Builtup Area Simulation Using MLP Markov Model

(A case study from Chaurjahari Municipality, West-Rukum, Nepal)

Tribhuvan University
Institute of Forestry, Pokhara Campus
Pokhara, Nepal

Researcher Advisor
Milan Budha Mr. Mahendra Singh Thapa
B.Sc. Forestry Assistant Professor
Tribhuvan University Department of Watershed
Institute of Forestry, Pokhara Management and
Campus Environmental sciences
Pokhara, Nepal IoFPC, Pokhara

A Project Paper submitted to Tribhuvan University, Institute of Forestry, Pokhara, Nepal

for the partial fulfillment of the requirement for the degree of Bachelor of Science in
Forestry
May 2019
© Milan Budha
2019
Contact: +977-9843664384
Email: [email protected]

Tribhuvan University
Institute of Forestry, Pokhara Campus
Pokhara Metropolitan city 15, Kaski,
Gandaki Province, Nepal
P.0. Box: 43,
Phone: 061-430469
Fax: 061-430387
Website: www.iofpc.edu.np

Citation:
Budha, M. (2019). Builtup Area Simulation Using MLP Markov Model (A case study from
Chaurjahari Municipality, West-Rukum, Nepal). A project paper submitted to Tribhuvan
University, Institute of Forestry, Pokhara campus, Pokhara for the partial fulfillment of the
requirement for the degree of Bachelor of Science in Forestry.
This thesis is dedicated to my Father Mr. Dhan Bahadur Budha, Mother
Mrs. Sumita Budha, family members and relatives.
 DECLARATION

I hereby declare that this project paper entitled "Builtup Area Simulation Using MLP Markov
Model (A case study from Chaurjahari Municipality, West-Rukum, Nepal)" is my own work
except otherwise acknowledged. I have not submitted it or any of its part to pursue other academic
degree. Errors if any are the responsibility of my own.

------------------
Milan Budha
BSc. Final year, 2019
Tribhuvan University
Institute of Forestry,
Pokhara, Nepal
Email: [email protected]

IV
LETTER OF ACCEPTANCE

V
 ACKNOWLEDGEMENTS

It was a long hard working five months to complete this ambitious thesis. This report is not my
solo effort but the synopsis of the wise ideas, experiences and suggestions from professionals. I
would like to express my sincere gratitude to many individuals and organizations whose support
and guided at different levels made me able to accomplish this research work

I am indebted to my Advisor Mr. Mahendra Singh Thapa, Assistant Professor at Institute of

Forestry, Pokhara campus (IoFPC) for providing his excellent supervision, knowledges and Ideas
for the completion of this research. I am pleased to this institute for providing me this platform to
pursue my undergraduate degree.
My immense acknowledgements goes to Himal Pun Magar, Anish Thapa and Dilip Budha for
their support in field data collection and moral help on the way to complete this Thesis.

I am obliged to United States Geological survey (USGS) for providing satellite and SRTM DEM
data , Department of Survey, Nepal, CBS, Chaurjahari Municipality, OSM and other organizations
that aided for secondary and primary data collection for the research. Without their springiness,
my research would have not been completed with this ease. I would like to express my honest
gratitude to ESRI (Environment systems research Institute) for providing the ArcMap 10.5.1, and
Clarks Labs, Clarks University for their technical support through TerrSet Geospatial Monitoring
and Modelling Software, Microsoft Office for providing such user-friendly software and tools.

A Special thank goes to my Roommate Mr. Bhuwan singh Bist, and SHIRISH GUYS for their
critical comments, suggestions and moral support during this research and on the course to
achieving undergraduate degree.

Huge thanks to my Batchmates (2071-75) and whole IoFPC Family for their support and guidance
during the different stages of study.

Finally, Once again the most important support, inspiration and encouragement that are never
noticed cannot be forgotten. My Parents, Sister Karuna and all the relatives have proved to be
invaluable in the completion of the thesis; I thank them profusely and dedicate this thesis to them.

Milan Budha
2019

VI
 ABSTRACT
Land use land cover changes are one of the foremost aspects of environmental changes caused by
human-induced activities mainly in rapidly developing areas. Land use land cover (LULC)
changes are the major issues and challenges for the eco-friendly and sustainable development for
the economic growth of any area. The rate of urbanization is increasing day by day in developing
cities parallel with infrastructure development and population pressure. Study regarding these
changes is required to document the changes for sustainable development. GIS and Remote
Sensing techniques are very useful and important for monitoring these changes. This research was
carried out in Chaurjahari Municipality of West Rukum District of Nepal. The major objectives
of this research were to generate LULC map, quantify changes and predict the future scenarios.
Landsat's high-level spectral reflectance data of the year 2001, 2011 and 2018 were used to
generate the Land use land cover map for respective years. Maximum Likelihood classifier was
adapted for classification of these images. Change analysis panel in Land Change Modeler (LCM)
was used along with ArcMap 10.5.1 to quantify and observe the change between the periods.
Explanatory Variables were obtained from reliable sources. MLP (Multi-layer perceptron Neural
Network) Markov Model has been the most accurate model to predict the future scenarios of
LULC last decades. This model was used to predict the LULC scenarios of 2030 and 2050.

