Environmental Challenges 2 (2021) 100017

Contents lists available at ScienceDirect

Environmental Challenges
journal homepage: www.elsevier.com/locate/envc

Land use and land cover change detection and prediction in Bhutan’s high
altitude city of Thimphu, using cellular automata and Markov chain
Sonam Wangyel Wang a, Lamchin Munkhnasan b, Woo-Kyun Lee b,∗
a
Ojeong Eco-Resilience Institute (OJERI), Division of Environmental Science and Ecological Engineering, College of Life Sciences, Korea University, Seoul 02841, South
Korea
b
Division of Environmental Science and Ecological Engineering, College of Life Sciences, Korea University, Seoul 02841, South Korea

a r t i c l e i n f o a b s t r a c t

Keywords: Rapid urbanization is changing landscapes often resulting in the degradation of ecosystem services and quality
Detection of urban life. Remote sensing and GIS tools can provide valuable information to deepen our understanding of the
Prediction dynamics of these changes to better plan and build sustainable cities for the future. Using remote sensing data,
Thimphu
socio-economic data, and field observations, we simulated spatiotemporal dynamics of land use and land cover
Bhutan
changes in the city of Thimphu. Simulation results reveal that the landscape of Thimphu city has changed con-
Remote sensing
GIS siderably during the study period and the change trend is predicted to continue into 2050. The study observed a
CA-Markov model significant increase (12.77%) in built-up area from 2002 (52.88%) to 2018 (65.5%), followed by a slight increase
in the cover of bare ground. On the contrary, forest cover declined drastically (15.25%) followed by agriculture
(1.01%). Rapid population growth triggered by rural urban migration coupled with hasty socio-economic devel-
opment post democracy are the main drivers of these changes. These changes have fragmentated forest cover,
increased soil/gully erosion, surface runoff, and storm induced floods of storm and sanitation drains, thereby
impinging on the overall quality of life in the city. Under the business as usual scenario, prediction analysis for
the year 2050 show that built up area will consume almost all of the city area (73.21%) with forest significantly
reduced to patches making up only about 16% of the city. These findings beg for an urgent need to implement
effective planning specially to protect the existing forest and water resources from further degradation.

Introduction services and human well-being (Mallupattu and Sreennivasula, 2013;

Kanal, et al., 2019; Mishra, et al., 2019. Generally, cities in develop-
Global cities, which are the engines of economic development ing nations are characterized by poor infrastructural plans, high immi-
(Sherbinin, et al., 2007) with large populations, are most vulnerable gration rates, growing squatter settlements, etc. This demand address-
to the impacts of land use and land cover (LULC) changes. The LULC ing unique challenges and opportunities for urban adaptation and mit-
change is a driver of global environmental change such as emission of igation responses and for mainstreaming them into urban development
greenhouse gases (MEA, 2005), habitat loss/fragmentation and biodi- plans.
versity loss (Lambin, et al., 2000, Sala, et al., 2000), and reduce the Under such scenarios, studies using high resolution images and GIS
quality of human wellbeing (Lambin, 1997; Boissiere, et al., 2009). can provide scientifically reliable information for planning environmen-
Rapid increase in human population and associated economic devel- tal and economic development programs that are sensitive to achieving
opment further exacerbates the rate of these changes especially in fast social and environmental goals (Weng, 2001; Rogan and Chen, 2004;
growing urban areas (Liping, et al., 2018) threatening their sustain- Suresh, et al., 2011; Dutta, 2012; Rimal, et al. 2018). For instance,
able growth. According to the figures from the UN (2016), over 54.4% Addo (2010) asserted that remote sensing and GIS tools for mapping
of the world population in 2016 lived in cities and it is projected to LULC changes can offer critical policy recommendations for sustainable
increase to 60% by 2030 exerting extra pressure on urban resources. land management. A review of global literature on the use of remote
This increase is progressing at a higher rate in the developing countries sensing and GIS for detecting and predicting LULC changes revealed
where cities are growing thereby rapidly changing urban landscapes. that a significant amount research (Kalnay and Cai, 2003; Turner, et al.,
Such rapid growth in the absence of adequate plan and infrastructure 2007; Huang, et al., 2008; Islam and Ahmed, 2011; Hansen, et al., 2013;
expedites LULC changes which are associated with degrading ecosystem Taubenbock, et al., 2014; Hassan and Nazem, 2015; Appiah, et al., 2015;

∗
Corresponding author.
E-mail addresses: [email protected] (S.W. Wang), [email protected] (L. Munkhnasan), [email protected], [email protected] (W.-K. Lee).

https://doi.org/10.1016/j.envc.2020.100017
Received 23 November 2020; Received in revised form 21 December 2020; Accepted 21 December 2020
2667-0100/© 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

