. 2016; Mishra and Rai 2016). The major influencing topo-
graphic parameters, such as elevation and slope, were identified ρn1 ρn2 ::: ρnn
using high-resolution DEM. However, the local scale should also !
X
n
be considered for simulation purposes, because each place has × 0 ≤ ρij ≤ 1; ρij ¼ 1; ði; j ¼ 1; 2; ..; nÞ ð2Þ
unique characteristics that must be considered when selecting other jþ1
factors (Fig. S3). The D8 approach from the DEM was used to
calculate flow direction (Erdbrügger et al. 2021). Flow accumula- The CA model’s publicly scalable structure and ability to inter-
tion was evaluated based on flow direction. Therefore, drainage face with other models are its key benefits (Clarke et al. 1997).

© ASCE 05024006-5 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 A contiguity filter (5 × 5 pixels) was applied to characterize the model parameters or hyperparameters (i.e., L, I, HL, and M) were
effect of surrounding cells on the central pixel (Wang et al. 2021). chosen based on the highest value of K c , and the lowest value of E
Each pixel was assigned to a class according to how well-suited it was considered the first error criterion. Furthermore, the final neu-
was for a given LULC class and how closely related it is to pixels in ral architecture was selected for the lowest value of ΔA , considered
that class. The transition rules in the model considered the state of the last error criterion. The sample size (S) was also chosen for
surrounding pixels and a latest LULC map. Despite its benefits, the simulation and prediction based on these values (Table 2).
CA model struggle for defining model structure (Rocha et al.
2007). The CA model was mathematically estimated as follows Integration of MC-CA-MLP-NN Model and
(Chopard 2009; Souza-e-Silva et al. 2009): Cross-Validation Strategy
This investigation employed the Markov chain and cellular au-
Sðt;tþ1Þ ¼ fðSðtÞ; NÞ ð3Þ
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

tomata methodologies to forecast initial land cover shifts within a

specific time interval utilizing historical data and accounting for
where SðtÞ = cellular state at time t; Sðt þ 1Þ = cellular state at time spatial limitations. The multilayer perceptron (MLP) model en-
t þ 1; functioned by the state probability of any time (N); and f = hanced these predictions by including supplementary factors and
transition rule. Therefore, the integrated model (MC-CA) ensured optimizing the transition probabilities. In addition to Markov chain
improved forecast accuracy and encompassed static and dynamic and cellular automata, a neural network took spatial constraints into
features of LULC changes (Kafy et al. 2021b; Li et al. 2015; Wang account—something the other models did not comprehensively
et al. 2021). The integrated MC-CA model can be calibrated and describe in isolation.
decomposed and can anticipate a spatiotemporal dynamic event This study used existing LULC maps with physical constraints
connected to LULC predictions (Saha et al. 2022). (i.e., slope, elevation, distance from roads, and distance from
Multilayer Perceptron–Neural Network (MLP-NN) Model streams) for k-fold cross-validation. The k-fold cross-validation
The multilayer perceptron–neural network (MLP-NN) model was technique is an essential technique for evaluating the performance
executed as a transition potential model in this study (Parsamehr of multilayer perceptron neural networks (MLP-NNs) (Islam et al.
et al. 2020). LULC pattern was predicted using an MLP-NN–based 2021; Roy and Saha 2022) when data sets are limited. This study
MC-CA model. Transformation operations performed the MLP- employed a fourfold cross-validation approach consisting of two
NN model to estimate TPM for the future. The MLP-NN model has iterations for forward simulation (i.e., F-1: from LULC 2008 and
been performed based on a backpropagation algorithm with a tanh 2013 to LULC 2016; F-2: from LULC 2013 and 2016 to LULC
sigmoid function, expressed as follows: 2020) and two iterations for backward simulation (i.e., F-3: from
LULC 2020 and 2016 to LULC 2013; F-4: from LULC 2016
e2x − 1 and 2013 to LULC 2008). The primary objective was to assess
fðxÞ ¼ tanhðxÞ ¼ ð4Þ
e2x þ 1 each simulated LULC map’s overall accuracy compared to the
existing LULC map. Last, LULC 2016 and 2020 were used to
The sampler uses both specific pixels and surrounding pixels simulate LULC 2024; LULC 2020 and 2024 were used to simu-
(3 × 3). The hidden layers (HL) were selected arbitrarily; 80% of late LULC 2028; and LULC 2024 and 2028 were used to simulate
the samples were used as a learning set, and 20% were used as a LULC 2032.
validation set for training the model. The samples were weight-
corrected as follows:
Estimation of Future Flash Flood Vulnerability and
wðI þ 1Þ ¼ ðL × dwðIÞÞ þ ðM × dwðI − 1ÞÞ ð5Þ Landslide Susceptibility
where w = weight vector of neurons; I = maximum iteration num- Land use change can significantly influence the occurrence of flash
ber; L = learning rate; and M = momentum. This study assessed floods and landslides in the Himalayan region. The Himalayas are a
several neural network architectures by varying parameters or complex and sensitive ecosystem, and alterations in land use can
hyperparameters. The best-fitted MLP-NN model was found by have profound consequences on the region’s hydrology and stabil-
evaluating the error criterion of these network architectures. The ity. In recent years, the proliferation of urban areas and the unregu-
accuracy of the transition potential model was estimated using the lated expansion of infrastructure have resulted in a notable increase
following parameters with their predefined ranges (Table 2). in surface runoff. Increased runoff contributes to the occurrence of
The optimized model for simulation and prediction was selected flash floods and landslides during the peak monsoon season. In this
based on these error parameters (Table 2). The transition potential study, future flash flood vulnerability was assessed by computing
FFVI (Roy et al. 2023) as follows:
N iSA
!
Table 2. Error criteria for defining the optimal model XNA X
Predefined FFVIpx;py ¼ Wi wik χAik ðCvpx;py Þi ð6Þ
i¼1 k¼1
Error criteria Equation range
Pn
Ψμ Future FFVI was computed by the weighted overlay method,
Minimum validation overall error (E) E ¼ i¼1 <0.02
n considering the predicted LULC with existing weights of all param-
PðAv Þ − PðEv Þ eters (Table S2). Other morphometric and physiographic parameters
Current validation kappa (K c ) Kc ¼ >0.80 were considered unchanged from the observed to the predicted
1 − PðEv Þ
period. The geometrical interval classifies the future FFVI reclassi-
Delta overall accuracy (ΔA ) ΔA ¼ jEm − Ec j <0.01 fied approach. Last, future FFVI was categorized into three FFVZs:
Note: Ψμ = mean of all Ψ on the learning and validation sets after every low, moderate, and high. The impact of LULCC on high FFVZ was
P
epoch; n = number of sets; PðAv Þ ¼ ci¼1 cij , cij represents the ith, jth assessed by comparing existing FFVZ with predicted FFVZ.
P
cells on the validation set of the contingency table; PðEv Þ ¼ ci¼1 cid cdj ; Flash floods and landslides are interconnected in certain situa-
cid and cdj = summations of the ith row and the jth column; and Em and Ec tions, especially in topographically harsh terrain. Heavy rainfall
represent the minimum and current validation errors. saturates the soil on slopes, making it unstable and prone to sliding

© ASCE 05024006-6 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 downhill. Waterlogged soil is more susceptible to landslides, espe- amount on ith day, respectively (mm); W s = amount of water en-
cially in areas with steep terrain. Conversely, landslides trigger tering the vadose zone from the soil profile on the ith day; and
flash floods if debris blocks rivers or streams, forming temporary Qg = amount of groundwater flow on the ith day (mm) (Neitsch
dams. When these dams fail or breach due to accumulated water et al. 2011).
pressure, the sudden release of water can result in a flash flood SWAT divided each subwatershed into 10 elevation bands to
downstream. Furthermore, deforestation and the deliberate removal consider orographic effects on temperature, precipitation, and solar
of natural vegetation for purposes such as agriculture, building, or radiation (Shukla et al. 2021). The Muskingum method was se-
infrastructure development diminish the stabilizing impact that tree lected for computing flow routing through the channel, and the
roots and vegetation have on the soil. In the absence of vegetation, Penman-Monteith equation was adopted to estimate potential
soil’s vulnerability to erosion increases, increasing the likelihood of evapotranspiration. The modified Soil Conservation Service runoff
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