Result showed that Builtup area has experienced the largest net change with 4.28% followed by
Agriculture (-4.03%), Forest (-0.96%), Bare area (0.63%) and Water (0.09%). Agriculture has
contributed most to the expansion of builtup land in the area. Calibration accuracy of MLP was
found to be 73.4% with skill measure of 0.6813. Hard predicted map of 2018 showed accuracy of
78.48%. So, the model was accepted to predict the future scenarios. Predicted map of 2030 and
2050 showed increase in builtup area with 9.02% and 12.97% coverage respectively. Agriculture
is observed to be decreasing from 42.88% in 2018 to 40.06% in 2030 and 37.37% in 2050.This
research is itself a pioneer research in Chaurjahari Municipality serve as baseline for future
researches. This could help to shape the urban planners, policy makers and related stakeholders
for better decision making in planning for future.

Keywords: MLP Markov, Future Prediction, LULC, Maximum likelihood classifier, Chaurjahari

VII
 Table of Contents
DECLARATION .....................................................................................................................IV
LETTER OF ACCEPTANCE .................................................................................................. V
ACKNOWLEDGEMENTS .....................................................................................................VI
ABSTRACT ........................................................................................................................... VII
List of Figures ........................................................................................................................... X
List of Tables ...........................................................................................................................IX
List of Annexes ......................................................................................................................... X
ACRONYMS ...........................................................................................................................XI
CHAPTER ONE
INTRODUCTION
1.1. Background ......................................................................................................................... 1
1.2 Rationale .............................................................................................................................. 3
1.3 Objectives ............................................................................................................................ 4
1.3.1 General objectives ..................................................................................................... 4
1.3.2 Specific objectives ..................................................................................................... 4
1.4 Limitations ........................................................................................................................... 4
CHAPTER TWO
LITERATURE REVIEW
2.1 Land Use Land Cover Change (LULC) ............................................................................... 5
2.2 Classification........................................................................................................................ 6
2.3 Researches in Nepalese context ........................................................................................... 7
2.4 LULC modelling ................................................................................................................ 10
CHAPTER THREE
METHODS AND METHODOLOGY
3.1 Study Area ......................................................................................................................... 14
3.1.1 Selection .................................................................................................................. 14
3.1.2 Intensive study area ................................................................................................. 14
3.1.3 Map .......................................................................................................................... 15
3.2 Data Collection .................................................................................................................. 16
3.2.1 Remotely sensed data .............................................................................................. 16
3.2.2 Topographic map ..................................................................................................... 16
3.2.3 Auxiliary data .......................................................................................................... 16

VIII
3.3 Data Analysis ..................................................................................................................... 17
3.3.1 Data processing ........................................................................................................ 17
3.3.2 Image classification ................................................................................................. 17
3.3.3 Land cover change detection ................................................................................... 19
3.3.4. Change prediction ................................................................................................... 19
CHAPTER FOUR
RESULTS AND DISCUSSIONS
4.1 Land Use Land cover Change ............................................................................................ 25
4.1.1 LULC 2001 .............................................................................................................. 25
4.1.2 LULC 2011 .............................................................................................................. 26
4.1.3 LULC 2018 .............................................................................................................. 27
4.1.4 Accuracy assessment ............................................................................................... 28
4.1.5 Land use land cover change ..................................................................................... 28
Overall Change between 2001, 2011 and 2018 ................................................................ 30
4.2 LULC Change Prediction .................................................................................................. 32
4.2.1 Model Building, Prediction and Validation ............................................................. 32
4.2.2. Future Prediction .................................................................................................... 36
CHAPTER FIVE
CONCLUSION AND RECOMMENDATIONS
5.1 Conclusion ......................................................................................................................... 41
5.2 Recommendations .............................................................................................................. 42
References ................................................................................................................................ 43
ANNEXES ............................................................................................................................... 58

List of Tables
Table 1: List of country-scale land use and land cover (LULC) data sources and mapping in Nepal
.................................................................................................................................................... 9
Table 2. Intensive study area ................................................................................................... 14
Table 3: Details of Satellite data used in the study .................................................................. 16
Table 4. Details about Land use land cover (LULC) classes. .................................................. 18
Table 5: Driver Variable Details .............................................................................................. 20
Table 6: Class specific LULC of Chaurjahari Municipality in 2001 ....................................... 25

IX
Table 7: Class specific LULC of Chaurjahari Municipality in 2011 ....................................... 26
Table 8: Class specific LULC of Chaurjahari Municipality in 2018 ....................................... 27
Table 9: Accuracy assessment results for different classified images ..................................... 28
Table 10: Land use Land cover Change between 2001 and 2011 ........................................... 29
Table 11: Land use Land cover change between 2011 and 2018 ............................................ 29
Table 12: Transition probability matrix for 2018 .................................................................... 33
Table 13: Agreement/Disagreement between Predicted and actual LULC of 2018. ............... 35
Table 14: Transition probability matrix for 2030 and 2050 .................................................... 36
Table 15: Overall change 2001 to 2050 ................................................................................... 40