Liping, et al., 2018; Wang, et al., 2020) using remote sensing and GIS to the south to over 7000 m amsl are associated with varying climate, eco-
model LULC dynamics has accumulated. Geospatial modelling analysis logical zones, high biological diversity, and diverse livelihood systems
lies at the heart of these studies (Rogan and Chen, 2004; Coppin, et al., (Wang et al., 2019b). Like many other mountainous countries, despite
2004). These models attempt to detect where the change occurred or the cautious development Bhutan continues to face multiple threats to
will potentially occur (Veldkamp and Lambin, 2001). Most of these mod- its ecological and livelihood systems including from rapid urbanization
els use historical land use data to assess the past land transformation and LULC changes (Yangchen, et al., 2015; Wang et al., 2019a). Rapid
and transition, which in combination with environmental variables can population growth triggered by high immigration (14.5%) looking for
predict future land use scenarios (Eastman, 2009). Predicted land use better opportunities and basic facilities (Walcot, 2011) have resulted
changes exhibiting major modifications and alternations can help land in unprecedented LULC in Bhutan’s cities. Recent transition to democ-
use planners, resource managers, and conservation officers in promot- racy in 2008 ushered in expedited socio-economic development that is
ing sustainable management and mitigating negative impacts. Conse- adversely changing landscapes especially in urban cores. For instance
quently, detecting and predicting LULC changes have become an impor- a study by Yangchen, et al. (2015) in their analysis of Bhutan’s LULC
tant consideration in a variety of fields including, modelling rural and change from 2000-2013, reported that the built up area has increased
urban plans (Theobald and Hobbs, 1998; Kumar, et al., 2013), identify- from 6% in 2006 (pre-democratic Bhutan) to 14% in a matter of 7
ing biodiversity hotspot landscapes (Wang, et al., 2020) for advancing years (post democracy). In addition, Bruggeman, et al. (2016) attributed
conservation efforts, studying dynamics of desertification, etc. Bhutan’s recent forest cover loss to the recent socio-economic and po-
These studies have generated highly credible information on the litical changes. Existing literature (NSSC and PPD, 2011; Hansen, et al.,
state of land use changes and their drivers, which can inform better 2013; Gilani, et al., 2015; Yangchen, et al., 2015; Bruggeman, et al.,
decision making for sustainable planning of cities (Lu, et al., 2004). 2016; Sharma, et al., 2016; FRMD, 2017; Wangda, 2017) collectively
Therefore, an accurate geo-spatial modelling of urbanization to under- confirm that the country has experienced rapid and increasingly pro-
pin the fundamentals of location, characteristics, and consequences of nounced LULC changes over the last 40 years. The studies also attributed
urban expansion is a pre-requisite for sustainable development of cities. major land use changes to a suite of drivers including expansion of urban
Geospatial analysis use various statistical and rule based modelling ap- cores.
proaches for detecting and predicting land use changes (Overmars, et al., The impact of LULC changes are more serious in the cities than other
2003; Lu, et al., 2004). Commonly used models in estimating land use areas mainly due to pressures from population, developmental pro-
changes are; statistical models (Hyandye, 2015), evolutionary models grams, and rate of urbanization (Gilani, et al., 2015). Such trends, if not
(Aitkenhead and Alders, 2009), cellular models (Singh, et al., 2015), mitigated could lead to irreversible changes in landscapes with serious
Markov models (Yang, et al., 2012), hybrid models (Subedi, et al., 2013), negative impacts on human wellbeing (settlement, livelihoods, health,
expert system models (Stefanov, et al., 2001), and multi-agent models etc), as well as natural resources and ecosystem services (MRC, 2010).
(Ralha, et al., 2013). Of these the most popularly used are the cellu- Yet, no study has attempted to simulate the dynamics of LULC changes
lar and Markov chain analysis and their hybrid model called the CA- for Bhutan’s rapidly growing cities, let alone predict future scenarios.
Markov model (Zhao and Peng, 2012; Sohl and Claggett, 2013). The This situation demands a deepened understanding of the LULC dynam-
Markov chain analysis is a random stochastic modeling approach that ics to identify technical and policy options that can guide planners and
is discrete in both time and state. The model defines the LULC tran- policy makers to build more sustainable cities in Bhutan. As the capi-
sition from one time (t1 ) to the next (t2 ) to predict future change at a tal city, Thimphu is the center of socio-economic and political activities
time (Sherbinin, et al., 2007) to project probabilities of land use changes in the country and presents itself as a typical representative of other
for the future. This is achieved by using the known probabilities from cities in Bhutan. The city is located between 2,347 m amsl to 2,438 m
the past changes to predict future changes. This is based on the as- amsl (Walcott, 2011) and like many others, this small mountain city
sumption that the probability of a system being in a certain state at is experiencing rapid urbanization that is changing its landscape and
certain time can be determined if its state at an earlier time is known degrading its social and ecological resilience. Rapid population growth
(Lambin, 1997). While Markov analysis has been used to simulate and (triggered by high immigration (14.5%) and the government’s directive
predict land use changes (Muller and Middleton, 1994), it is best used for the city to expand and build infrastructures topped with growing
for short term projections (Lambin, 1997; Sinha and Kumar, 2013) as its tourist numbers, and expat workers), and associated rise in vehicle num-
analysis is not spatially explicit (Sklar and Costanza, 1991). In addition, bers are exerting increasing the rate of land use changes. Consequently,
it does not consider spatial information allocation within each class and the city is grappling with classic problems of growing slums, ecosystem
the probabilities of change between landscape states are not constant. degradation, insufficient basic facilities including drinking water, and
Hence, it can offer the right magnitude, but not the right direction of increasing frequency of disasters such as landslides, forest fires, flash
change. This short coming is mitigated by creating a superior CA-Markov floods, etc. (Wang et al., 2019a). Despite the growing challenges faced
model that combine the Markov model with a more dynamic and empir- by the city, its planners, and the inhabitants, no studies have attempted
ical cellular automata model (Subedi, et al., 2013). Cellular automata to understand the dynamics of LULC changes and their drivers. Hence
is a bottom up dynamic model that incorporates the spatial dimension this study represents the first attempt to simulate LULC in Thimphu city
and thus adds modeling direction (Subedi, et al., 2013). Together CA– using remote sensing and GIS techniques.
Markov model has the ability to stimulate and predict spatial transi-
tions in complex land use systems and has outperformed regression- Methodology
based models (Theobald and Hobbs, 1998). Thus CA-Markov model is
widely used by scholars to understand urban expansion and landscape Study area
dynamics globally (Araya and Cabral, 2010; Islam and Ahmed, 2011;
Guan, et al., 2011; Sohl and Claggett, 2013; Kityuttachai, et al., 2013; Thimphu city, the capital of Bhutan is the crucible of socio-economic,
Dutta, 2012; Puertas, et al., 2014; Hassan and Nazem, 2015; Han, et al., political, and environmental development in the country. Administra-
2015; Li, et al., 2016; Keshtkar, et al., 2017; Rimal, et al., 2017). Given tively, the city is composed of central core, southern extension, and
its robustness and popularity, the present study used CA-Markov analy- northern extension. Geographically, the city is located at 27˚ 29’N lati-
sis to carry out the geo-spatial analysis of Bhutan’s Thimphu city. tude and 89˚ 36’E longitude between an elevation of about 2347–2438
The Kingdom of Bhutan is located in the rugged landscape of the m amsl (Walcott, 2011) (Fig. 1). With an area of 26.2 km2 (Wang et al.,
eastern Himalayas between 88˚ 54’ and 92˚ 10’ E and 26˚ 40’ and 28˚ 15’ 2019a), the city stretches for about 15 km from Dechencholing in the
N (MoWHS, 2008; Banerjee and Bandopadhyay, 2016; Wangda, 2017). north to Babesa in the South along the narrow banks of the Wang chhu
Rapid changes in elevation from 100 amsl (above mean sea level) in river (“chhu” means river in Bhutanese language”) which drains south

2
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

Fig. 1. Map of study area showing local areas.

into India (MoWHS, 2008). Except for this narrow valley leading to the Table 1
south, the city is surrounded by mountains rising over 2,648m on all Characteristics of data collected.
sides. The mountains are mostly covered in forests composed of blue Sensor Month/Day/Year Resolution Path /Row
pine, hemlock, spruce, oak, transitioning into alpine meadows, which
Landsat 7 ETM+ 11/07/2002_Thimphu 30 m 138 / 41
forms important watersheds that drains into river Wang chhu. Like the
Landsat 8 OLI 11/27/2018_Thimphu 30 m 138 / 41
rest of the country, Thimphu city experiences 4 seasons. Winter temper-
atures range from freezing at high elevations (at night) to 12.3˚C during Source: (Source: http://glovis.usgs.gov)
the day, while summers are warmer with temperatures ranging from
21˚C (night) to 32˚C during the day (Walcott, 2011). The city on the
average receives about 813 mm of rainfall annually with much of the (i) Approval of city extension: Government approval of the comprehen-
rainfall happening during monsoon period from June through Septem- sive structural plan for Thimphu city in 2002, approved the expan-
ber (MoWHS, 2008). Combination of elevational gradient topped with sion of the core area towards the south and north (Norbu, 2020 un-
different types of forests has endowed the city and its periphery with di- published). This expansion rapidly replaced traditional buildings,
verse flora and fauna (Wang et al., 2019a). Riparian vegetation, marsh- agricultural fields, forests, and wetlands by buildings, roads, and
land vegetation, and recreational parks are the dominant types of vege- other artificial infrastructures.
tation within the city and its periphery. A number of ornamental plants (ii) Transition to democracy in 2008: Bhutan transition to democracy in
and trees such as willow, poplar, juniper, beech, and hibiscus have also 2008, resulted in policy changes that favored rapid socio-economic
been planted within the city. development to partly impress voters. This move expediated the
With over 15.75% (114,551) of Bhutan’s entire population (727,145) construction of public infrastructures such as roads, parking lots,
living in Thimphu, it is the most populated of all the cities of Bhutan recreational spaces, and other basic facilities in the city. In addi-
(NSB, 2017). This represents a rapid population growth of about 62.04% tion, changes in fiscal policies allowed easy access to housing loans
from 2002 (43,479) excluding armed forces and monk body (Norbu, thereby sparking a construction boom in the city. These develop-
2020 unpublished). Rural urban migration at 14.53% (Walcott, 2011) ments exacerbated the rate of LULC mainly from agriculture and
is primarily responsible for rapid population growth which currently forests to built. The construction projects recruited expat work-
stands at 3.7%. According to the statics from the National Statistics Bu- ers from India, who together with already existing population and
reau (2017), 54% of Thimphu city’s population is composed of people tourists also contribute to the LULC in the city.
who have immigrated from rural areas. This population is further in-
flated by a steadily growing population of expat workers and tourists Data acquisitions and preparation
(TCB, 2016). Rapidly growing urban population coupled with urban ex-
pansion and development is stretching the ecological and infrastructural Methodological framework of the simulation has been illustrated in
carrying capacity of the city. Already, the city is facing problems of inad- Fig. 2. Spatial and socio-economic data further validated by the first au-
equate infrastructure, squatter settlements, traffic congestion, flooding thor’s expert experience and field observation in the study area were
along riverbed (triggered by rainstorms, glacial lake outburst floods), used to simulate the LULC changes in the study area. Landsat-7 The-
and low-lying pockets in the city due to inadequate storm and sewage matic Mapper (TM) and Landsat-8 Operational Land Imager (OLI) im-
drains, as well as landslides. International Environment Development ages at a resolution of 30 m were acquired for the years 2002 and
(IIED) has classified Thimphu city as one of the 15 most vulnerable cities 2018 respectively, from the United States Geological Survey (USGS)
in the world (IIED, 2009). Center for Earth Resources Observation and Science (EROS) found at
http://glovis.usgs.gov/. The specifications of the satellite data acquired
Land policy and regime changes for change analysis are given in Table 1. Post monsoonal data were used
as these periods represent clear skies with least clouds.
Historically, Thimphu city started in 1961 with some 25 shops or so These data sets were imported into TerrSet satellite image process-
which currently falls within the central core of the city. Prior to this, the ing software to create a false color composite. Other geospatial data col-
area was occupied by terraced agricultural fields and some 13 villages. lected from Thimphu city office include, digital elevation model (DEM
Planned development for Thimphu started in 1985 with the establish- 30 m resolution), administrative boundaries, and infrastructure data
ment of National Urban Development Commission. Since then, two ma- such as road networks, drainage networks, water bodies, buildings, and
jor policy interventions have expediated the growth of the city thereby other important establishments in the city. Additionally, field consulta-
changing its LULC: tions with select communities, experts, and planners from the city office