landslides during periods of intense precipitation. Heightened soil curve number (SCS-CN) was chosen for estimating surface runoff
erosion can also contribute to the deposition of sediment in river volume considering land use, soil type, and antecedent moisture
channels, which can result in flash floods during periods of severe conditions (Slack and Welch 1980). In this study, the Ranikhola
precipitation when a river’s ability to transport water surpasses its watershed was divided into 25 subwatersheds, including 124 HRUs.
capacity. Therefore, assessing LULCC around the collected LIPs The nonzero HRU threshold method was employed, notwithstand-
is crucial for landslide susceptibility mapping. Moreover, the pres- ing the potential loss of information on the river basin landscape.
ence of overlapping LIPs in the projected LULC maps could be The threshold was applied for land cover (15%), soil (10%), and
beneficial for the purpose of validating future FFVZs and gaining slope (10%) in defining HRUs in order to improve computational
insight into potential scenarios of increased future susceptibility to efficiency (Femeena et al. 2022; Her et al. 2015; Srinivasan et al.
landslides. 2010).
Initially, the SWAT model was performed from 2015 to 2020
Optimized SWAT-Based Runoff Estimation and on a daily scale using IMD data sets. The simulation model was
Prediction Using the CMIP6 Model calibrated and validated for daily discharge by comparing it to the
observed discharge at the outlet gauge site, Singtam (Fig. 1),
from 2015 to 2020. The daily discharge data from the gauge sta-
CMIP6 Model Selection for Hydrological Modeling
tion for 2015–2018 was used for model calibration, and valida-
GCM-CMIP6–based rainfall and temperature data are helpful for
tion was carried out for 2019–2020 for subwatershed 25. A total
the proper representation of climate scenarios. This study focused
of 22 parameters were designated based on their sensitivities to
on selecting the best possible region-specific GCM and using it
observed streamflow and other relevant watershed conditions,
for hydrological modeling. The uncorrected GCMs that most ef-
such as LULC, topography, and hydrological state of the water-
fectively represented climatological variables—including number
shed during the calibration process. The calibration process in-
of zero rainfall days, average rainfall per day, number of rainy days,
volved 20 iterations with 200 simulations each. Subsequently,
peak value, annual rainfall, and monsoon rainfall—with observed
the optimized SWAT model was incorporated with uncorrected
data were identified without considering any statistical methods
CMIP6 data sets and existing LULC information. A comparison
(Lovino et al. 2021; Mesgari et al. 2022; Ngoma et al. 2021; Wang
of the observed discharge from 2015 to 2020 was used to cali-
et al. 2021). This can also enhance the reliability of unadjusted,
brate and validate the available SSPs for daily discharge. Statis-
more appropriate GCM-CMIP6 data compared to observed data
tical matrices, including coefficient of determination (R2 ), percent
for hydrological analyses conducted at the watershed or subwa-
bias (PBIAS), Nash-Sutcliffe efficiency (NSE), and the ratio of
tershed level. In addition, the better-fitted GCM-CMIP6 data were
the root-mean square error to the standard deviation of the ob-
selected by computing the average, maximum, and minimum de-
served data (RSR), were used to define the effectiveness of using
viations from observed data for the rainfall parameters considered
finer uncorrected GCM meteorological data sets. Furthermore,
for this analysis.
higher values for R2 and NSE and lower values for PBIAS and
Semidistributed Hydrological Modeling for Runoff RSR assisted in choosing the best-suited SSP for future runoff
Estimation estimation (Table 3).
This study used SWAT as a physically based, semidistributed
hydrological model (Arnold et al. 1998, 2012) for runoff esti-
mation. A soil properties map, topographical map (i.e., slope Table 3. Expression and range of statistical matrices considered for runoff
and elevation), and LULC were the primary spatial input for estimation and prediction
the SWAT model (Swain et al. 2022). The watershed was divided Name Expression Range
into hydrologic response units (HRUs) to indicate spatial hetero- Pn
½ ðyobs − yobs Þðysim − ysim Þ2
geneity in the modeling. An HRU has uniform properties in terms R2 R2 ¼ Pn i¼1obs i obs u2 Pn i sim u sim 2 0 to 1
of topography, LULC, and soil type (Sun et al. 2021; Xu et al. i¼1 ðyi − yu Þ i¼1 ðyi − yu Þ
2023). Due to the HRU architecture, the meteorological and spa- Pn 
i¼1 ðyi − ysim Þ
obs
tial data were generalized, which affected the model’s resolution. PBIAS PBIAS ¼ 100 × Pn obs i −∞ to ∞
i¼1 yi
The water balance equation served as the foundation for the Pn 
2
SWAT model’s simulation of the hydrological cycle, expressed ðyobs − ysim i Þ
NSE NSE ¼ 1 − Pi¼1 i
obs 2
−∞ to 1
i¼1 ðyi − yu Þ
n obs
as follows:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pn ffi
X
t
i¼1 ðyi
obs
− ysim i Þ
2
SW t ¼ SW o þ ðRday − Qs − Ea − W s − Qg Þ ð7Þ RSR RSR ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pn ffi 0 to ∞
2
i¼1 ðyi − yobs u Þ
i¼1 obs

where SW t = final soil water content (mm); t = time (days); Note: yobs
i and ysim
i = observed and simulated runoff values for each
SW o , Rday , Qs , and Ea = initial soil water content, precipitation day, respectively; and ysim
u and yobs
u = mean of simulated and observed
amount, amount of surface runoff, and evapotranspiration (ET) values, respectively.

© ASCE 05024006-7 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 The best-suited Centro Euro-Mediterraneo per I Cambiamenti 2016 and 2020. Other lands barely changed during this period, in-
Climatici Earth System Model Version 2 Coupled Model Inter- dicating past infrastructure development.
comparison Project phase 6 (CMCC-ESM2CMIP6) data sets The transition probability matrices for the different periods
(rainfall and maximum–minimum temperature) and predicted are shown in Table 8. The values of diagonal cells of matrices re-
LULC information were used to simulate future runoff for the present the proportion of unchanged pixels between two periods. In
Ranikhola watershed. Monsoon (i.e., May to October) runoff contrast, a specific diagonal cell’s row and column values are ex-
and peak magnitude were assessed for both periods (observed press the proportion of changing pixels from one class to another
and predicted). between subsequent years. The transition matrices show that the
diagonal value for urban regions is constant, indicating that they
neither decreased nor shifted to another class. The LULCC maps
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Results and Discussion reveal that significant portions of forest cover around Gangtok,
Rumtek, and Singtam were converted to scrubland and urban re-
Land Use/Land Cover Dynamics and Prediction gions in subsequent years (Fig. 3). Maximum forest-to-urban-areas
(5.80 km2 ) transition was observed from 2016 to 2020, while the
Existing Land Use/Land Cover Dynamics maximum increase in scrubland (3.83 km2 ) and agricultural land
The resolution of satellite data has an impact on geospatial analy- (1.08 km2 ) was from 2008 to 2013 (Table S3). The clearing of for-
sis for a small mountainous watershed like Ranikhola. Therefore, est land for construction may have produced a subtle but intriguing
high-resolution LISS-IV (5.8 m) data were utilized to prepare the transition from forest to scrubland and other land, as seen in the
existing LULC (as input) (Fig. S1). According to the classifica- transition matrices between 2013 and 2020. Other land (especially,
tion results, the highest percentage of forest was found in 2008 barren land) transitioned to agricultural land, and scrubland was
(85%), followed by 2013 (80%), 2016 (77%), and 2020 (72%). significantly greater between 2016 and 2020, followed by 2013
The lowest percentage of scrubland was found in 2016 (5%), and and 2016. Most of the bare ground in the northeastern extension
other years were relatively identical (Fig. S2, Tables 5–7). How- of the watershed transitioned to a built-up area in this period.
ever, maximum agricultural areas (3%) were observed in 2020, The TPMs from 2008 to 2020 served as the foundation for the
followed by 2016 (2%) and urban areas (14%), followed by development of the TPMs from 2020 to 2032.
2016 (11%), 2013 (9%) and 2008 (7%). Other land cover areas
(i.e., sand, barren land, and snow cover) increased during the
period from 2008 (1%) to 2020 (6%). The accuracy of the clas-
Table 5. Classwise LULC statistics from 2008 to 2013
sified images was assessed with the help of training sites pre-
pared from ground truth data (2020) and a high-resolution virtual LULC 2008 2013 Δ 2008 2013 Δ
earth database (2008, 2013, and 2016). The overall accuracy and classes (km2 ) (km2 ) (km2 ) (%) (%) (%)
classwise producer accuracy of the classified images are shown in Forest 215.477 204.647 −10.830 84.676 80.420 −4.256
Table 4. Scrubland 14.941 14.550 −0.391 5.871 5.718 −0.154
The classified images were reasonably satisfactory when the Agricultural 2.063 3.504 1.441 0.811 1.377 0.566
MLC algorithm–based supervised classification technique was land
used. The classification results amply demonstrate that the degra- Urban areas 17.578 21.879 4.301 6.908 8.598 1.690
dation of forest cover (14%) and an increase in urban area (97%) Others 3.732 9.289 5.556 1.467 3.650 2.183
Water bodies 0.681 0.594 −0.088 0.268 0.233 −0.035
in the study area is a primary concern from 2008 to 2020. Forest
degradation has resulted from population growth, residential devel-
opment, and infrastructure development for the tourism industry
around the central and southwestern parts of the watershed over Table 6. Classwise LULC statistics from 2013–2016
the last decade.
LULC 2013 2016 Δ 2013 2016 Δ
Transition of Land Use/Land Cover (2008–2020) classes (km2 ) (km2 ) (km2 ) (%) (%) (%)
LULC maps were assessed between the subsequent years and the Forest 204.647 195.461 −9.186 80.420 76.810 −3.610
classwise areal distributions were computed (Tables 5–7). Also, the Scrubland 14.550 12.049 −2.501 5.718 4.735 −0.983
transition of LULC classes (Fig. 3) and TPMs were constructed Agricultural 3.504 4.537 1.033 1.377 1.783 0.406
(Table 8) to find trends in LULCC from 2008 to 2020. The maxi- land
mum growth in urban area (an expansion of 7.267 km2 ) was ob- Urban areas 21.879 27.487 5.608 8.598 10.802 2.204
served from 2016 to 2020. The maximum decrease in forest cover Others 9.289 14.226 4.938 3.650 5.591 1.940
(11.13 km2 ) and agricultural land expansion (1.96 km2 ) was ob- Water bodies 0.594 0.681 0.087 0.233 0.268 0.034
served during the same period. Scrubland only increased between