List of Figures
Figure 1. Satellite image classification methods........................................................................ 6
Figure 2. Architecture of the Multi-layer perceptron network applied to land use change modelling
.................................................................................................................................................. 11
Figure 3: Map of Study area .................................................................................................... 15
Figure 5: Conceptual Framework for Future Prediction using LCM ...................................... 24
Figure 6: LULC Map of 2001 .................................................................................................. 25
Figure 7: LULC map of 2011 .................................................................................................. 26
Figure 8: LULC map of 2018 .................................................................................................. 27
Figure 9: Overall change in Percentage between 2001, 2011 and 2018. ................................. 30
Figure 10: Class wise contribution to Builtup area Expansion ................................................ 31
Figure 11: Transition Potential maps. ...................................................................................... 32
Figure 12: Predicted LULC map 2018..................................................................................... 34
Figure 13: Predicted LULC Map of 2030 ................................................................................ 37
Figure 14: Pie-chart showing the Percentage coverage of each LULC Classes in Predicted map
2030.......................................................................................................................................... 37
Figure 15: Predicted LULC map of 2050 ................................................................................ 38
Figure 16: Pie chart showing the coverage of each LULC Classes in Predicted map 2050. ... 38
Figure 17: Trend of Expansion in Builtup Area (2001-2050) ................................................. 39
Figure 18: Annual change rate (2001-2050) ............................................................................ 39

List of Annexes
Annex-I Description of Topographic Map used in the study .................................................. 58
Annex-II Accuracy Assessment ............................................................................................... 59
Annex-III Land change matrix................................................................................................. 60
Annex-IV ................................................................................................................................. 61
Annex-V MAPS ....................................................................................................................... 62

X
 ACRONYMS
CBS Central bureau of statistics

CA Cellular automata

DDC District Development Committee

DEM Digital elevation Model

DN Digital Number

ESRI Environment Systems Research Institute

ETM Enhanced Thematic mapper

ETM+ Enhanced Thematic Mapper plus

GIS Geographical information system

GSFC Goddard Space Flight Center

IoFPC Institute of Forestry Pokhara Campus

LCM Land change modeler

LEDAPS Landsat Ecosystem Disturbance Adaptive Processing System

LULC Land use land cover

MEaSUREs Making Earth System Data Records for Use in Research Environments

MLP Multi layer Perceptron

MLC Maximum Likelihood Classifier

MODIS Moderate Resolution Imaging Spectroradiometer

OLI Operational Land Imager

OLI TIRS Operational Land Imager Thermal Infrared sensor

OSM Open Street Map

SRTM Shuttle Radar Topography Mission

TM Thematic mapper

USGS United states government services

UNDESA United Nations, Department of Economic and Social Affairs

VDC Village development committee

XI
 CHAPTER ONE
INTRODUCTION

1.1. Background
Land cover is defined by the attributes of the earth’s land surface captured in the distribution of
vegetation, water, desert and ice and the immediate subsurface, including biota, soil, topography,
surface and groundwater and it also includes those structures created solely by human activities
such as mine exposures and settlement (Lambin et al., 2003; Baulies and Szejwach, 1997). Land use
is the intended employment of land management strategy placed on the land cover by human
agents or land managers to exploit the land cover and reflects human activities such as industrial
zones, residential zones, agricultural fields, grazing, logging and mining among many others
(Zubair, 2006). Land use land cover (LULC) change is a dynamic process-taking place on the
biophysical surfaces that have taken place over a period and space is of enormous importance in
natural resource studies. LULC change dynamics are important elements for monitoring,
evaluating, protecting and planning for earth resources. These changes are the major issues and
challenges for the eco-friendly and sustainable development for the economic growth of any area.
With the population explosion, human activities such as deforestation, soil erosion, global
warming and pollution are very harmful for the environment. This causes changes in LULC with
the demand and supply of land in different activities. Change detection in LULC can be performed
on a temporal scale such as a decade to assess landscape change caused due to anthropogenic
activities on the land (Gibson and Power, 2000). Various natural and human activity processes
influence LULC change. In order to improve the economic condition of the area without further
deteriorating the bio-environment, every bit of the available land has to be used in the most rational
way. This requires the present and the past LULC data of the area (Chaurasia et al., 1996). LULC
dynamics are widespread, accelerating and significant processes being driven by human actions
but also produce changes that impact humans (Agarwal et al., 2002).