3
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

Fig. 2. Methodological framework for data processing.

were held to collect expert explanations and perceptions on the causes Table 2
of land use and land cover changes including climate change. Land cover classification scheme.
The years 2002 and 2018 were selected to correspond to major socio- LULC Types Description
political and policy changes in Bhutan that has significant bearings on
Water body Rivers, streams, ponds, lakes, reservoirs, and marshlands
LULC in the city. Expansion of Thimphu city in 2002 added local areas
Bare ground Exposed soils, landfills, and areas of active excavation
of Lanjophaka, Heo-Samteling, Jungshi-Pamtsho, Taba and Dechenchol- Built-up area Commercial areas, urban settlements, industrial areas,
ing in the north and Changjiji, Lungtenphu, Simtokha, and Babesa in the government and institutional buildings, roads, hard surfaces,
south (Fig. 1). Since then, what was previously lush paddy fields, fruit parking areas, recreational areas, and golf courses,
Agriculture Crop land (wetlands and drylands), and orchards
orchards, and fringe forests dotted with traditional Bhutanese houses
Forest Plantations, broadleaf forests, conifer forests, and riparian
were rapidly replaced by modern concrete buildings and other infras- vegetations
tructure such as roads, parking lots, etc., resulting in significant changes
in Thimphu’s landscape. The initiation of constitutional democracy in
2008, further accelerated the rate of infrastructure development includ-
ing aggressive construction of commercial and residential areas, roads, For unsupervised classification, hyper clustering (use of much higher
service infrastructure. These hasty events triggered rapid conversion of number of clusters then the desired classes) was used as exact number
land with little regard to social and environmental safeguards and as of pixels were not known (Cihlar, 2000). These clusters were labelled
such a study to detect LULC changes would yield useful information for as water body, bare ground, built-up area, forest, and agriculture on
planners and policy makers. the google earth observation and other land use maps (MoWHS, 2008;
Norbu, 2020 unpublished) of the study site. Spectrally similar classes
Pre-processing and classification of the identical land use type were merged. For each of the land use
types, training samples were randomly generated from the satellite im-
Pre-processing is an important step to establish direct affiliation be- agery. The signature points were then tested for statistical similarities
tween the acquired data and biophysical phenomenon. Remotely sensed (Rojas, et al., 2013) which indicated a good degree of similarity-based
data are susceptible to radiance, geometric, and atmospheric distortions on spectral distance. This indicated minimum confusion between the
due to acquisition systems and platform movements. One major function land uses and as such considered satisfactory for the study (Gao, 2012).
of pre-processing is to remove such distortions especially when optical Visual interpretation was also used to resolve any issues related to mixed
sensor data is used. Atmospheric correction involves removal of haze pixels and enhance classification accuracy.
primarily originating from water vapor, fog, dust, smoke, or other par- The outputs were then subjected to supervised classification using
ticles in the atmosphere (Makarau, et al., 2014), which otherwise leads the Maximum Likelihood Classifier (MLC). MLC classifier was used as
to distortion of reflectance. In addition, topographic normalizations are it is a robust and popular algorithm increasingly used for classification
important for mountain environment such as Thimphu valley, as the of LULC (Lillesand, and Kiefer, 1999). MLC procedure uses a paramet-
presence of steep slopes can trigger variations in illuminations of identi- ric statistical approach to prepare the probability density distribution
cal features (Vanonckelen, et al., 2015). We imported the satellite data functions for each individual land use class (Thom, 1994; Pielke, et al.,
into TerrSet software and GIS and used its ATCOR (Atmospheric Topo- 2002). Compared to other methods, MLC is considered to be more ac-
graphic CORection) feature to remove haze and correct topographic nor- curate as it calculates the total amount of variance and the correlation
malization (Richter and Schlapfer, 2002) to generate images with true of the spectral values of different bands according to the specimen and
reflectance. To provide more accurate land use classification scheme uses this property for the association of pixels classified into one of the
both unsupervised and supervised (hybrid classification) approaches groups and is based on the most similarity between the pixels. In addi-
were used to derive five land use classes for Thimphu city (Table 2). tion MLC also reflects the intensity of land use changes and the visual