Table 7. Classwise LULC statistics from 2016 to 2020

Table 4. Accuracy assessment of the LULC map in 2008, 2013, 2016,
and 2020 LULC 2016 2020 Δ 2016 2020 Δ
classes (km2 ) (km2 ) (km2 ) (%) (%) (%)
Producer accuracy (%)
Overall
Forest 195.461 184.333 −11.128 76.810 72.437 −4.373
Agricultural Urban Water accuracy
Scrubland 12.049 13.485 1.436 4.735 5.299 0.564
Year Forest Scrubland land areas Others bodies (%)
Agricultural 4.537 6.500 1.962 1.783 2.554 0.771
2008 99.98 88.11 84.93 73.17 93.76 83.91 98.29 land
2013 99.98 94.74 81.56 78.31 96.93 81.43 98.66 Urban areas 27.487 34.754 7.267 10.802 13.657 2.856
2016 99.87 87.77 83.66 86.31 96.37 84.29 98.11 Others 14.226 14.650 0.423 5.591 5.757 0.166
2020 99.56 84.24 98.62 85.10 96.37 86.38 98.05 Water bodies 0.681 0.683 0.002 0.268 0.268 0.001

© ASCE 05024006-8 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Land use/land cover change maps: 2008–2013, 2013–2016, and 2016–2020.

Prediction of LULC for 2032

Table 8. TPMs of LULC classes from 2008–2013, 2013–2016, and
2016–2020 The best-fitted transition potential model was determined by
training the model under the set of parameters/hyperparameters
LULC Agricultural Urban Water (i.e., L ¼ 0.001, I ¼ 100, HL ¼ 10, M ¼ 0.05, and S ¼ 5,000)
classes Forest Scrubland land areas Others bodies
(Tables S5 and S6). The error criterion values (i.e., E ¼ 0.0152,
2008–2013 K c ¼ 0.8274, ΔA ¼ 0.003) revealed a high efficacy of the model
Forest 0.950 0.018 0.005 0.014 0.013 0.000 for generating future LULC maps. Therefore, existing LULC
Scrubland 0.000 0.710 0.019 0.077 0.193 0.001 maps with the considered spatial constraints were used for training,
Agricultural 0.000 0.000 0.992 0.008 0.000 0.000
testing, and validation. A cross-validation strategy (i.e., fourfold,
land
Urban areas 0.000 0.000 0.000 1.000 0.000 0.000 F-1 to F-4) was adopted to establish the robustness of the model.
Others 0.000 0.052 0.038 0.029 0.866 0.015 The simulated LULC maps were validated by computing their
Water bodies 0.000 0.004 0.019 0.102 0.285 0.590 overall accuracy (Foody 2020) compared to existing LULC maps.
Fig. 4 depicts typical forward (F-2) and backward (F-4) simulation
2013–2016
process examples. The overall accuracy values for forward simu-
Forest 0.954 0.005 0.001 0.020 0.020 0.000
Scrubland 0.000 0.732 0.016 0.071 0.180 0.001
lations were 92.85% and 93.68%; the values for backward simu-
Agricultural 0.000 0.000 0.994 0.006 0.000 0.000 lations were 90.06% and 91.38%. In addition, producer accuracy of
land classwise values was satisfactory (>85%). The overall accuracy
Urban areas 0.000 0.000 0.000 1.000 0.000 0.000 values of simulated LULC maps define the execution of the
Others 0.000 0.060 0.083 0.043 0.805 0.009 MC-CA-MLP-NN model was better in the forward simulation than
Water bodies 0.000 0.010 0.007 0.110 0.103 0.770 in the backward simulation. Therefore, LULC 2016 and LULC
2016–2020 2020 were used to simulate LULC 2032. In order to predict LULC
Forest 0.943 0.003 0.002 0.030 0.021 0.001 for 2032, the optimal neural architecture was employed for the sim-
Scrubland 0.000 0.995 0.000 0.005 0.000 0.000 ulation process. During the simulation process, the initial error
Agricultural 0.000 0.000 0.984 0.016 0.000 0.000 requirements were satisfied (i.e., E < 0.01, K c > 0.9, ΔA < 0.002),
land suggesting a high level of spatial accuracy in the anticipated
Urban areas 0.000 0.000 0.000 1.000 0.000 0.000 outcome.
Others 0.000 0.082 0.131 0.049 0.733 0.005 The rapid growth of the tourism industry and other struc-
Water bodies 0.000 0.007 0.003 0.137 0.218 0.636
tural developments around the central southwestern part of the

© ASCE 05024006-9 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Typical examples of the K-fold cross-validation strategy adopted for LULC map simulation.

Ranikhola watershed have affected the forest cycle in recent years. locations around the Ranikhola River and Gangtok, the maximum
According to our analysis, there was a spatial trend of decreasing forest cover has transitioned to urban areas and scrubland. In con-
forest cover area from 2008 to 2020 (a 14.5% decrease), and this trast, in the northeastern part of the watershed, minimal land use
trend will continue until 2032. The areal distribution in LULC transitions were observed around the LIPs (Fig. 6). According
classes is depicted by Δ (km2 and %) (Table S4). The rates of to the LULC maps for 2020 and 2032, around 60% and 70% of
change in forest cover and other land cover areas are projected the LIPs are in urban areas. Gangtok, the capital city, has a height-
to decrease by 4.70% and 0.30%, respectively, while the rates ened susceptibility because of its highly populated areas situated on
of change in scrubland and urban areas are projected to increase unstable hills.
by 1.86% and 3.04%, respectively, over the next 12 years. How- Recently (July 14, 2023), a major landslide occurred in the 9th
ever, agricultural lands and water bodies showed minimal changes Mile region along Jawaharlal Nehru Road, disconnecting Nathula
during this period. The transition of scrub, agricultural lands, and Pass, which serves as a link between India and China, and pop-
urban areas remained constant; forests showed the maximum tran- ular tourist attractions like Tsomgo Lake and Baba Mandir from
sition over this time period (Table S3). The prediction results in- Gangtok, the capital city of Sikkim. On October 9, 2022, the was
dicate that the maximum LULC transition from 2020 to 2032 a disruption in the northeastern state’s transportation infrastructure
will be from forest to urban areas (6.93 km2 ), followed by forest due to ongoing heavy rainfall that led to a series of landslides along
to scrubland (4.63 km2 ), and other land to urban areas (0.80 km2 ). NH-10 in the vicinity of Singtam and Ranipool. The increased
The changes in LULC from 2020 to 2032 and spatial prediction danger of landslides during the monsoon is significantly influenced
results for the Ranikhola watershed are shown in Fig. 5. In addition, by the expansion of impervious surfaces from forests on the steep
the increase in urban areas signifies hasty population growth, which slopes of the watershed (Quevedo et al. 2023). In addition, this
may continue to be a problem until environmentally responsible causes an increase in surface runoff, which leads to flash floods, soil
land use practices are put in place to balance demand. erosion, and silt deposition on riverbeds. Proper planning and land
use regulations are essential in mitigating the impact of urbanization
on landslides. Conducting detailed geotechnical assessments before
Impact of Predicted LULCC on Landslides and Flash
urban development in the future is crucial in identifying and avoiding
Flood Vulnerable Areas
high-risk landslide zones (i.e., around LIPs).

Impact of LULCC on Landslide Prone Areas Future Flash Flood Vulnerability

The Himalayan watershed of Sikkim, Ranikhola, is susceptible to The predicted LULC map was used to develop future FFVZs for
rainfall-related earthquakes and landslides. In landslide-prone the Ranikhola watershed (Fig. 6), assuming that its physiographical

© ASCE 05024006-10 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Predicted LULC map for 2032 and LULCC map from 2020 to 2032 with LIPs.

Fig. 6. Predicted (2032) and existing FFVZ (2020) maps.