Land use and land cover are the outcome of interaction between man and environment. Some of
the land uses are directly related to cultures, and social and economic conditions of the people
(Vink, 1975). Land use and land cover change, as one of the main driving forces of global
environmental change, is central to the sustainable development debate. Land use/land cover
change has been reviewed from different perspectives in order to identify the drivers of land use

1
land cover change, their process and consequences. This problem needs to be seriously studied,
through multi-dimensional fields in order to preserve the precious and limited agricultural lands.
The increased urbanization may have several impacts on infrastructure, energy use and economy
of the country (Hegazy and Kaloop, 2015). The rapid changes of land use and cover than ever
before, particularly in developing nations, are often characterized by rampant urban sprawling,
land degradation, or the transformation of agricultural land to shrimp farming ensuing enormous
cost to the environment (Sankhala and Singh, 2014).

Change detection in land use and land cover is the measure of the separable data layout and
noticeable change in information that can lead to more visible insight into subtle process enclosing
land cover and land use changes than the data observed from the usual change’ (Singh et al., 2013).
Change detection is the procedure of discovering and observing the difference in a body or
phenomenon by visualizing at difference times (Mouat et al., 1993).

Information ‘regarding land cover is essential to outdo the problems arising from regular,
uncontrolled progress and destruction of environmental quality, loss of basic water bodies and
their associated life forms while planning the national development’ (Anderson et al., 1976). ‘A
high quality understanding is necessary to ensure sustainable land use management’ (Latham,
2014). To amend proper development plans a highly maintained and timely managed data
regarding land cover and detection is must (Dhinwa et al., 1992).

Application of remotely sensed data made possible to study the changes in land cover in less time,
at low cost and with better accuracy (Kachhwaha, 1985) in association with Geographical
Information System (GIS) that provide suitable platform for data analysis, update and retrieval
(Star et al. 1997; Chilar 2000). Space borne remotely sensed data may be particularly useful in
developing countries where recent and reliable spatial information is lacking. Remote sensing
technology and GIS provide efficient methods for analysis of land use issues and tools for land
use planning and modeling. By understanding the driving forces of land use development in the
past, managing the current situation with modern GIS tools, and modeling the future, one is able
to develop plans for multiple uses of natural resources and nature conservation. The change in any
form of land use is largely related with both the external forces and the pressure builtup within the
system (Bisht and Kothyari, 2001).

2
For all the models available, Markov model has been widely used to model LULC changes
including both urban and non-urban areas at large spatial scales. It studies the initial occupation
and transition probabilities of different states, accordingly to determine the trend of development
and predict the state of the future. It is widely used in the analysis of different LULC types and
dynamic changes. Its applicability and feasibility had been testified in many research papers and
its results approximately conform to observed results (Dadhich & Hanaoka 2010; Dongjie et al.
2008; Huang et al. 2008; Zhang et al. 2011).

1.2 Rationale
The impacts of rapid urbanization in the Himalayan cities, particularly sprawl and other types of
pell-mell urban growth, go beyond the urban footprints (Joshi and Tiwari, 2014). Monitoring of
land use land cover (LULC) in these cities, therefore, is not only a pragmatic way to detect and
quantify landscape-level transformation, but also a window to understanding the complex social-
ecological relationships in the region characterized by its fragility and sensitivity to hazards and
disaster, such as earthquakes, landslides (UNDESA, 2015). The present scenario of the major towns
of Nepal shows a rapidly changing pattern of urban land use (Poudel, 2008). Nepal is recorded as
one of the top ten fastest urbanizing countries in the world (UNDESA, 2015).

Accurate information on land cover change and the forces behind it is essential for designing a
sound environmental planning and management (Regmi et al., 2017). Policy makers in developing
countries face unprecedented challenges with regard to governing, urban planning and LULC
management because of the prevailing high dynamic growth. Therefore, knowledge concerning
past, current and future growth plays an important role in the decision-making process (Moghadam
and Helbich, 2013). The conversion of natural habitats into agricultural land for the maintenance
of human livelihoods has been identified as the greatest driver of global environment change
(Ramakutty and Foley, 1999). Lack of adequate research for further urban planning for developing
cities like Chaurjahari may lead to unsustainable and unmanageable crisis in future. So, Research
is needed to specify the root cause of LULC change and the variables affecting it. This research
will serve as an important benchmark for urban planners, policy makers and other researchers.

3
1.3 Objectives
1.3.1 General objectives
The general objectives of this research is to predict the LULC (Land use Land cover) change of
Chaurjahari Municipality.

1.3.2 Specific objectives

 To generate the LULC maps for 2001, 2011 and 2018 for Chaurjahari Municipality,
 To observe the LULC change between these periods,
 To predict the future (2030, 2050) Scenarios.

1.4 Limitations
Limitations were minimized as possible for better outcome of this research. Nevertheless, as
always there are always certain limitations for the study.

Some major limitations are listed below:

 Lack of adequate digital data for perfect variable analysis,
 Time, cost and resource limitations,
 Low accuracy in secondary data collected,
 Lengthy process in analysis but limited time.