4
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

differences in land use types and considers not only the cluster cen- The Markovian chain analysis is represented as, S(t,t+1) = Pij × S(t),
ter but also its shape, size, and orientation (Hailemariam, et al., 2016; where, S(t) is the system status at time of t, S(t+1) is the system status
Molla, et al., 2018). Resultant thematic raster layers were then used for at time t+1; Pij is the transition probability matrix in a state, which is
post classification change detection. The use of high resolution google calculated using the following formula (Li, et al., 2016; Khanal, et al.,
earth map and firsthand knowledge of the study area by the principal 2019):
investigator considerably improved the results obtained using the super-
⎛𝜌11 𝜌12 ⋯ 𝜌1 𝑛 ⎞
vised algorithm. ⎜ ⎟
𝜌 𝜌22 ⋯ 𝜌2 𝑛 ⎟ ( )
𝜌𝑖𝑗 = ⎜ 21 0 ≤ 𝜌𝑖𝑗 ≤ 1 (2)
⎜ ⋮ ⋮ ⋮ ⋮ ⎟
Assessment of classification accuracy ⎜𝜌 𝜌𝑛 2 ⋯ 𝜌𝑛𝑛 ⎟⎠
⎝ 𝑛1
Assessment of classification accuracy is a prerequisite for classifi- where, 𝜌 is the Markov probability matrix, and 𝜌𝑖𝑗 stands for the prob-
cation data to credibly detect changes (Wang, et al., 2020). Accuracy ability of converting from current state i to another state j in next time
assessment was carried out on the resulting classified imagery using period. Low transition will have a probability near (0) and high tran-
error matrix and kappa index (Congralton, 1991; Keshtkar, et al., 2017) sition probability near (1). The 2002 LULC image of the Thimphu city
to test the precision and accuracy of imagery and comparing them with was used as the base (t1 ) image while 2018 LULC map as the later (t2 )
actual points from the field supplemented by 150 systematic points from image in this model to obtain the transition matrix between 2002 and
high resolution google earth data. Kappa coefficient was calculated us- 2018. In this study, ArcGIS cross-tabulation functionality was used to
ing the formula confirmed and used by Congralton and Green (2009); generate transitional area matrices by multiplying each column in the
∑𝑘 ∑ transition probability matrix by the number of pixels of corresponding
𝑖=1 𝑛𝑖𝑖 − 𝑘𝑖=1 𝑛𝑖𝑖 (𝐺𝑖𝐶𝑖) class in the later image.
Kappa coef f icient = ∑ (1)
𝑛2 − 𝑘𝑖=1 𝑛𝑖𝑖 (𝐺𝑖𝐶𝑖)
Transitional suitability maps
Where, i is the class number, n is the total number of classified pix- Multi-Criteria Evaluation (MCE) was used to build transition suitabil-
els that are being compared to actual data, nii is the number of pixels ity maps which show the probability of a pixel to change to another class
belonging to the actual data class i, that were classified with a class i, Ci or remain unchanged. MCE integrates various driving factors to derive
is the total number of classified pixels belonging to class i and Gi is the a single index of evaluation (El-Hallaq and Habboub, 2014). These driv-
total number of actual data pixels belonging to class i. ing factors differ according to the conditions of the region under study
In order to deepen the understanding of major land uses and drivers (Rimal, et al., 2018). As this is the first study of its kind for Thimphu city,
of their changes and socio-environmental implications, the principal in- authors firsthand knowledge in combination with complexity of the ter-
vestigator carried out field visits in 2019. During the field visits, field rain, socio-economic development, urban regulations, political changes,
observations and consultations were held with the stakeholders includ- physical proximity to current LULC were used to define transition rules.
ing communities, experts and local officials to collect biophysical and Accordingly, distance to main road, distance to water bodies, distance
climatic data. Discussions were also held to acquire information about to forests, distance to built areas, slope, were used in calculating transi-
urban expansion, evolution of land use, possible reasons for the observed tion suitability maps. The maps of road, and other infrastructure, along
changes and resident perception on socio-environmental resilience. In with DEM were obtained from the city office. Fuzzy membership func-
addition, records of major incidents such as disasters (fire, floods), plan- tions were used to standardize factor maps into 0-1, where 0 represents
tation efforts, policy changes, etc. were also noted. These information unsuitable locations and 1 represents ideal locations. The Analytic hier-
and firsthand experience from the field were used to further validate the archy process (AHP) was used to derive the weights of driving factors.
classified images.
The CA–Markov model
Land use and land cover change analysis The transition probability matrix and transition area matrix from
2002 to 2018 calculated in Markov chain analysis. The area of each
CA-Markov is a robust model that has outperformed other methods land class to be converted to another LULC classes was estimated based
for simulating and predicting LULC types (Theobald and Hobbs, 1998; on the transition probabilities. These areas were divided by the total
Kamusoka, et al., 2009; Guan, et al., 2011). Hence, this study selected number of iterations (16 years) for the cellular automation to generate
the CA-Markov model for simulating and predicting LULC change. This the areas to be converted per iteration. We applied a contiguity filter of
process involved (1) carrying out Markovian chain analysis on the 2002 5 × 5 pixels to define the effect of neighboring cells on the central cell.
and 2018 LULC maps to generate transition area matrices; (2) generat- The future assignment to a specific LULC class for each pixel was based
ing transitional area maps of LULC; (3) evaluating the model accuracy on how much the pixel is suitable for that LULC class and how close the
to simulate future changes based on Kappa indices; and (4) predicting pixel is to other pixels of the same class. The new policy and political
spatial distribution of LULC in 2050. developments after 2001 and 2008 were assumed as in simulating fu-
ture scenarios. The period, 2002–2018 represents a rapid expansion of
Markov chain analysis the city accompanied by expediated socio-economic development that
The Markov model simulated changes in LULC from one time to an- led to significant changes in the city’s landscape. Assuming that these
other in order to predict future change. Markov chain analysis built in probabilities (2002-2018) will hold and using the interpreted map of
module at TerrSet was used to generate transition probability matrix 2018 as the base map, we finally predicted LULC in 2050.
(where the probabilities of transition represent the probability that a
pixel of a given class will move to some other cell class in the next Results and discussion
time period) and transition area matrix (which represents the total area
(in cells) expected to change from one LULC class to another over the Classification accuracy
prescribed number of time units). The transition probability matrix is
expressed in a text file that records the likelihood of moving each land Model validation is an important precondition for studies that at-
use and land cover category to some other category, while the transition tempt to predict LULC changes (Appiah, et al., 2015). The Kappa statis-
area matrix, also represented in a text file records the number of pixels tic is one of the most popularly accepted methods for quantifying a
required to transition from one land use and land cover class to another model’s predictive power (Maingi and March, 2002; Hua, 2017) and can
over the specified number of time unit (Fig. 2). take any values from -1 to +1 (Congralton and Green, 2009). Studies by

5
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

Fig. 3. Land use and land cover map of 2002

and 2018

Table 3
Composite table of area statistics (km2 ) of Thimphu city 2002 and 2018.

LULC 2002 2018 Over all change 2002 - 2018

Types
Area (km2 ) Area (%) Area (km2 ) Area (%) Area (km2 ) Area (%)

Water body 0.45 1.70 0.45 1.72 0.00 0.02

Bare ground 0.63 2.38 1.56 5.94 0.93 3.55
Built-up area 13.86 52.88 17.20 65.65 3.34 12.77
Forest 10.94 41.76 6.94 26.50 -4.00 15.25
Agriculture 0.32 1.20 0.05 0.19 -0.26 1.01
Total 26.20 100 26.20 100

Manomani and Suganya (2010) and Maingi and Marsh (2002) have cat- along the riparian zones. Bare ground is found scattered across the city
egorized kappa values into several groups: <0 indicate no agreement, while agriculture is confined to certain pockets only. In 2002, the largest
0-0.2 as slight, 0.0-0.41 as poor, 0.41-0.60 as moderate, 0.60-0.80 as land use types were built-up (52.88%), followed by forest (41.76%), and
significant, and 0.81-1.0 as almost perfect agreement. Kappa analysis bare ground (2.38%). By 2018, city’s built-up area increased the most
for this study generated values well over 70, qualifying them as signifi- from 13.86 km2 to 17.20 km2 (65.5%) reflecting an overall increase of
cant predictors (Manonmani and Suganya, 2010). 12.77% (3.34 km2 ) in 2018 (Fig. 4). Concrete buildings for residential
and commercial purposes, road networks, pavement, parking lots, recre-
Analysis of land use and land cover types ational facilities, and other impervious surfaces were responsible for
the expansion of built-up area in the city. This has significantly reduced
LULC analysis allows us to understand biophysical changes such the forest cover by 15.25% (4 km2 ; Fig. 4) which is mostly confined
as: loss of productive ecosystems/biodiversity, deterioration of environ- to the periphery of the town especially in the south and north (Fig. 3).
mental quality, loss of forest and agricultural lands what are important Currently, forest coverage of the city is only about 26% compared to
information for planning sustainable cities (Mallupattu and Screeniva- 42% in 2002 (Table 3). Between 2002 and 2018, bare ground area has
sula, 2013). also increased by 3.55% (Table 3) due to increasing excavation, land-
Our study applied latest remote sensing and GIS techniques to quan- fills, and exposure of soils. The results also show that the conversion of
tify LULC in Thimphu city. Using the outputs from remote sensing im- forest to other land use types remain high during the study period with
agery, field surveys, and topped expert knowledge of the study area, built-up area consuming over 28% of the total forest loss (Fig. 4).
five LULC types were classified; water body, bare ground, built-up area, These expediated changes can be attributed to: i) governments ap-
forest, and agriculture (Fig. 3). LULC analysis show that the extent of proval for expansion of the city in 2002, followed by the hastened con-
land cover types varied across the years and we explored the signif- struction of city infrastructure post Bhutan’s transition to democracy in
icant transformations between 2002 and 2018 (Table 3, Fig. 3). The 2008. The resultant improvement in basic facilities combined with cre-
area of land use types was calculated (Table 3) from the classified im- ation of additional jobs further attracted more immigration from rural
ages of 2002 (a) and 2018 (b) presented in Fig. 3. Fig. 3 confirm that areas and lesser developed cities, causing a boom in real estate and in-
the built-up is most intense at the core of the city and radiates out- formal settlements and triggering significant changes in land uses.
wards. Forests are mostly confined to the marginal fringe lands located Studies by Bruggeman, et al. (2016) and Yangchen, et al. (2015) also
at higher grounds with some spotty patches scatted through the city and confirmed infrastructure development especially by rapidly growing

6
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

Table 4
Land use class transition matrix from 2002 to 2018.