© ASCE 05024006-11 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 and morphometrical character remains unchanged from 2020 to hydrological model. Therefore, daily rainfall data (1991–2032)
2032. The areal distributions of FFVZs in 2020 to 2032 are shown were collected for the available CMIP6 models: Beijing Climate
in Table 9. Between 2020 and 2032, high FFVZs are expected to Center Climate System Model version 2 Medium Resolution
increase by 9% due to increased urbanization (3.04%) and de- (BCC-CSM2-MR) (China), Norwegian Earth System Model
creased forest cover (4.70%). The increase in high FFVZs also version 2–Medium Resolution (NorESM2-MM) (Norway),
portends an increase in landslide incidents, because land- Meteorological Research Institute Earth System Model version
slides are interrelated with flash floods and arise on a sequence 2-0 (MRI-ESM2-0) (Japan), Centro Euro-Mediterraneo per I Cam-
of occasions. biamenti Climatici Earth System Model version 2 (CMCC-ESM2)
The flash floods that occurred during the early hours of October (Italy) for the study area. This study did not employ bias correction
4, 2023, significantly impacted 41,870 individuals residing in or other statistical approaches because of its focus on evaluating the
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

four districts of Sikkim, namely Gangtok, Mangan, Namchi, and uncorrected CMIP6 models’ ability to reflect temporal rainfall pat-
Pakyang. In these districts, around 30,000 people experienced the terns. Therefore, the following parameters were analyzed to find the
most severe consequences of this catastrophic event (The Hindu best-suited GCM-CMIP6 model by comparing them with observed
2023). The recent flash flood significantly impacted the central data sets (i.e., IMD) for the Ranikhola watershed: number of zero
(around Gangtok and Ranipool) and lower (around Rumtek and rainfall days, average rainfall per day, number of rainy days, peak
Singtam) regions of the watershed. This extreme hydrological event value, annual rainfall, and monsoon rainfall (i.e., from May to
disrupted National Highway 10, causing damage to bridges across October).
the Ranikhola River and destroying facilities like power supplies, Table 10 depicts the yearly deviation of the rainfall parameters
hotels, and tourist attractions. Based on our investigation, these from the observed data considered for this study. The maximum
regions (especially southwestern part of the watershed) were al- variations in annual rainfall, monsoon rainfall, and peak value were
ready classified as high FFVZs. Furthermore, the projected LULC lower for the NorESM2-MM model. However, the CMCC-ESM2
scenario indicates that vulnerability is expected to be higher for model showed a reduced deviation from observed rainfall param-
regions where forest cover transitions to the other LULC classes. eters apart from those values. This model’s average, maximum, and
minimum deviations for climatological parameters were also within
Estimation and Prediction of Monsoon Runoff an acceptable range compared to the other GCM-CMIP6 models.
Using SWAT Therefore, the CMCC-ESM2 model’s meteorological data sets
were utilized for further hydrological analysis.
Identification of Better-Fitted GCM-CMIP6 Model
CMIP6-Based Runoff Estimation
Daily meteorological input data, especially rainfall data, is the pri-
The semidistributed hydrological model (i.e., SWAT) was imple-
mary requirement for estimating future runoff using an optimized
mented for the Ranikhola watershed using high-resolution spatial
data sets (Fig. S4). For HRU production, physiographic factors
(i.e., elevation and slope), soil properties information, and LULC
Table 9. Areal distributions of FFVZs in 2020 and 2032
were used. Runoff was simulated at each subwatershed outlet using
FFVZ area (%) the observed meteorological data sets (i.e., IMD). In this study,
Year Low Moderate High the Ranikhola watershed was divided into 25 subwatersheds,
2020 41.50 36.41 22.10
including 124 HRUs. The nonzero HRU threshold method was
2032 28.21 40.79 31.00 employed, notwithstanding the potential loss of information on
the river basin landscape. The performance of the model was

Table 10. Average, maximum, and minimum deviation of GCM-CMIP6 rainfall parameters from observed data
Deviations of parameters from the IMD data set
Avg. No. of
Annual Monsoon rainfall Peak Peak rainy No. of
rainfall rainfall per day value rainfall days zero-rainfall
GCM-CMIP6 (mm) (mm) (mm) (mm) day (day) days (day)
Average
CMCC-ESM2 −241.72 −392.35 −0.65 −35.54 −15.54 6.50 −2.46
BCC-CSM2-MR −2,150.64 −1,818.65 −5.89 −154.72 −200.58 −120.38 −169.04
MRI-ESM2-0 −1,488.99 −1,332.79 −4.08 −76.63 −191.96 −120.79 −133.17
NorESM2-MM −1,712.02 −1,760.54 −4.03 −111.33 −216.04 −115.63 −176.71
Maximum
CMCC-ESM2 956.63 880.01 2.65 50.32 81.00 38.00 38.00
BCC-CSM2-MR −1,639.47 −1,347.19 −4.49 −87.40 −144.00 −102.00 −140.00
MRI-ESM2-0 −1,007.14 −840.45 −2.76 −30.43 −136.00 −90.00 −100.00
NorESM2-MM −674.23 −587.97 −2.91 −24.31 −112.00 −68.00 −125.00
Minimum
CMCC-ESM2 −1,371.93 −1,500.68 −3.74 −98.30 −28.00 −97.00 −60.00
BCC-CSM2-MR −3,150.45 −2,690.57 −8.63 −261.78 −137.00 −255.00 −187.00
MRI-ESM2-0 −2,008.91 −1,814.07 −5.50 −109.87 −150.00 −252.00 −183.00
NorESM2-MM −2,552.40 −2,719.53 −5.59 −194.69 −163.00 −310.00 −250.00
Note: Bold represents minimum values.

© ASCE 05024006-12 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 Table 11. Sensitivity analysis and range with fitted values of parameters
Fitted Minimum Maximum
Parameter name Description t-statistic p-value value value value
GW_REVAP.gw Groundwater “revap” coefficient −0.013 0.989 0.021 0.011 0.027
GWQMN.gw Threshold depth of water in the shallow −0.014 0.989 −0.168 −0.349 −0.020
aquifer required for return flow to occur (mm)
SMFMN.bsn Minimum melt rate for snow during the year −0.036 0.971 3.876 3.174 4.226
CH_K2.rte Effective hydraulic conductivity in main −0.050 0.960 18.781 16.364 20.991
channel alluvium
RCHRG_DP.gw Deep aquifer percolation fraction 0.059 0.953 0.468 0.442 0.481
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

GW_DELAY.gw Groundwater delay time (days) −0.069 0.945 217.627 212.513 233.605
OV_N.hru Manning’s n value for overland flow 0.088 0.930 0.549 0.501 0.565
SOL_AWC.sol Available water capacity of the soil layer 0.090 0.929 1.219 1.131 1.491
PLAPS.sub Precipitation lapse rate 0.100 0.921 154.433 147.924 158.297
ALPHA_BF.gw Baseflow alpha factor (days) 0.105 0.916 0.116 0.099 0.123
SFTMP.bsn Snowfall temperature −0.134 0.894 3.606 3.378 4.780
SNOCOVMX.bsn Minimum snow water content that −0.140 0.889 292.162 269.006 321.933
corresponds to 100% snow cover
TLAPS.sub Temperature lapse rate −0.161 0.872 −2.189 −2.394 −0.904
CN2.mgt SCS runoff curve number 0.174 0.862 −0.711 −0.721 −0.675
ESCO.hru Soil evaporation compensation factor 0.175 0.862 0.765 0.720 0.787
SNO50COV.bsn Snow water equivalent that corresponds to 0.177 0.860 0.436 0.402 0.469
50% snow cover
SURLAG.bsn Surface runoff lag time 0.198 0.843 0.629 0.561 0.650
SMTMP.bsn Snow melt base temperature −0.233 0.816 2.588 2.389 2.746
CH_N2.rte Manning’s n value for the main channel −0.251 0.802 0.147 0.144 0.154
TIMP.bsn Snowpack temperature lag factor −0.443 0.659 0.591 0.543 0.605
SOL_K.sol Saturated hydraulic conductivity 0.528 0.598 −0.698 −0.707 −0.670
SMFMX.bsn Maximum melt rate for snow during the year 0.592 0.555 17.133 16.382 17.233
Note: The fitted, minimum, and maximum values are the relative values, and the existing parameter value is multiplied by (1 þ a given value).

satisfactory, as evidenced by overall p-factor and r-factor values that the calibrated hydrological parameters (Table 11) were static.
of >0.70 and 0.10, respectively. The greater p-value and smaller The findings indicated that, consistent with the prevailing pattern
t-statistic value indicate the sensitivity and relevance of the param- observed between 2015 and 2020, there is a projected increase
eters. Groundwater “revap” coefficient (GW_REVAP.gw) was in the daily runoff trend for 2021–2032 (Fig. 8). In the upcom-
found to be the most sensitive parameter for the Ranikhola water- ing years (2021–2032), the average peak magnitude (59.46 m3 =s)
shed (Table 11). of monsoon runoff will rise by 26.00% over the calibrated pe-
The SWAT model occasionally overestimated runoff compared riod (2015–2020). The peak monsoon runoff value indicates an
to observed data during the calibration and validation periods, as expected period of surface hydrological extremes in this topo-
indicated by the SWAT model’s positive PBIAS value (see Appendix). graphically harsh watershed. Furthermore, the average monsoon
The higher R2 (0.82), NSE (0.80), and lower RSR (0.44) values (May to October) runoff magnitude (5,969.02 m3 =s) is expected
indicate that the model was satisfactorily optimized. The optimized to rise by 32.50% in the projected period. Based on the current peak
SWAT model with fitted values of the considered parameters was magnitude and average monsoon runoff, 2021, 2023, 2025, 2026,
used to simulate runoff for 2015–2020 using CMCC-ESM2 meteoro- 2027, 2028, and 2031 are classified as extremely vulnerable (>25%
logical data sets and LULC 2020. Meteorological data sets for each increase) to hydrological extremes. The most recent catastrophic
SSP were used to estimate the runoff for this period in order to find flash flood occurred on October 4, 2023. This highlights the tem-
the best-fitted SSP for predicting future runoff for the Ranikhola poral applicability of the current analysis. Because the watershed
watershed for 2015–2020 (Figs. S5 and S6). The statistical matrices is one of the most susceptible to flash floods and landslides in
of simulated runoff versus observed values computed (Table 12) at northeastern India, future runoff estimation helps identify and
the outlet from 2015 to 2020 were used to calibrate the available SSPs understand potential changes in flood patterns. Assessment of pro-
for daily discharge. jected runoff amounts aids in comprehending the availability and
Compared to the observed data, all scenarios overstated runoff distribution of water resources in the Ranikhola watershed. The
values according to the positive PBIAS values. However, SSP2-4.5
produced entirely satisfactory results in capturing temporal patterns
of observed runoff (Fig. 7). According to the statistical evaluation,
SSP2-4.5 had a higher R2 and NSE with lower RSR values, making Table 12. Statistical evaluation for each scenario of CMCC-ESM2 (Italy)
it possible to utilize it to simulate future runoff for the Ranikhola CMCC-ESM2 (Italy)
watershed. Performance
indicators SSP1-2.6 SSP2-4.5 SSP5-8.5
Future Runoff Prediction R2 0.24 0.60 0.58
The predicted LULC 2032 and the meteorological data sets of PBIAS 0.02 0.01 0.02
CMCC-ESM2 (Italy) SSP2-4.5 were employed to simulate future NSE −2.01 0.59 0.38
runoff for 2021–2032. In this context, the optimized SWAT model RSR 1.74 0.64 0.78
was appointed for runoff prediction on a daily scale, assuming Note: Bold value represents better values for all performance indicators.