4
 CHAPTER TWO
LITERATURE REVIEW

2.1 Land Use Land Cover Change (LULC)

Land use and land cover change is an important tool to assess global change in different
spatiotemporal scales (Lambin, 1997). It is a widespread, accelerating, and significant process,
driven by human actions, and, in many cases, it also drives changes that affect humans (Agarwal
et al., 2001). The impacts of LULC change on the sustainability of the ecosystems are becoming
increasingly important issues in global changes research. Human actions seem to lead to the
greatest changes in the current state of the earth’s surface. Alterations in the surface cover result
in changes to the balance of energy, water, and the geochemical fluxes at the local, regional and
global level and these changes will inevitably influence the sustainability of natural resources and
socio-economic activities (Vescovi et al., 2002). With the increase in growth of population, pressure
on limited natural resources of a country increases and contributes to the changes in land cover.
Land cover corresponds to the physical condition of the ground surface, such as forest, grassland,
agriculture land and so on, while land use reflects human activities like the use of the land for
different purposes such as industrial zones, residential zones, and agricultural fields. There is a
direct link between land cover and the actions of people in their environment, i.e. land cover
change may result from land use (Phong, 2004). Changes in land-use and land cover have impacts
on soil and water quality, biodiversity, and global climatic systems and, thus, have important
consequences on natural resources (Awasthi et al., 2002). Land use involves the manner in which
the natural land cover is manipulated by a human agent as well as the intent of manipulation or
the purpose for which the land is utilized. The Land use pattern is a resultant of the interaction of
various physical, socio-economic, environmental and political factors and people’s
responsiveness towards these forces. The shift in the intent and utilization of the existing land use
in response to the economic, physical, environmental or other factors constitute the land use
change.

Land use Land cover change information has an important role to play at local and regional as
well as at macro level planning. The planning and management task is hampered due to
insufficient information on rates of LULC. The land cover changes occur naturally in a progressive
and gradual way, however sometimes it may be rapid and abrupt due to anthropogenic activities.

5
The growing population and human activities are increasing the pressure on the limited land and
soil resources for food, energy and several other needs. As the population increases particularly
in the urban areas by attracting job opportunities and city spreads outward from its limit,
encroachment on the surrounding available land starts. Due to increasing number of population,
agricultural land starts converting into built up area and forest areas starts converting into
agricultural land, built up etc. Thus, spatial and temporal analysis technologies are very useful in
generating scientifically based statistical spatial data for understanding the land ecosystem
dynamics (Singh et al, 2013).

2.2 Classification
Classification involves mathematical grouping of pixel values (pre-processed DN) into themes,
which correspond to particular land cover types on the Earth’s surface (Venkataratnam, 1980;
Schowengerdt, 2012). Land cover classification involves the discrimination of land cover types
through different classification methods developed in the field of remote sensing (Ahmad et al.,
1992; Lu and Weng, 2007). The development of these methods was largely attributed to the
improvements in Landsat images, advancement of computer technology, development of
geographic information systems (GIS) and the Landsat free access policy (Steiner, 1970;
Thompson and Mikhail, 1976). Land cover is the physical substance covering the Earth’s surface,
for example forests, water and grasslands (Campbell and wynne, 2011). Land cover classification
using Landsat images has evolved over the last four decades.

Figure 1: Satellite image classification methods

6
Darius Phiri and Justin Morgenroth (2017) classified different methods of land cover classification
of Landsat data into six major types based on temporal and developmental basis. These are:
(a)Early Landsat Land Cover Classification: Visual Approach,
(b) Landsat Land Cover Classification Using Digital Format,
(c) Developments of Computer-Based Land Cover Classification Methods,
(d) Pixel-Based Classification,
(e) Sub-Pixel Image Classification,
(f) Object-Based Approach.

2.3 Researches in Nepalese context

Land use and Land cover in Nepal has undergone constant change over the past few decades due
to the major changes caused by anthropogenic and natural factors and their impacts on the national
and regional environment and climate (Paudel et al., 2016). Nepal is a mountainous country,
covering two thirds of the Himalayan region (Rokaya et al., 2012), where the main economic
activities are based on agriculture. Nepal comprises five physiographic regions: High Mountain,
Middle Mountain, Hill, the Shiwalik range and Tarai (LRMP, 1986). The Middle Mountain and
High Mountain regions are more sensitive to LUCC, and are more seriously affected by it, even
with small changes, than the low land area of Tarai (Khanal, 2002). This type of impact is not
limited to regions where change takes place, but also easily spreads with further impact in the
plains and low land areas, due to the high gradient of Nepal’s mountain slopes (Becker and
Bugmann, 2001).

With regard to the overall historical records of the sectors of LULC mapping in Nepal, the first
attempt was made in 1964 by the forest resource survey office (FRSO) in the form of forest cover
mapping based on 1953–58 and 1963–64 aerial photographs at the national level. That project
mapped forestland using visual interpretation of aerial photographs with field verification, with
the forest categorized into commercial and noncommercial forest (Acharya and Dangi, 2009).
Aerial photographs (1978/79 and 1985) were the basis of another national-scale mapping of forest
cover change conducted as a major part of LULC in 1985 (Nield, 1985).