LULC 2018 image (km2 )

Types
Water body Bare ground Built-up area Forest Agriculture Total

2002 image Water body 0.17 0.00 0.13 0.14 0.00 0.45
(km2 ) Bare ground 0.00 0.11 0.49 0.02 0.00 0.62
Built-up area 0.10 1.17 11.83 0.71 0.00 13.82
Forest 0.15 0.55 4.25 6.04 0.00 10.94
Agriculture 0.00 0.01 0.21 0.08 0.01 0.31
Total 0.66 1.56 16.92 6.99 0.01 26.14

Fig. 4. Gains and losses between 2002 and 2018.

cities such as Thimphu as a primary cause of forest loss in Bhutan. Glob- the most stable land use type with a transition probability of 0.82. This
ally, the loss of forest and other land types in Thimphu city is a typ- implies that it the probability of built up area transitioning into other
ical trend observed in urbanization especially in developing countries land use types is very low. Although low, it’s encouraging to see forest
(Sun, et al., 2013; Pandey and Seto, 2015; Kegazy and Kaloop, 2015; type retaining some stability, this is probably due to some plantation
Bruggeman, et al., 2016). It is generally proven that the loss of forest efforts that the Government have initiated recently. The most dynamic
and vegetative cover will degrade ecosystem services by reducing wa- land use types are bare ground, agriculture, and forest with transition
ter retention, drying of water sources, loss of habitat for biodiversity, probabilities of 0.10, 0.01, and 0.34 respectively. Most land types were
reduced sequestration rate of CO2 , and enhancing the magnitude and also converted to built-up areas with the largest area coming from forest
frequency of disasters such as flash floods, landslides, and scarcity of (4.25km2 ) followed by bare ground (0.49km2 ).
water and other natural resources, which are vital for human wellbeing The potential for these trends to continue remain high with transi-
(Woldeamlak, 2002). Loss of forest and vegetative cover can amplify is- tion probabilities of 0.83 for bare ground and 0.58 for forest to transition
land heat affect in cities such as Thimphu which is located in a deep val- to built-up area. Although small, agriculture land shows a significant
ley surrounded by high mountains on all sides. Gogoi, et al. (2019) sim- probability to transition to built-up area which has been the trend in
ulated the impact of LULC on surface temperature over eastern India many cities (Pandey and Seto, 2015; Rimal, et al., 2018; Ishtiaque, et al.,
and found that 20–50% of the observed warming were associated with 2017), and represent future threat to existing agricultural land adjoining
LULC. They further confirmed that largest changes were linked to chang- the city. A slight but interesting transition of forest to water body could
ing vegetation cover. Given that the current trend of LULC is mainly have been the result of clearing forests along the river and streams ex-
triggered by building infrastructure in the face of limited size of various posing larger water surfaces. Fig. 5 confirms these transitions and show
land use types, it won’t be long before the city’s growth will use the the transitions spatially. It is clear from Fig. 5 that most of the con-
available bare ground and expand development into existing forest, ri- versions are occurring in the newly extended areas in the north and
parian areas, and may even continue encroaching into the government south as well as along the fringes of the core area. This is because the
reserve forests on the surrounding hills. Such a scenario is highly unde- central core is already built to almost its full capacity forcing infras-
sirable as it will further degrade the fragile watersheds and destabilize tructure construction to extend south and north after 2002. Most in-
ecological functions including providing water which is already in short tense conversion from forest to built-up took place in Babesa and Ser-
supply. The city is currently grappling with problems of water shortage, bithang areas in the southern extension of the city. A lot of the bare
expansion of squatter settlements, sanitation, flooding of storm drains ground in the northern extension of the city transitioned to built-up
during rainy seasons, etc. Some sections of the city including those in in- area. Bare grounds in this part of the city were previously agricultural
formal settlements have irregular or no supply of drinking water. Under land left fallow in anticipation for development after inclusion in the city
the business as usual scenario and expected impacts from global climate in 2002.
change, there is an urgent need to step up adaptation and mitigation in- The findings indicate that rapid development of urban built will
terventions including massive plantation in vacant spaces and roads, as lead to sharp declines in forest and agriculture and increase in urban
well replace hard surfaces with vegetation. population. These developments could jeopardize ecosystem health, hu-
man well-being, and food security (Rahman, 2016; Khanal, et al., 2019).
LULC transition between years Further, most of forest conversions are taking place in higher grounds
and watersheds, which could expose bare ground and make them vul-
To deepen our understanding of the evolving nature of the LULC nerable to gully erosion (Miheretu and Yimer, 2017) which can cause
types in the city, we created confusion matrices for changes in 2002 serious damages especially in fragile and rugged terrain like Thimphu.
and 2018 (Table 4). All unchanged pixels are presented diagonally in In addition, increased population could result in increased pollution
bold fonts. The details of land cover transition between the 5 land use and excessive extraction of natural resources which could lead to detri-
types are mapped in Fig. 5. Corresponding transitional probability is mental consequences including urban heat wave (Gogoi, et al., 2019),
also tabulated into Table 5. During the study period, built-up area is landslides, and health problems (Hansen, et al., 2013).

7
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

Fig. 5. LULC transition map of Thimphu city.

Table 5
Land use transition probability matrix from 2002–2018.

2018 image (km2 )

Water body Bare ground Built-up area Forest Agriculture

2002 Water body 0.17 0.03 0.49 0.31 0.00

im- Bare ground 0.01 0.10 0.83 0.06 0.00
age Built-up area 0.01 0.09 0.82 0.08 0.00
(km2 ) Forest 0.02 0.08 0.58 0.32 0.00
Agriculture 0.01 0.07 0.72 0.19 0.01

Table 6 city especially along streams, rivers, and marshlands. Fig. 6 also re-
Predicted land use and land cover for 2050. confirm the findings that most of what is covered by forest in 2018
LULC types Area (km2 ) Area (%) will transition into built-up and bare ground. The presence of bare
ground indicates soil excavations probably from the last wave of con-
Water body 0.46 1.74
struction and landfills. These findings are reflective of the current ur-
Bare ground 2.20 8.40
Built-up area 19.24 73.43 banization trend and government policy. The predicted trends of urban
Forest 4.28 16.32 built slowly or rapidly taking over other important land types such as
Agriculture 0.03 0.12 forest and agriculture are similar to those reported by other urbanization
studies (Huang, et al., 2008; Han ad Song, 2015; Appiah, et al., 2015;
Rimal, et al., 2018; [13,22,26,51] in developing countries. If Thimphu
city is to avoid irreversible problems that are plaquing many global
Prediction of land use land cover change cities, the city must shift form present business model to urgently adopt
ecosystem based approach to urban development.
Results of LULC prediction using CA-Markov analysis are shown in
Table 6 and Fig. 6. Predicted figures for 2050 show further decrease in Conclusion
the forest (16.32%) and agriculture (0.12%) land use types which com-
pared to 2018 figures (Table 3) represent a loss of 10.18%, and 0.07%, This study represents a first ever attempt to simulate the process of
respectively. These losses are gained by built-up area (7.78%) and bare land use dynamics and the effects of physical, demographic, and socio-
ground (2.46%) taking their total coverage to 19.24km2 (73.43%) and economic driving forces on LULC in Bhutan’s capital city of Thimphu
2.20km2 (8.40%) respectively. using remote sensing and GIS technology. Results clearly confirm sig-
Fig. 6 show the predicted spatial destruction of LULC in 2050. Over- nificant changes in LULC from the start of expansion to current area in
all, forest cover is fragmented with larger patches remaining mainly in 2002 to 2018. Significant increase in built-up areas were associated with
the south of the city along the river and on the fringes on higher grounds. parallel loss of important land cover types such as forest and agricul-
Small and narrow patches of forests are also seen scattered across the ture. Significant changes in Thimphu’s landscapes are correlated with