© ASCE 05024006-13 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 simulate LULC dynamics was analyzed. The efficacy of the
model was entirely satisfactory in terms of the error criterion of
the model and the overall accuracy (>90%) of the simulated
LULC map. The increasing trend in urbanization (142%) and
the decreasing trend in forest cover (20%) from 2008 to 2032
is a critical factor in the watershed’s hydrological and environ-
mental vulnerabilities.
• The vulnerability of the watershed to potential landslides and
flash floods is expected to rise due to increasing urbanization
(3.04%) and decreasing forest cover (4.70%). According to this
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

study, high FFVZs for this high-altitude Himalayan watershed

are expected to rise by 9% over the next decade. Furthermore,
the predicted alteration in forest cover surrounding the LIPs
exposes the region (Gangtok, Singtam, Ranipool) to heightened
landslide susceptibility.
• The efficacy of using uncorrected CMCC-ESM2-SSP2-4.5 data
for runoff estimation carried out by optimized SWAT (i.e., IMD-
based) was entirely satisfactory in terms of capturing temporal
Fig. 7. Daily discharge observed and simulated using CMCC-ESM2- patterns of observed runoff and statistical evaluation (R2 ¼ 0.60
SSP2-4.5. and NSE ¼ 0.59). The significant finding of this analysis is
that, between 2021 and 2032, there is an expected increase
in both average monsoon runoff (from May to October) and
average monsoon runoff peak magnitude (32.50% and 26.00%,
respectively).
• According to this analysis, the years 2021, 2023, 2025, 2026,
2027, 2028, and 2031 are particularly sensitive to hydrolog-
ical extremes. The most recent catastrophic flash flood
occurred in the vicinity of and within the watershed on
October 4, 2023, confirming the temporal relevance of this
analysis.
The present research aimed to spatiotemporally quantify the
impacts of LULCC and subsequently drew the conclusion that
the Himalayan watersheds necessitate targeted hydrological
considerations. This study assumed that the physiological and
morphological characteristics were unchanged for the purpose
of delineating high FFVZs from 2020 to 2032. Furthermore,
the calibrated parameters value of the optimized hydrological
model during the observed period were assumed to be constant
for the predicted period. However, this study may serve as a
helpful benchmark to adopt better decisions and reframe
land use and water resource management policies within the
watersheds.
This study suggested that forest cover transition should be
avoided or prevented in the coming years for these locations.
Sustainable urban planning should be implemented, especially
around Gangtok, Ranipool, and Singtam. Restoration of the bio-
Fig. 8. Daily rainfall and runoff scenario from 2021 to 2032.
diversity and ecology of the watershed by afforestation should be
encouraged among the local people as beneficial to the study
area in terms of harmonizing climatic conditions and mitigating
creation of efficient flood management methods, such as the imple- the impact of hydrological extremes. The overall methodology
mentation of early warning systems, the construction of flood con- adopted for this study can be used for spatially and temporally
trol structures, and the execution of new land use policies, is made extensive analyses. Flash flood vulnerability maps can be gen-
possible by this analysis. This study demonstrates that the influence erated considering changes in components due to severe occur-
of the projected LULCC is one of the critical causes of an increased rences (e.g., rapid snow melt in the hilltops). Additional criteria
frequency of flash floods and landslides and rising monsoon sur- could be defined as spatial constraints for land use changes and
face runoff. parameters for SWAT modeling re-assessed based on the particu-
larities of the study locations. The calibration and validation of
the hydrological model were carried out for the Ranikhola water-
Conclusions shed with limited data. The performance of the model is expected
to improve with an increase in observed daily streamflow data.
This study makes the following conclusions: However, employing high-resolution spatial information enables
• The effectiveness of integrated GIS and high-resolution remote a precise representation of the temporal variability in runoff es-
sensing data sets with a hybrid model (MC-CA-MLP-NN) to timation and prediction.

© ASCE 05024006-14 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 Appendix. Statistical Evaluation of IMD-Based 1987–2022 and identifying areas at risk of flooding due to the impact of
Runoff Estimation urbanization in Lao Cai city.” Ecol. Inf. 72 (Jul): 101912. https://doi.org
/10.1016/j.ecoinf.2022.101912.
SWAT Arnold, J., R. Srinivasan, C. Santhi, R. D. Harmel, and A. Van Griensven.
Performance 2012. “SWAT: Model use, calibration, and validation.” Trans. ASABE
indicators Calibration Validation 55 (4): 1491–1508. https://doi.org/10.13031/2013.42256.
R2 0.664 0.817 Arnold, J. G., R. Srinivasan, R. S. Muttiah, and J. R. Williams. 1998.
PBIAS 0.001 0.011 “Large area hydrologic modeling and assessment Part I: Model devel-
NSE 0.648 0.803 opment’ basin scale model called SWAT (soil and water speed and stor-
RSR 0.593 0.444 age, advanced software debugging policy to meet the needs, and the
management to the tank model.” JAWRA J. Am. Water Resour. Assoc.
Note: Bold represents higher values.
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

34 (1): 73–89. https://doi.org/10.1111/j.1752-1688.1998.tb05961.x.