7
The first detailed LULC mapping was carried out in 1986 by the Land Resource Mapping Project
(LRMP), which created a number of datasets (geology, land system, land utilization and land
capability) on the basis of 1978/79 aerial photographs at a scale of 1:50,000 (LRMP, 1986). It
mapped all LULC categories with reports on each aspect. Based on LRMP records and forest data
from FRSO, the Ministry of Forests implemented a master plan for the forestry sector (MPFS) in
1986, which was designed to update the LRMP information (MPFS, 1989a; 1989b). The Nepal
government survey department prepared and published a topographical map of Nepal based on
aerial photographs taken in 1989 for the eastern part, and in 1996 for the western part of the
country, at two different scales (1:25,000 in Tarai and the Hill area and 1:50,000 in the Mountain
area) (Chhatkuli, 2004). This showed different categories of LULC for the whole country. The
National Forest Inventory (NFI) was launched in 1990 to measure forestland change and was
completed in 1994 (Gautam et al., 2004). NFI used Landsat satellite imagery, aerial photographs
and field measurements to discover forest change scenarios in Nepal. National-level detailed land
cover mapping has recently been completed (Uddin et al., 2015), and a land cover database of
Nepal was developed in 2010 using Landsat TM 30m resolution satellite imagery. Major national-
level mapping of LULC is set out in Table 1. In addition, there have been attempts to map LULC
in Nepal on the scale of regions, districts, watersheds, river basins and micro-areas. Some of these
have been, published and some are unpublished reports or thesis work. Most research in Nepal
has used aerial photographs and satellite imagery for LULC mapping at watershed and river basin
levels. Khanal, 2002 developed LULC map in the Madi watershed in 2002 and assessed LUCC
scenarios of 1956–1996 (Khanal, 2002). Another satellite imagery-based forest cover mapping of
Tarai’s 20 districts was carried out by the Department of Forest and Soil Conservation in 2005
(MoFSC, 2009), covering the periods 1990/91 and 2000/01. Similarly, satellite imagery-based
mapping of the Koshi river basin was carried out in 2000 (Sharma et al., 2000) and 2012 (Gao,
2012). These two studies presented detailed mapping of LULC in 1992, 2000 and 2010 in the
Koshi river basin.

8
Table 1: List of country-scale land use and land cover (LULC) data sources and mapping in
Nepal

Research types Spatial coverage Data resources Temporal Sources/References

coverage
LULC Whole of Nepal Aerial photographs 1978/79 and (Nield, 1985)
1985/86
LULC Whole of Nepal Aerial photographs 1978/79 (LRMP, 1986)
LULC Whole of Nepal Aerial photographs 1986 (MPFS 1989a, 1989b)
Topographical Whole of Nepal Aerial photographs 1989 to (Chhatkuli, 2004)
Map 2001
Forest land Whole of Nepal Aerial photographs, 1990-1994 (Gautam et al., 2004)
(NFI) Landsat satellite
imagery
Forest land Whole of Nepal Aerial photographs 1953-58 and (Acharya and Dangi,
(FSRO) 1963/64 2009)
Snow cover Whole of Nepal Satellite imagery 2000-2005 (Shrestha and Joshi,
2009)
Glacier cover Whole of Nepal Aerial photographs, 2001 and (Bajracharya et al.,
Topo maps and 2010 2011)
satellite imagery
Glacier cover Whole of Nepal Landsat satellite 1980, 1990, (Bajracharya et al.,
imagery 2000 and 2014)
2010
Land cover Whole of Nepal Landsat satellite 2010 (Uddin et al., 2015)
imagery
Land cover Global Landsat satellite 2010 (Chen et al., 2015)
imagery
Source: (Paudel et al., 2016)

9
2.4 LULC modelling
Many researchers have attempted to develop and describe different approaches for taking land
cover change models for further advancements (Turner, 1987; Baker, 1989; Singh, A. 1989; Coppin
and Bauer, 1996; Lambin, 1997; Mas, 1999; Lambin et al., 2003; Theobald and Hobbs, 1998, Brown
et al., 2000; Daryei, 2003; Halil et al., 2006). Different approaches have been attempted in spatial
modelling in recent years such as fractal models (De Cola, 1989; Rees, 1992; De Jong and Burrough,
1995), artificial neural network models (Maier and Dandy, 2000; Li and Yeh, 2002; Pijanowski et
al., 2002; Mas et al., 2004), CA models (Deadman et al., 1993; Clarke et al., 1997; Singh, 2003) and
agent based models (Sudhira, 2004).