8
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

Fig. 6. Predicted LULC map of Thimphu city in 2050.

the government’s decision to expand the city northwards and south- sion, host biodiversity, as well as regulate temperature and pollution.
wards. This expansion received a boost after the institution of constitu- Current capacities of storm drains and water reservoirs must also be
tional democracy in 2008 which brought in expediated development in enhanced. Informal settlements must either be formalized by providing
the city. These two policy events topped with rapid population growth all basic facilities or moved to formal settlements. We also caution the
mainly from rural urban migration for better opportunities and develop- city to restrain from land filling flood plains and use the claimed land
ment activities expediated the LULC changes in the city. Combined these for building infrastructure as these areas are not only prone to flood-
factors have reduced and fragmented forest cover which will degrade ing especially global lake outburst floods but also important for biodi-
ecosystem services that are vital for maintaining the quality of human versity. We encourage the city planners and experts to use the findings
wellbeing in the city. Reports from the city residents also confirm that and recommendations from our study to avert irreversible changes to its
the impacts of these changes especially in the face of climate change LULC.
are disproportionately felt by the most vulnerable groups such as low The accuracy of the classification results was not perfect due to the
income, women, and children. The results of prediction for 2050 also hilly nature of the study area, where ground changes in altitude can
did not bear well for Thimphu city, with forest cover diminishing to a impact the image pixel value. However, it is significant and as such
mere 4.28km2 (16.32%) which are also highly fragmented. Loss of forest study findings and recommendations can be applied to other cities in
cover will degrade important ecological services and increase vulnera- Bhutan and other mountainous countries which share similar ecological
bility of the city to landslides, gulley erosion, drying of water sources, and socioeconomic characteristics.
worsened air pollution, and loss of important biodiversity. More stud-
ies are recommended to especially monitor the impact of these changes Declaration of Competing Interests
and identify adaptive and mitigative interventions for a more sustain-
able city. Authors have no competing interests.
Study also show that remote sensing and GIS are effective tools for
References
simulating urban changes, which are useful for guiding urban planning
and management. The findings of this study are useful to policy makers, Addo, K.A., 2010. Urban and Peri-urban agriculture in developing countries studied using
urban planners, and citizens to adopt better environmental management remote sensing and in situ methods. Remote Sens. 2, 497–513.
practices including adaptation and mitigation strategies for the city and Aitkenhead, M.L., Alders, H.H., 2009. Predicting land cover using GIS, Bayesian and evo-
lutionary algorithm methods. J. Environ. Manag. 90, 236–250.
its surrounding areas. The study recommend an ecosystem based adapta- Appiah, D.O., Schroder, D., Forkuo, E.K., Bugri, J.T., 2015. Application of geo-informa-
tion policies and other legal frameworks should be developed and prac- tion techniques in land use and land cover change analysis in a peri-urban district of
tically implemented to protect the current forest cover as well as reha- Ghana. Int. J. Geo Inf. 4, 1265–1289.
Araya, Y.H., Cabral, P., 2010. Analysis and modelling of urban land cover change in Se-
bilitate and improve the existing green spaces in and around the urban tubal and sesimbra. Portugal. Remote Sens. 30, 1549–1563.
centers. Such efforts must focus on planting fruit trees in urban spaces Banerjee, A., Bandopadhyay, R., 2016. Biodiversity hotspot of Bhutan and its sustainabil-
such as parking lots, between buildings, back yards, and along roads. ity. Curr. Sci. 110, 521–527.
Boissiere, M., Sheil, D., Basuki, I., Wan, M., Le, H., 2009. Can engaging local people’
These will significantly, increase water retention, protect soil from ero- interest reduce forest degradation in Central Vietnam? Biodiv. Cons. 18, 2743–2757.