Astuti, I. S., K. Sahoo, A. Milewski, and D. R. Mishra. 2019. “Impact of
land use land cover (LULC) change on surface runoff in an increas-
Data Availability Statement ingly urbanized tropical watershed.” Water Resour. Manage. 33 (12):
4087–4103. https://doi.org/10.1007/s11269-019-02320-w.
Some or all data, models, or code that support the findings of this Bewket, W. 2002. “Land cover dynamics since the 1950s in Chemoga
study are available from the corresponding author upon reasonable watershed, Blue Nile basin, Ethiopia.” Mt. Res. Dev. 22 (3): 263–269.
request. https://doi.org/10.1659/0276-4741(2002)022[0263:LCDSTI]2.0.CO;2.
Bhattacharjee, K., and B. Behera. 2017. “Forest cover change and flood
hazards in India.” Land Use Policy 67 (Dec): 436–448. https://doi.org
Acknowledgments /10.1016/j.landusepol.2017.06.013.
Chalise, S. R., S. R. Kansakar, G. Rees, K. Croker, and M. Zaidman. 2003.
We are thankful to the CWC, Sikkim Investigation Division, “Management of water resources and low flow estimation for the
government of India, for providing daily observed discharge data Himalayan basins of Nepal.” J. Hydrol. 282 (1): 25–35. https://doi.org
(2015–2020). In addition, we are appreciative of Copernicus C3S /10.1016/S0022-1694(03)00250-6.
for making their open data archives freely accessible data sets avail- Chandan, M. C., G. Nimish, and H. A. Bharath. 2020. “Analysing spatial
able. We are thankful to the NRSC in Hyderabad, India, for pro- patterns and trend of future urban expansion using SLEUTH.” Spatial
viding the high-resolution LISS-IV data. We are also grateful to Inf. Res. 28 (1): 11–23. https://doi.org/10.1007/s41324-019-00262-4.
the ASF, SoI, IMD, the Centre for Interdisciplinary Studies of Chopard, B. 2009. “Cellular automata modeling of physical systems.”
Mountain and Hill Environment, University of Delhi, India, Envir- In Encyclopedia complexity system science, edited by R. A. Meyers,
865–892. New York: Springer.
onmental Information System (ENVIS) Centre, the Ministry of
Clarke, K. C., S. Hoppen, and L. Gaydos. 1997. “A self-modifying cellular
Environment and Forest, and the government of India for their automaton model of historical urbanization in the San Francisco Bay
freely available data sets in their open data archives. area.” Environ. Plann. B: Plann. Des. 24 (2): 247–261. https://doi.org
Author contributions: Dibyandu Roy: conceptualization, metho- /10.1068/b240247.
dology, investigation, software, visualization, and writing–original Cunha, E. R., C. A. G. Santos, R. M. da Silva, V. M. Bacani, and A. Pott.
draft. Bharath H. Aithal: supervision, conceptualization, visualiza- 2021. “Future scenarios based on a CA-Markov land use and land cover
tion, and writing. Anirban Dhar: supervision, conceptualization, and simulation model for a tropical humid basin in the Cerrado/Atlantic
visualization. Venkappayya R. Desai: data and software acquisition forest ecotone of Brazil.” Land Use Policy 101 (Feb): 105141. https://
and visualization. doi.org/10.1016/j.landusepol.2020.105141.
Daide, F., R. Afgane, A. Lahrach, and A. A. Chaouni. 2022. “Beht water-
shed (Morocco) rainfall-runoff simulation with the HEC-HMS hydro-
Supplemental Materials logical model.” Ecol. Eng. Environ. Technol. 23 (1): 142–155. https://
doi.org/10.12912/27197050/142942.
Tables S1–S6 and Figs. S1–S6 are available online in the ASCE Dale, V. H. 1997. “The relationship between land-use change and climate
Library (www.ascelibrary.org). change.” Ecol. Appl. 7 (3): 753–769. https://doi.org/10.1890/1051-0761
(1997)007[0753:TRBLUC]2.0.CO;2.
Dale, V. H., R. A. Efroymson, and K. L. Kline. 2011. “The land use–climate
References change–energy nexus.” Landscape Ecol. 26 (Jul): 755–773. https://doi
.org/10.1007/s10980-011-9606-2.
Abbaszadeh, M., O. Bazrafshan, R. Mahdavi, E. R. Sardooi, and S. Dey, A., R. Remesan, and R. Kumar. 2023. “Blue and green water re-
Jamshidi. 2023. “Modeling future hydrological characteristics based on distribution dependency on precipitation datasets for a tropical Indian
land use/land cover and climate changes using the SWAT model.” Water River basin.” J. Hydrol.: Reg. Stud. 46 (Apr): 101361. https://doi.org/10
Resour. Manage. 37 (10): 4177–4194. https://doi.org/10.1007/s11269 .1016/j.ejrh.2023.101361.
-023-03545-6. Dey, N. N., A. Al Rakib, A. Al Kafy, and V. Raikwar. 2021. “Geospatial
Abdulkareem, J. H., W. N. A. Sulaiman, B. Pradhan, and N. R. Jamil. 2018. modelling of changes in land use/land cover dynamics using multi-
“Relationship between design floods and land use land cover (LULC) layer perceptron Markov chain model in Rajshahi City, Bangladesh.”
changes in a tropical complex catchment.” Arabian J. Geosci. 11 (14): Environ. Challenges 4 (Aug): 100148. https://doi.org/10.1016/j.envc
376. https://doi.org/10.1007/s12517-018-3702-4. .2021.100148.
Aburas, M. M., Y. M. Ho, M. F. Ramli, and Z. H. Ash’aari. 2016. “The Din, A. U., W. Basharat, A. Malik, and R. H. Gavit. 2023. “Dynamics
simulation and prediction of spatio-temporal urban growth trends using of land use and land cover in Northern India: A systematic review.”
cellular automata models: A review.” Int. J. Appl. Earth Obs. Geoinf. GeoJournal 88 (Mar): 4297–4324. https://doi.org/10.1007/s10708-023
52 (Oct): 380–389. https://doi.org/10.1016/j.jag.2016.07.007. -10870-1.
Aithal, B. H., M. C. Chandan, S. Vinay, and T. V. Ramachandra. 2018. Dinka, M. O., and A. Klik. 2019. “Effect of land use–land cover change on
“Urbanization in India: Patterns, visualization of cities, and greenhouse the regimes of surface runoff—The case of Lake Basaka catchment
gas inventory for developing an urban observatory.” In Urban remote (Ethiopia).” Environ. Monit. Assess. 191 (5): 1–13. https://doi.org/10
sensing, 151–172. Boca Raton, FL: CRC Press. .1007/s10661-019-7439-7.
Anh, T., T. Do, A. Ngoc, T. Do, and H. Duc. 2022. “Ecological informatics Erdbrügger, J., I. van Meerveld, K. Bishop, and J. Seibert. 2021. “Effect
quantifying the spatial pattern of urban expansion trends in the period of DEM-smoothing and-aggregation on topographically-based flow

© ASCE 05024006-15 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 directions and catchment boundaries.” J. Hydrol. 602 (Nov): 126717. Hyandye, C. 2015. “GIS and logit regression model applications in land
https://doi.org/10.1016/j.jhydrol.2021.126717. use/land cover change and distribution in Usangu catchment.” Am. J.
Faisal Koko, A., W. Yue, G. Abdullahi Abubakar, R. Hamed, and A. A. Remote Sens. 3 (1): 6. https://doi.org/10.11648/j.ajrs.20150301.12.
Noman Alabsi. 2021. “Analyzing urban growth and land cover change Hyandye, C., and L. W. Martz. 2017. “A Markovian and cellular automata
scenario in Lagos, Nigeria using multi-temporal remote sensing data land-use change predictive model of the Usangu catchment.” Int. J.
and GIS to mitigate flooding.” Geomatics Nat. Hazards Risk 12 (1): Remote Sens. 38 (1): 64–81. https://doi.org/10.1080/01431161.2016
631–652. https://doi.org/10.1080/19475705.2021.1887940. .1259675.
Femeena, P. V., R. Karki, R. Cibin, and K. P. Sudheer. 2022. “Reconceptu- Islam, M. D., K. S. Islam, R. Ahasan, M. R. Mia, and M. E. Haque. 2021.
alizing HRU threshold definition in the soil and water assessment tool.” “A data-driven machine learning-based approach for urban land cover
J. Am. Water Resour. Assoc. 58 (4): 508–516. https://doi.org/10.1111 change modeling: A case of Khulna city corporation area.” Remote
/1752-1688.13000. Sens. Appl.: Soc. Environ. 24 (Sep): 100634. https://doi.org/10.1016/j
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Feng, B., Y. Zhang, and R. Bourke. 2021. “Urbanization impacts on flood .rsase.2021.100634.
risks based on urban growth data and coupled flood models.” Nat. Haz- Kafy, A., N. N. Dey, A. Al Rakib, Z. A. Rahaman, N. M. R. Nasher, and A.
ards 106 (1): 613–627. https://doi.org/10.1007/s11069-020-04480-0. Bhatt. 2021a. “Modeling the relationship between land use/land cover
Foody, G. M. 2020. “Explaining the unsuitability of the kappa coefficient and land surface temperature in Dhaka, Bangladesh using CA-ANN
in the assessment and comparison of the accuracy of thematic maps algorithm.” Environ. Challenges 4 (Jun): 100190. https://doi.org/10
obtained by image classification.” Remote Sens. Environ. 239 (Jan): .1016/j.envc.2021.100190.
111630. https://doi.org/10.1016/j.rse.2019.111630. Kafy, A., M. N. H. Naim, G. Subramanyam, A. Al Faisal, N. U. Ahmed,
Geetha Raveendran Nair, A. N., S. D. Shamsudeen, M. G. Mohan, and A. A. Al Rakib, M. A. Kona, and G. S. Sattar. 2021b. “Cellular automata
Sankaran. 2023. “Basin-scale streamflow projections for greater Pamba approach in dynamic modelling of land cover changes using RapidEye
river basin, India integrating GCM ensemble modelling and flow images in Dhaka, Bangladesh.” Environ. Challenges 4 (Mar): 100084.
accumulation-weighted LULC overlay in deep learning environment.” https://doi.org/10.1016/j.envc.2021.100084.
Sustainable 15 (19): 1–27. https://doi.org/10.3390/su151914148. Kozak, J., U. Gimmi, T. Houet, and J. Bolliger. 2017. “Current prac-
Gharaibeh, A., A. Shaamala, R. Obeidat, and S. Al-Kofahi. 2020. “Im- tices and challenges for modelling past and future land use and land
proving land-use change modeling by integrating ANN with cellular cover changes in mountainous regions.” Reg. Environ. Change 17 (8):
automata-Markov chain model.” Heliyon 6 (9): e05092. https://doi 2187–2191. https://doi.org/10.1007/s10113-017-1217-2.
.org/10.1016/j.heliyon.2020.e05092. Kumar, M., D. M. Denis, A. Kundu, N. Joshi, and S. Suryavanshi. 2022.
Gidey, E., O. Dikinya, R. Sebego, E. Segosebe, and A. Zenebe. 2017. “Understanding land use/land cover and climate change impacts on
hydrological components of Usri watershed, India.” Appl. Water Sci.
“Cellular automata and Markov chain (CA_Markov) model-based
12 (3): 1–14. https://doi.org/10.1007/s13201-021-01547-6.
predictions of future land use and land cover scenarios (2015–2033)
Lehmann, P., J. von Ruette, and D. Or. 2019. “Deforestation effects on
in Raya, northern Ethiopia.” Model. Earth Syst. Environ. 3 (4):
rainfall-induced shallow landslides: Remote sensing and physically-
1245–1262. https://doi.org/10.1007/s40808-017-0397-6.
based modeling.” Water Resour. Res. 55 (11): 9962–9976. https://doi
Gogoi, P. P., V. Vinoj, D. Swain, G. Roberts, J. Dash, and S. Tripathy. 2019.
.org/10.1029/2019WR025233.
“Land use and land cover change effect on surface temperature over
Li, G., Y. Lei, H. Yao, S. Wu, and J. Ge. 2017. “The influence of land
eastern India.” Sci. Rep. 9 (1): 1–10. https://doi.org/10.1038/s41598
urbanization on landslides: An empirical estimation based on Chinese
-019-45213-z.
provincial panel data.” Sci. Total Environ. 595 (Oct): 681–690. https://
GSI (Geological Survey of India). 2023. “Bhukosh is a gateway to all geo-
doi.org/10.1016/j.scitotenv.2017.03.258.
scientific data.” Accessed February 1, 2024. https://bhukosh.gsi.gov.in.
Li, S. H., B. X. Jin, X. Y. Wei, Y. Y. Jiang, and J. L. Wang. 2015. “Using
Guan, D., H. Li, T. Inohae, W. Su, T. Nagaie, and K. Hokao. 2011.
CA-MARKOV model to model the spatiotemporal change of land use/
“Modeling urban land use change by the integration of cellular automa- cover in Fuxian Lake for decision support.” ISPRS Ann. Photogramm.
ton and Markov model.” Ecol. Modell. 222 (20): 3761–3772. https://doi Remote Sens. Spatial Inf. Sci. 2 (4): 163–168. https://doi.org/10.5194
.org/10.1016/j.ecolmodel.2011.09.009. /isprsannals-II-4-W2-163-2015.
Guo, L., L. Di, C. Zhang, L. Lin, F. Chen, and A. Molla. 2022. “Evaluating Liu, J., Z. Wu, and H. Zhang. 2021. “Analysis of changes in landslide sus-
contributions of urbanization and global climate change to urban land ceptibility according to land use over 38 years in Lixian county, China.”
surface temperature change: A case study in Lagos, Nigeria.” Sci. Rep. Sustainability 13 (19): 1–23. https://doi.org/10.3390/su131910858.
12 (1): 1–11. https://doi.org/10.1038/s41598-022-18193-w. Lovino, M. A., M. J. Pierrestegui, O. V. Müller, E. H. Berbery, G. V. Müller,
Hannah, D. M., S. R. Kansakar, A. J. Gerrard, and G. Rees. 2005. and M. Pasten. 2021. “Evaluation of historical CMIP6 model simula-
“Flow regimes of Himalayan rivers of Nepal: Nature and spatial pat- tions and future projections of temperature and precipitation in
terns.” J. Hydrol. 308 (1): 18–32. https://doi.org/10.1016/j.jhydrol.2004 Paraguay.” Clim. Change 164: 1–24.
.10.018. Maher, M., T. Nasrallah, M. Rabah, and F. Abdelhaleem. 2022. “Watershed
Harik, G., I. Alameddine, M. A. Najm, and M. El-Fadel. 2023. “Modified delineation and runoff estimation of Wadi Tayyibah, South Sinai using
SWAT to forecast water availability in mediterranean mountainous Arc-GIS and HEC-HMS model.” Port-Said Eng. Res. J. 26 (2): 61–71.
watersheds with snowmelt dominated runoff.” Water Resour. Manage. https://doi.org/10.21608/pserj.2022.111930.1155.
37 (5): 1985–2000. https://doi.org/10.1007/s11269-023-03466-4. Mansour, S., M. Al-Belushi, and T. Al-Awadhi. 2020. “Monitoring land use
Hassan, Z., R. Shabbir, S. S. Ahmad, A. H. Malik, N. Aziz, A. Butt, and S. and land cover changes in the mountainous cities of Oman using GIS
Erum. 2016. “Dynamics of land use and land cover change (LULCC) and CA-Markov modelling techniques.” Land Use Policy 91 (Jun):
using geospatial techniques: A case study of Islamabad Pakistan.” 104414. https://doi.org/10.1016/j.landusepol.2019.104414.
Springerplus 5 (1): 1–11. https://doi.org/10.1186/s40064-016-2414-z. Mas, J. F., M. Kolb, M. Paegelow, M. T. Camacho Olmedo, and T. Houet.
Her, Y., J. Frankenberger, I. Chaubey, and R. Srinivasan. 2015. “Threshold 2014. “Inductive pattern-based land use/cover change models: A
effects in HRU definition of the soil and water assessment tool.” Trans. comparison of four software packages.” Environ. Modell. Software
ASABE 58 (2): 367–378. https://doi.org/10.13031/trans.58.10805. 51 (Sep): 94–111. https://doi.org/10.1016/j.envsoft.2013.09.010.
Hertel, T. W., S. K. Rose, and R. S. J. Tol. 2009. Economic analysis of land Maurya, S., P. K. Srivastava, L. Zhuo, A. Yaduvanshi, and R. K. Mall.
use in global climate change policy. Oxfordshire, UK: Routledge. 2023. “Future climate change impact on the streamflow of Mahi River
Hosseini, S. H., and M. R. Khaleghi. 2020. “Application of SWAT model basin under different general circulation model scenarios.” Water
and SWAT-CUP software in simulation and analysis of sediment uncer- Resour. Manage. 2023 (Jan): 1–22. https://doi.org/10.1007/s11269-022
tainty in arid and semi-arid watersheds (case study: The Zoshk– -03372-1.
Abardeh watershed).” Model. Earth Syst. Environ. 6 (4): 2003–2013. Meneses, B. M., S. Pereira, and E. Reis. 2019. “Effects of different land
https://doi.org/10.1007/s40808-020-00846-2. use and land cover data on the landslide susceptibility zonation of road