These approaches are often combined together to create a hybrid model. With recent advances in
geographic information systems (GIS), integration of land-use change models into GIS has been
possible. Of these models, cellular automata and hybrid models are used widely for land-use
change modeling (Wood et. al, 1997; Breur et al., 2006). Land-cover mapping determines the
current composition and distribution of land surface attributes, and this is subsequently used as
the basis for assessing future change (McDermid et al., 2005; Miller and Rogan, 2007; Schulz et al.,
2010; Carmelo et al., 2012). However, modeling the complex dynamic systems of the land surface
features has been challenging (Ahmed, 2011). Some popular tools have been developed to model
urban growth and land cover changes including GEOMOD (Pontius and Spencer, 2005), SLEUTH
(Silva and Clarke, 2002), Land Use Scanner (Hilferink and Rietveld, 1999), Environment Explorer
(Verburg et al., 2004), SAMBA (Castella et al., 2005), Land Transformation Model (Pijanowski et
al., 2000), and CLUE (Kok and Veldkamp, 2001). These tools make use of a number of methods
such as Markov Chain (Balzter, 2000), Cellular Automata (Sante et al., 2010), Logistic Regression
(McConnell et al., 2004), and Artificial Neural Network (ANN) (Civco, 1993). More details on the
characteristics of each tool are discussed in the literature (Pontius et al., 2008). This study used the
Multi-layer Perceptron (MLP), a type Neural Network method with more than one hidden layer,
in order to model and project future land cover change scenarios (Ahmed and Ahmed, 2012). The
MLP decides about the parameters to be used and how they should be changed to better model the
data. It undertakes the classification of remotely sensed imagery using the back propagation
algorithm. The MLP also performs a non-parametric regression analysis between input variables
and one dependent variable with the output containing one output neuron (Atkinson and Tatnall,
1997).

10
Multi-Layer Perceptron Markov Model
The term “Artificial Neural Network (ANN)” has been inspired by human biological nervous
system (karul and soyupak, 2006). In a typical ANN model, simple nodes are connected together
to form a network of nodes. Some of these nodes are called input nodes; some are output nodes
and in between there are hidden nodes (Atiknson and Tatnall, 1997). Multi-Layer Perceptron
(MLP) is a feed-forward Neural Network with one or more layers between input and output layers.
The great advantage of using MLP perceptron neural network is that it gives the opportunity to
model several or even all the transitions at once (Eastman, 2009).

MLP is an artificial neural network model that maps sets of input data onto a set of appropriate
outputs. It is composed of an input layer, output layer, and hidden layers, which are between input
and output layer (Figure 2). MLP works in such a way that activity of the input units represents
the raw information that is fed into the network. The behavior of the output units depends on the
activity of the hidden units and the weights between the hidden and output units. The activity of
each hidden unit is determined by the activities of the input units and the weights on the
connections between the input and the hidden units (Bayes, 2012; Basyal, 2013; Sahalu, 2014; Gala
et al., 2015).

Figure 2: Architecture of the Multi-layer perceptron network applied to land use change modelling

11
The Feed-Forward Concept of Multi-Layer Perceptron Neural Network
MLP neural network uses the back propagation (BP) algorithm. The calculation is based on
information from training sites (Eastman, 2009). Back propagation involves two major steps,
forward and backward propagation. The input that a single node receives is weighted as:

netj = Σ wjiOi (1)

where, wij = the weights between node i and node j; Oi = the output from the node i.
The output from a given node j is computed as:

Oi= f(netj) (2)

f = a non-linear sigmoid function that is applied to the weighted sum of inputs before the signal
passes to the next layer.
This is known as “Forward Propagation”. Once it is finished, the activities of the output nodes are
compared with their expected activities. In normal circumstances, the network output differs from
the desired output (a set of training data, e.g., known classes). The difference is termed as the error
in the network (Atiknson and Tatnall, 1997). The error is then back propagated through the
network. Now the weights of the connections are corrected as follows:

Δwji(n+1)= η (δjOi)+ α Δwji (3)

η = the learning rate parameter; δj = an index of the rate of change of the error; α = the momentum
parameter.

The process of the forward and backward propagation is repeated iteratively, until the errors of
the network minimized or reaches an acceptable magnitude (Atiknson and Tatnall, 1997). The
purpose of training the network is to get proper weights both for the connection between the input
and hidden layer, and between the hidden and the output layer for the classification of unknown
pixels (Eastman, 2009). Several factors affect the capabilities of the neural network to generalize
(Atiknson and Tatnall, 1997). These include:

12
Number of Nodes
In general, the larger the number of nodes in the hidden layer, the better the neural network
represents the training data (Atiknson and Tatnall, 1997).The number of hidden layer nodes is
estimated by the following equation:

Nh=INT(√Ni*No) (4)
where, Nh = the number of hidden nodes; Ni = the number of input nodes; No = the number of
output nodes

Number of Training Samples and Iterations

The number of training sample also affects the training accuracy. Too few samples may not
represent the pattern of each category while too many samples may cause overlap. Again too many
iterations can cause over training that may cause poor generalization of the network (Eastman,
2009). Over training can be prevented by early stopping of training (Karul and soyupak, 2006).
The acceptable error rate is evaluated based on the Root Mean Square (RMS) Error:

RMS=Σ(ei)2/N= Σ(ti- ai)2 (5)

where, N = the number of elements; i = the index for elements; ei = the error of the ith element; ti
= the target value (measured) for ith element; ai = the calculated value for the ith element.