9
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

Bruggeman, D., Meyfroidt, P., Lambin, E.F., 2016. Forest cover changes in Bhutan: Revis- Maingi, J.K., Marsh, S.E., 2002. An Accuracy Assessment of 1992 Landsat-MSS Derived
iting the forest transition. Appl. Geog. 67, 49–60. Land Cover for the Upper San Pedro Watershed (U.S./Mexico). United States Environ-
Cihlar, J., 2000. Land cover mapping of large areas from satellites: status and research mental Protection Agency, Washington, DC, USA, p. 29.
priorities. Int. J. Remote Sens. 21, 1093–1114. Makarau, A., Richter, R., Muller, R., Reinartz, P., 2014. Haze detection and removal in re-
Congralton, R.G., Green, K., 2009. Assessing the Accuracy of Remotely Sensed Data: Prin- motely sensed multispectral imagery. IEEE Trans. Geo. Remote Sens. 52, 5895–5905.
ciples and Practices, second ed. Taylor and Francis Group, LLC, Abingdon, UK. Mallupattu, P.K., Sreenivasula Reddy, J.R., 2013. Analysis of land use/land cover changes
Congralton, R.G., 1991. A review of assessing the accuracy of classifications of remotely using remote sensing data and GIS at an urban area, Tirupati, India. The Sci. World
sensed data. Remote Sens. Environ. 37, 35–46. J..
Coppin, P., Jonckheere, K., Nackaerts, K., Muys, B., 2004. Digital change detection meth- Manonmani, R., Suganya, G.M.D., 2010. Remote sensing and GIS application in change
ods in ecosystem monitoring: a review. Int. J. Remote Sens. 25, 1565–1596. detection study in urban zone using multi temporal satellite. Int. J. Geomat. Geo Sci.
Dutta, V., 2012. Land use dynamics and Peri-urban growth characteristics reflections on 24, 60–65.
master plan and urban suitability from a sprawling North Indian city. Environ. Urban MEA., 2005. Ecosystems and Human Well-Being: Opportunities and Challenges for Busi-
3 277–30. ness and Industry, World Resources Institute, Washington, SC. USA.
Eastman, J.R., 2009. IDRISI Guide to GIS and Image Processing Accessed in IDRISI Selva Miheretu, B.A., Yimer, A.A., 2017. Land use/land cover changes and their environmental
17. Clarkson University, Worcester, MA 182e185. implications in the Gelana sub-watershed of northern highlands of Ethiopia. Environ.
El-Hallaq, M.A., Habboub, M.D., 2014. Using GIS for time series analysis of the dead sea Sys. Res. 6, 1–12.
from remotely sensing dat. Open J. Civil Eng. 4, 386–396. Mishra, R.K., Rai, A., Rai, S.C., 2019. Land use and land cover change detection using
FRMD., 2017. Land use and land cover assessment for Bhutan 2016: Technical Report. geospatial techniques in the Sikkim Himalaya. India. Egyptian J. Remote Sens. Space
Forest Resource Management Division, Department of Forest and Park Services, Royal Sci. 23, 133–143.
Government of Bhutan. ISBN: 978-99936-743-3-7. Molla, M.B., Ikporukpo, C., Olatubara, C., 2018. The spatio-temporal pattern of urban
Gao, J., 2012. Land Cover Change and Its Relationship with Climate Change in Koshi River green spaces in Southern Ethiopia. Am. J. Geog Info. Sys. 7, 1–14.
Basin of Central Himalaya. Institute of Geographic Sciences and Natural Resources. MoWHS., 2008. Thimphu City Development Strategy. Ministry of Works and Human Set-
Graduate University of Chinese Academy of Sciences, Beijing, China, p. 160. tlement (MoWHS). Royal Government of Bhutan, Thimphu.
Gilani, H., Shrestha, H.L., Murthy, M.S.R., Phuntso, P., Pradhan, S., Bajracharya, B., MRC., 2010. State of the basin report 2010. Mekong River Commission (MRC), Vientiane,
Shrestha, B., 2015. Decadal land cover change dynamics in Bhutan. J. Environ. Manag Lao PDR. ISBN: 978-993-2080-57-1.
148, 91–100. Muller, M.R., Middleton, J.A., 1994. Markov model of land-use change dynamics in the
Gogoi, P.P., Swain, D., Roberts, G., Dash, J., Trapahty, S., 2019. Land use and land cover Niagara Region. Ontario, Canada. Landsc. Ecol. 9, 151–157.
change effect on surface temperature over Eastern India. Scient. Reports 9, 8859. Norbu G., 2020 unpublished. Urban Planning in Thimphu – its contribution to Develop-
Guan, D., Li, H., Inohae, T., Su, W., Nagaie, T., Hokao, K., 2011. Modelling urban land ment in Neither Tale (s) nor Poems but Some Concept Papers.
use change by the integration of cellular automation and markov model. Ecol. Model NSB., 2017. 2017 Population and Housing Census of Bhutan. National Statistical Bureau,
222, 3761–3772. Royal Government of Bhutan, Thimphu ISBN: 987-99936-28-64-4.
Hailemariam, S.N., Soromessa, T., Teketay, D., 2016. Land use and land cover change in NSSC., PPD., 2011. Bhutan Land Cover Assessment 2010. Thimphu: National Soil and
the Bale Mountain eco-region of Ethiopia during 1985 to 2015. Land 5, 41. Services Centre and Policy and Planning Division, Ministry of Agriculture and Forests,
Han, H., Yang, C., Song, J., 2015. Scenario simulation and the predition of land use and Royal Government of Bhutan Tech. report.
land cover change in Beijin, China. Sustainability 7, 4360–4370. Overmars, K.P., de Koning, G.H.J., Veldkamp, A., 2003. Spatial autocorrelation in multi-
Hansen, M.C., Potapov, P.V., Moore, R., Hancher, M., Turubanova, S.A., Tyukavina, A., -scale land use models. Ecol. Model. 164, 257–270.
Stahman, S.V., Goetz, S.J., Loveland, T.R., Kommareddy, A., Egorov, A.V., 2013. High- Pandey, B., Seto, K.C., 2015. Urbanization and agricultural land loss in India.: comparing
-resolution global maps of 21st century forest cover change. Science 342, 850–853. satellite estimates with census data. J. Environ. Manag. 148, 53–66.
Hassan, M.M., Nazem, M.N.I., 2015. Examination of land use/land cover changes, urban Pielke Sr, R.A., Marland, G., Betts, R.A., Chase, T.N., Eastman, J.L., Niles, J.O.,
growth dynamics, and environmental sustainability in Chittagong city, Bangladesh. Niyogi, D.D.S., Running, S.W., 2002. The influence of land-use change and landscape
Environ. Dev. Sustain. 18, 9672–9678. dynamics on the climate system: relevance to climate-change policy beyond the ra-
Hua, A.K., 2017. Land use land cover changes in detection of water quality: A study based diative effect of greenhouse gases. Philos. Trans. R. Soc. Lond. Series A: Mathe. Phy.
on remote sensing and multivariate statistics. J. Environ. Publ. Heal., 7515130. Eng. Sci. 360, 1705–1719.
Huang, W., Luan, Q., Jiang, Q., Liu, J., Liu, H., 2008. Detection and prediction of land use Puertas, O.L., Henriquez, C., Meza, F.J., 2014. Assessing spatial dynamics of urban growth
change in Beijing based on remote sensing and GIS. The international archives of the using an integrated land use model. Application in Santiago metropolitian area. Land
photogrammetry. Remote Sens. Spat. Inf. Sci. 37, 75–82. Use Pol. 38 425-425.
Hyandye, C., 2015. GIS and logit regression model applications in land use and land cover Rahman, M., 2016. Detection of land use/land cover changes and urban sprawl in Al-Kho-
change and distribution in Usangu catchment. Am. J. Remote Sens. 3, 6–16. bar, Saudi Arabia: an analysis of multi-temporal remote sensing data. Int. J. Geo info.
IIED., 2009. Climate change and the urban poor: Risk and Resilience in 15 of the World’s 5, 15.
most Vulnerable Cities. International Institute of Environment and Development Ralha, C.G., Abreu, C.G., Coelho, C.G.C., Zaghetto, A., Macchiavello, B., Machado, R.B.,
(IIED). London, United Kingdom. 2013. A multi-agent model system for land-use change simulation. Remote Sens. En-
Ishtiaque, A., Shrestha, M., Chhetri, N., 2017. Rapid urban growth in the Kathmandu viron. 42, 30–46.
valley, Nepal: monitoring land use land cover dynamics of a Himalayan city with Richter, R., Schlapfer, D., 2002. Atmospheric/Topographic Correction for Satellite Im-
landsat imageries. Environ 4, 72. agery ATCOR-2/3 User Guide to Remote Sensing Applications. Schläpfer: Langegg-
Islam, Md.S., Ahmed, R., 2011. Land use change prediction in Dhaka city using GIS aided weg, Switzerland.
Markov chain Modelling. J. Life Earth Sci. 6, 81–89. Rimal, B., Zhang, L., Keshtkar, H., Haak, B.N., Rijal, S., Zhang, P., 2018. Land Use/Land
Kalnay, E., Cai, M., 2003. Impact of urbanization and land-use change on climate. Nature Cover Change Dynamics and Modeling of Urban Land expansion by the Integration
423, 528–531. of Cellular Automata and Markov Chain. Int. J. Geo Info. 7, 1–21.
Kamusoka, C., Aniya, M., Adi, B., Manjoro, M., 2009. Rural sustainability under threat in Rimal, B., Zhang, L., Keshtkar, H., Wang, N., Lin, Y., 2017. Monitoring and modeling
Zimbabwe – simulation of future land use/cover changes in the Bidura district based spatio-temporal urban expansion and landuse and land cover change using integrated
on the Markov-cellular automate model. Appl. Geo. 29, 435–447. markov chain cellular automata model. Int. J. Geo. Info. 6, 1–21.
Keshtkar, H., Voigt, W., Alizadeh, E., 2017. Land-cover classification and analysis of Rogan, J., Chen, D., 2004. Remote sensing technology for mapping and monitoring land–
change using machine-learning classifiers and multi-temporal remote sensing im- cover and land-use change. Progr. Plann. 61, 301–325.
agery. Arab. J. Geosci. 10, 154. Rojas, C., Pino, J., Basnou, C., Vivanco, M., 2013. Assessing land-use and cover changes
Khanal, N., Uddin, K., Matin, M.A., Tenneson, K., 2019. Automatic detection of spatiotem- in relation to geographic factors and urban planning in the metropolitan area of con-
poral urban expansion patterns by using OSM and landsat data in Kathmandu. Remote cepcion’ (Chile). Implications for biodiversity conservation. Appl. Geogr. 39, 93–103.
Sens. 11, 2296. Sala, O.E., Chapin, F.S., Armesto, J.J., Berlow, E., Bloomfield, J., Dirzo, R., Huber-San-
Kityuttachai, K., Tripathis, N., Tipdecho, T., Shretha, R., 2013. CA-Markov analysis of wald, E., Huenneke, L.F., Jackson, R.B., Kinzig, A., Leemans, R., Lodge, D.M.,
constrained coastal urban growth modelling. Hua hin seaside city. Thailand. Sustain. Mooney, H.A., Martin, O., Poff, N.L., Sykes, M.T., Walker, B.H., Walker, M.,
5, 1480. Wall, D.H., 2000. Global biodiversity scenarios for the year 2100. Science 287,
Kumar, P., Jain, S., Gurjar, B.R., Sharma, P., Khare, M., Morawska, L., Britter, R., 2013. 1770–1774.
New directions: can a “blue sky” return to Indian megacities? Atmos. Environ. 71, Sharma, K., Robeson, S.M., Thapa, P., Saikia, A., 2016. Land-use/land-cover change and
198–201. forest fragmentation in Jigme Dorji National Park, Bhutan. Phy. Geog. 38, 18–35.
Lambin, E.F., 1997. Modelling and monitoring land-cover change processes in tropical Sherbinin, A.de., Schillar, A., Pulsipher, A., 2007. The vulnerability of global cities to
regions. Progr. Phys. Geogr. 21, 375–393. climate hazards. Environ. Urbaniz. 19, 39–64.
Lambin, E.F., Rounsevell, M.D.A., Geist, H.J., 2000. Are agricultural land use models able Singh, S.K., Mustak, S., Srivastava, P.K., Szabo, S., Islam, T., 2015. Predicting spatial and
to predict changes in land use intensity? Agri. Ecosys. Environ. 82, 321–331. decadal LULC changes through cellular automata Markov Chain models using earth
Li, L., Lu, L., Kuang, D., 2016. Examining urban impervious surface distribution and its observation datasets and geo-information. Environ. Proc. 2, 61–78.
dynamic change in Hangzhou Metropolis. Remote Sens. 8, 1–23. Sinha, P., Kumar, L., 2013. Markov land cover change modelling using pairs of time series
Lillesand, T.M., Kiefer, R.W., 1999. Remote Sensing and Image Interpretation. Wiley, New satellite images. Photogramm. Eng. Remote Sens. 79, 1037–1051.
York, NY, USA. Sklar, F.H., Costanza, R., 1991. The development of dynamic spatial models for land-
Liping, C., Yujun, S., Saeed, S., 2018. Monitoring and predicting land use and land cove scape ecology: a review and prognosis. In: Turner, M.G., Gardner, R.H. (Eds.).
changes using remote sensing and GIS techniques – A case study of a hilly area. Jian- In: Ecological Studies, Quantitative Methods in Landscape Ecology, 82. Springer,
gle, China. Plosone 13, 1–23. Berlin/Heilderberg, Germany, pp. 239–288 1991.
Lu, D., Mausel, P., Brondı ´zio, E., Moran, E., 2004. Change detection techniques. Int. J. Sohl, T.L., Claggett, P.R., 2013. Clarity versus complexity: land-use modeling as a practical
Remote Sens. 25, 2365–2401. tool for decisionmakers. J. Environ. Manage. 129, 235–243.