© ASCE 05024006-16 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 networks.” Nat. Hazards Earth Syst. Sci. 19 (3): 471–487. https://doi Roy, D., A. Dhar, and V. R. Desai. 2023. “A grey fuzzy analytic hierarchy
.org/10.5194/nhess-19-471-2019. process-based flash flood vulnerability assessment in an ungauged
Merz, J., P. M. Dangol, M. P. Dhakal, B. S. Dongol, G. Nakarmi, and R. Himalayan watershed.” Environ. Dev. Sustainability 2023 (Jun): 1–26.
Weingartner. 2006. “Rainfall-runoff events in a middle mountain catch- https://doi.org/10.1007/s10668-023-03385-9.
ment of Nepal.” J. Hydrol. 331 (3): 446–458. https://doi.org/10.1016/j Roy, J., and S. Saha. 2022. “Ensemble hybrid machine learning methods
.jhydrol.2006.05.030. for gully erosion susceptibility mapping: K-fold cross validation ap-
Mesgari, E., S. A. Hosseini, M. S. Hemmesy, M. Houshyar, and L. G. proach.” Artif. Intell. Geosci. 3 (Mar): 28–45. https://doi.org/10.1016/j
Partoo. 2022. “Assessment of CMIP6 models’ performances and pro- .aiig.2022.07.001.
jection of precipitation based on SSP scenarios over the MENAP re- Saha, P., R. Mitra, K. Chakraborty, and M. Roy. 2022. “Application of
gion.” J. Water Clim. Change 13 (10): 3607–3619. multi layer perceptron neural network Markov chain model for LULC
Mishra, V. N., and P. K. Rai. 2016. “A remote sensing aided multi-layer change detection in the sub-Himalayan north Bengal.” Remote Sens.
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

perceptron-Markov chain analysis for land use and land cover change Appl.: Soc. Environ. 26 (Nov): 100730. https://doi.org/10.1016/j.rsase
prediction in Patna district (Bihar), India.” Arabian J. Geosci. 9 (4): .2022.100730.
1–18. https://doi.org/10.1007/s12517-015-2138-3. Sajikumar, N., and R. S. Remya. 2015. “Impact of land cover and land
Mondal, M. S., N. Sharma, P. K. Garg, and M. Kappas. 2016. “Statistical use change on runoff characteristics.” J. Environ. Manage. 161 (Sep):
independence test and validation of CA Markov land use land cover 460–468. https://doi.org/10.1016/j.jenvman.2014.12.041.
(LULC) prediction results.” Egypt. J. Remote Sens. Space Sci. 19 (2): Sharma, P., and S. C. Rai. 2004. “Streamflow, sediment and carbon trans-
259–272. https://doi.org/10.1016/j.ejrs.2016.08.001. port from a Himalayan watershed.” J. Hydrol. 289 (1): 190–203. https://
National Remote Sensing Center. 2023. “Bhuvan: Indian Geo Platform of doi.org/10.1016/j.jhydrol.2003.11.023.
ISRO.” Accessed October 10, 2023. https://bhuvan-app1.nrsc.gov.in. Shi, L., P. Feng, B. Wang, D. L. Liu, H. Zhang, J. Liu, and Q. Yu. 2022.
Neitsch, S., J. Arnold, J. Kiniry, and J. Williams. 2011. Soil & water assess- “Assessing future runoff changes with different potential evapotranspi-
ment tool theoretical documentation version 2009, 1–647. College ration inputs based on multi-model ensemble of CMIP5 projections.”
Station, TX: Texas Water Resources Institute. J. Hydrol. 612 (Aug): 128042. https://doi.org/10.1016/j.jhydrol.2022
Ngoma, H., W. Wen, B. Ayugi, H. Babaousmail, R. Karim, and V. Ongoma. .128042.
2021. “Evaluation of precipitation simulations in CMIP6 models over Shukla, S., S. K. Jain, and M. L. Kansal. 2021. “Hydrological modelling
Uganda.” Int. J. Clim. 41 (9): 4743–4768. of a snow/glacier-fed western Himalayan basin to simulate the current
and future streamflows under changing climate scenarios.” Sci. Total
Pai, D. S., L. Sridhar, M. Rajeevan, O. P. Sreejith, N. S. Satbhai, and B.
Environ. 795 (Nov): 148871. https://doi.org/10.1016/j.scitotenv.2021
Mukhopadhyay. 2014. “Development of a new high spatial resolution
.148871.
(0.25° × 0.25°) long period (1901–2010) daily gridded rainfall data set
Singh, V., N. Bankar, S. S. Salunkhe, A. K. Bera, and J. R. Sharma. 2013.
over India and its comparison with existing data sets over the region.”
“Hydrological stream flow modelling on tungabhadra catchment:
Mausam 65 (1): 1–18. https://doi.org/10.54302/mausam.v65i1.851.
Parameterization and uncertainty analysis using SWAT CUP.” Curr. Sci.
Parsamehr, K., M. Gholamalifard, and Y. Kooch. 2020. “Comparing three
104 (9): 1187–1199.
transition potential modeling for identifying suitable sites for REDD+
Sisodia, P. S., V. Tiwari, and A. Kumar. 2014. “Analysis of supervised
projects.” Spatial Inf. Res. 28 (2): 159–171. https://doi.org/10.1007
maximum likelihood classification for remote sensing image.” In
/s41324-019-00273-1.
Proc., Int. Conf. on Recent Advances and Innovations in Engineering
Patil, N. S., and M. Nataraja. 2020. “Effect of land use land cover changes (ICRAIE-2014), 9–12. New York: IEEE.
on runoff using hydrological model: A case study in Hiranyakeshi
Slack, R. B., and R. Welch. 1980. “Soil conservation service runoff
watershed.” Model. Earth Syst. Environ. 6 (4): 2345–2357. https://doi curve number estimates from Landsat data.” JAWRA J. Am. Water
.org/10.1007/s40808-020-00808-8. Resour. Assoc. 16 (5): 887–893. https://doi.org/10.1111/j.1752-1688
Pisano, L., V. Zumpano, M. C. M. Rosskopf, and M. Parise. 2017. .1980.tb02504.x.
“Variations in the susceptibility to landslides, as a consequence of land Soheili Majd, M., E. Simonetto, and L. Polidori. 2012. “Maximum
cover changes: A look to the past, and another towards the future.” Sci. likelihood classification of high-resolution SAR images in urban area.”
Total Environ. 601–602 (Dec): 1147–1159. https://doi.org/10.1016/j Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. 44 (Jun): 309–312.
.scitotenv.2017.05.231. https://doi.org/10.5194/isprsarchives-xxxviii-4-w19-309-2011.
Prasad, A. S., B. W. Pandey, W. Leimgruber, and R. M. Kunwar. 2016. Souza-e-Silva, H., W. Savino, R. A. Feijóo, and A. T. R. Vasconcelos.
“Mountain hazard susceptibility and livelihood security in the upper 2009. “A cellular automata-based mathematical model for thymocyte
catchment area of the river Beas, Kullu Valley, Himachal Pradesh, development.” PLoS One 4 (12): e8233. https://doi.org/10.1371/journal
India.” Geoenviron. Disasters 3 (1): 1–17. https://doi.org/10.1186 .pone.0008233.
/s40677-016-0037-x. Srinivasan, R., X. Zhang, and J. Arnold. 2010. “SWAT ungauged: Hydro-
Puente-Sotomayor, F., A. Mustafa, and J. Teller. 2021. “Landslide sus- logical budget and crop yield prediction in the upper Mississippi river
ceptibility mapping of urban areas: Logistic regression and sensitivity basin.” Trans. ASABE 53 (5): 1533–1546. https://doi.org/10.13031
analysis applied to Quito, Ecuador.” Geoenviron. Disasters 8 (1): 19. /2013.34903.
https://doi.org/10.1186/s40677-021-00184-0. Strahler, A. N. 1964. “Quantitative geomorphology of drainage basins
Quevedo, P., A. Velastegui-Montoya, N. Montalván-Burbano, F. Morante- and channel networks.” In Handbook of applied hydrology, 4–39.
Carballo, O. Korup, and C. Daleles Rennó. 2023. “Land use and land New York: McGraw-Hill.
cover as a conditioning factor in landslide susceptibility: A literature Subedi, P., K. Subedi, and B. Thapa. 2013. “Application of a hybrid cellular
review.” Landslides 20 (5): 967–982. https://doi.org/10.1007/s10346 automaton–Markov (CA-Markov) model in land-use change prediction:
-022-02020-4. A case study of saddle creek drainage basin, Florida.” Appl. Ecol. En-
Rahman, M. S., M. S. Anower, M. R. Hasan, M. B. Hossain, and M. I. viron. Sci. 1 (6): 126–132. https://doi.org/10.12691/aees-1-6-5.
Haque. 2017. “Design and numerical analysis of highly sensitive Au- Sun, Z., T. Lotz, and Q. Huang. 2021. “An ET-based two-phase method
MoS2 -graphene based hybrid surface plasmon resonance biosensor.” for the calibration and application of distributed hydrological models.”
Opt. Commun. 396 (Jan): 36–43. https://doi.org/10.1016/j.optcom.2017 Water Resour. Manage. 35 (3): 1065–1077. https://doi.org/10.1007
.03.035. /s11269-021-02774-x.
Rocha, J., J. C. Ferreira, J. Simões, and J. A. Tenedório. 2007. “Modelling Swain, S., S. K. Mishra, A. Pandey, A. C. Pandey, A. Jain, and S. K. Chauhan.
coastal and land use evolution patterns through neural network and 2022. “Hydrological modelling through SWAT over a Himalayan catch-
cellular automata integration.” J. Coastal Res. 2007 (50): 827–831. ment using high-resolution geospatial inputs.” Environ. Challenges
Rohan, T., E. Shelef, B. Mirus, and T. Coleman. 2023. “Prolonged influ- 8 (Jun): 100579. https://doi.org/10.1016/j.envc.2022.100579.
ence of urbanization on landslide susceptibility.” Landslides 20 (Mar): Tayal Senzeba, K., A. Bhadra, and A. Bandyopadhyay. 2015. “Snowmelt
1433–1447. https://doi.org/10.1007/s10346-023-02050-6. runoff modelling in data scarce Nuranang catchment of eastern

© ASCE 05024006-17 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006

 Himalayan region.” Remote Sens. Appl. Soc. Environ. 1 (5): 20–35. Wang, S. W., L. Munkhnasan, and W. K. Lee. 2021. “Land use and land
https://doi.org/https://doi.org/10.1016/j.rsase.2015.06.001. cover change detection and prediction in Bhutan’s high altitude city of
Teng, H., S. Chen, B. Hu, and Z. Shi. 2023. “Future changes and driving Thimphu, using cellular automata and Markov Chain.” Environ. Chal-
factors of global peak vegetation growth based on CMIP6 simulations.” lenges 2 (Dec): 100017. https://doi.org/10.1016/j.envc.2020.100017.
Ecol. Inf. 75 (Nov): 102031. https://doi.org/10.1016/j.ecoinf.2023.102031. Xu, Y., Y. Chen, Y. Ren, Z. Tang, X. Yang, and Y. Zhang. 2023. “Attribu-
Thavhana, M. P., M. J. Savage, and M. E. Moeletsi. 2018. “SWAT model tion of streamflow changes considering spatial contributions and driver
uncertainty analysis, calibration and validation for runoff simulation interactions based on hydrological modeling.” Water Resour. Manage.
in the Luvuvhu river catchment, South Africa.” Phys. Chem. Earth 37 (5): 1859–1877. https://doi.org/10.1007/s11269-023-03459-3.
105 (Jan): 115–124. https://doi.org/10.1016/j.pce.2018.03.012. Xue, P., C. Zhang, Z. Wen, E. Park, and H. Jakada. 2022. “Climate vari-
The Hindu. 2023. “Sikkim flash floods: Eight soldiers killed; search on for
ability impacts on runoff projection under quantile mapping bias cor-
14 missing.” October 7, 2023.
rection in the support CMIP6: An investigation in Lushi basin of
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 02/06/25. Copyright ASCE. For personal use only; all rights reserved.

Tola, S. Y., and A. Shetty. 2021. “Land cover change and its implication to
China.” J. Hydrol. 614 (5): 128550. https://doi.org/10.1016/j.jhydrol
hydrological regimes and soil erosion in Awash River basin, Ethiopia:
A systematic review.” Environ. Monit. Assess. 193 (12): 1–19. https:// .2022.128550.
doi.org/10.1007/s10661-021-09599-6. Yadav, B. P., R. Saxena, A. K. Das, S. K. Manik, and S. K. Asok Raja.
Twisa, S., and M. F. Buchroithner. 2019. “Land-use and land-cover (LULC) 2017. Rainfall statistics of India-2017. New Delhi, India: Hydromet Di-
change detection in Wami River basin, Tanzania.” Land 8 (9): 136. vision, Indian Meteorological Department, Ministry of Earth Sciences.
https://doi.org/10.3390/land8090136. Zhao, Y., D. Xie, X. Zhang, and S. Ma. 2021. “Integrating spatial Markov
Veh, G., O. Korup, S. von Specht, S. Roessner, and A. Walz. 2019. chains and geographically weighted regression-based cellular autom-
“Unchanged frequency of moraine-dammed glacial lake outburst floods ata to simulate urban agglomeration growth: A case study of the
in the Himalaya.” Nat. Clim. Change 9 (5): 379–383. https://doi.org/10 Guangdong–Hong Kong–Macao greater bay area.” Land 10 (6): 633.
.1038/s41558-019-0437-5. https://doi.org/10.3390/land10060633.

© ASCE 05024006-18 J. Hydrol. Eng.

J. Hydrol. Eng., 2024, 29(3): 05024006