10
S.W. Wang, L. Munkhnasan and W.-K. Lee Environmental Challenges 2 (2021) 100017

Stefanov, W.L., Ramsey, M.S., Christensen, P.R., 2001. Monitoring urban land cover Vanonckelen, S., Lhermitte, S., Van Rompaey, A., 2015. The effect of atmospheric and
change: an expert system approach to land cover classification of semiarid to arid topographic correction on pixel-based image composites: improved forest cover de-
urban centers. Remote Sens. Environ. 77, 173–185. tection in mountain environments. Int. J. Appl. Earth Obs. Geo info. 35, 320–328.
Subedi, P., Subedi, K., Thapa, B., 2013. Application of a hybrid Cellular Automaton— Veldkamp, A., Lambin, E.F., 2001. Predicting land use change. Agri., Ecosys. Environ. 85,
Markov (CA-Markov) model in land-use change prediction: a case study of saddle 1–6.
creek drainage basin. Florida. Sci. & Edu. 1, 126–132. Walcott, S.M., 2011. One of a kind: Bhutan and the modernity challenge. Natl. Identities
Sun, C., Wu, Z.F., Lv, Z.Q., Yao, N., Wei, J.B., 2013. Quantifying different types of urban 13, 253–266.
growth and the change dynamic in Guangzhou using multi-temporal remote sensing Wang, S.W., Gebru, B.M., Lamchin, M., Kayastha, R.B., Lee, W.K., 2020. Land use and
data. Int. J. Appl. Earth Obs. Geo info. 21, 409–417. land cover change detection and prediction in the Kathmandu district of Nepal using
Suresh, Y., Balachandar, D., Murthy, K.R., Muruganandam, R., Kumaraswamy, K., 2011. remote sensing and GIS. Sustainability 122-18.
Land use/land cover change detection through using remote sensing and GIS technol- Wang, S.W., Lee, W.K., Kim, J.G., 2019a. Assessing barriers and opportunities for ecosys-
ogy—A case study of St. Thomas Mount Block, Kancheepuram District, Tamil Nadu. tem-based adaptation in high altitude city of Thimphu. Urb. Stud. Pub. Adm. 2, 1–21.
Int. J. Curr. Res. 3, 501–504. Wang, S.W., Manjur, B., Kim, J.G., Lee, W.K., 2019b. Assessing socio-economic impact of
Taubenbock, H., Wiesner, M., Felbier, A., Marconcini, M., Dech, T.E., 2014. New dimen- agricultural subsidies: a case study from Bhutan. Sustainability 11, 1–12.
sions of urban landscapes: The spatio-temporal evolution from a polynuclei area to a Wangda, P., 2017. Rangelands Areas of Bhutan. National Research Center for Animal Nu-
mega-region based on remote sensing data. Appl. Geogr. 47, 137–153. trition Bumthang. Royal Government of Bhutan.
TCB., 2016. Bhutan Tourism Monitor. An Annual Report. Tourism Council of Bhutan Weng, Q., 2001. A remote sensing-GIS evaluation of urban expansion and its impact on
(TCB). Royal Government of Bhutan. Thimphu. surface temperature in the Zhujiang Delta, China. Int. J. Remote Sens. 22, 1999–2014.
Theobald, D.M., Hobbs, B.T., 1998. Forecasting rural land use change: a comparision of Woldeamlak, B., 2002. Land cover dynamics since the 1950s in Chemoga watershed, Blue
regression and spatial transition-based models. Geogr. Environ. Model. 2, 65–82. Nile Basin, Ethiopia. Mount. Res. Dev. 22, 263–269.
Thom, B., 1994. Special Issue on the October 1993 Conference on Land-Use and Land– Yang, X., Zheng, X-Q., Lv, L-N., 2012. A spatiotemporal model of land use change based
Cover in Australia-Living with Global Change-Preface. John Wiley & sons ltd, baffins on ant colony optimization, Markov chain and cellular automata. Ecol. Model 233,
lane Chichester, W. Sussex, England PO19 1UD. 11–19.
Turner, B.L., Lambin, E.F., Reenberg, A., 2007. The emergence of land change science Yangchen, U., Thinley, U., Wallentin, G., 2015. Land use land cover changes in Bhutan:
for global environmental change and sustainability. Proce. Nat. Aca. Sci USA. 104, 2000-2013. Occas. Publi. 1, 37–46.
20666–20671. Zhao, L., Peng, Z-R., 2012. LandSys: an agent-based cellular automata model of land use
UN., 2016. The World’s Cities in 2016—Data Booklet (ST/ESA/ SER.A/392). Department change developed for transportation analysis. J. Transp. Geog. 25, 35–49.
of Economic and Social Affairs, Population Division, United Nations.

11