13
 CHAPTER THREE
METHODS AND METHODOLOGY

3.1 Study Area

Chaurjahari Municipality, located in West Rukum of Karnali province now, was Chaurjahari
Municipality (Bijyeswori and Kotjahari), Kholagaun and some part of nuwakot and Purtimkada
VDC then. It was declared as Municipality by coalision of two VDCs Bijeyswori and Kotjahari
in 2072/06/01. It has an area of 107 square kilometer and population 27438 in census 2011.It is
one of the 10 new cities as planned by Government of Nepal along the midhills pusphalal highway.
Located at the bank of Bheri River, it is and considered as main business center to surrounding
Salyan, Jajarkot and Dolpa districts. Plain alluvial soil and flatness is the potential provider of
food security to the region (Chaurjahari Municipality, 2018). Tribeni Rural Municipality (in the
east), Sani bheri Rural Municipality (in North), Bherimalika Municipality (in West) and
Kumakhmalika Rural Municipality, Darma Rural Municipality (in the south), surround
Chaurjahari Municipality.

3.1.1 Selection
Study area was selected based on few assumptions and critical overview:
 It is one of the urbanizing city in Nepal with great potential,
 Lack of any studies carried out on this topic ,
 Feasibility with respect to researchers time, cost, public relations and miscellaneous.

3.1.2 Intensive study area

Further details regarding the study area is mentioned in table below.

Table 2. Intensive study area

Geographical Coverage
˚
Latitude: 28 34'45.62''N to 28˚41'29.56''N Longitude:82˚11'4.63''E to 82˚21'22.51''E
Based on latest Provincial structure of Nepal
Previous Municipality/VDC Previous ward No. New ward No.
Chaurjahari Municipality 1,2 1
Chaurjahari Municipality 3 2
Chaurjahari Municipality 4,5 3

14
 Chaurjahari Municipality 6 4
Chaurjahari Municipality 7 5
Chaurjahari Municipality 8 6
Purtimkada 4,5
Chaurjahari Municipality 9 7
Chaurjahari Municipality 10 8
Purtimkada 9
Chaurjahari Municipality 11 9
Kholagaun 1,2,9 10
Kholagaun 3,4,5 11
Kholagaun 6,7,8 12
Nuwakot 5,6,7,8 13
Nuwakot 1,2,4,9 14
Source: (Chaurjahari Municipality, 2018)

3.1.3 Map

Figure 3: Map of Study area

15
3.2 Data Collection
3.2.1 Remotely sensed data
Landsat's Surface reflectance data on demand, of different time 2001, 2011 and 2018, which
spatial resolution is 30m*30m and are atmospherically corrected (Young et al., 2017) were
downloaded from USGS website (https://earthexplorer.usgs.gov/).These datas (Table 3) were
downloaded based on low cloud cover and seasonal fluctuations. Due to this reason, imageries
from winter season were main sources of data for this study (Uddin and Gurung, 2010).

Table 3: Details of Satellite data used in the study

Satellite Sensor Path/ Data No. of Data source Cloud
ID row acquired bands cover (%)
Landsat 7 ETM 143/40 2001-02-08 8 https://earthexplorer.usgs.gov/ 4
Landsat 7 ETM 143/40 2010-12-18 8 https://earthexplorer.usgs.gov/ 1
Landsat 8 OLI_TIR 143/40 2018-01-14 11 https://earthexplorer.usgs.gov/ 2.77
S

3.2.2 Topographic map

Topographic map of 1999 with 1:25000 scale were obtained from http://pahar.in/nepal-topo-maps/
(Pahar Website). Study area overlays on four topo sheets. Therefore, these topo sheets were
Georeferenced and mosaicked to cover the area. Finally, clipped topo map was obtained to assist
in further analysis in the research. Details about these sheets is presented in Annex-I.

3.2.3 Auxiliary data

Field data for training sites for signature file creation and accuracy assessment were collected. In
case of field verification, for 2001, Topographic map was used. For 2011, High-resolution google
earth satellite image was used. For present 2018, High-resolution google earth satellite image and
field data were used.

Based on expert's opinions, literature review and researchers critical thinking major variables
directing the LULC change were summed up and dates for those variables were collected
accordingly. These major variables are described more detailed in Table 5.

16
3.3 Data Analysis
3.3.1 Data processing
The Surface Reflectance data product is generated from specialized software called
Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS). LEDAPS was
originally developed through a National Aeronautics and Space Administration (NASA) Making
Earth System Data Records for Use in Research Environments (MEaSUREs) grant by NASA
Goddard Space Flight Center (GSFC) and the University of Maryland (Masek et al., 2006).

These high-level products provided by USGS are atmospherically corrected (Young et al., 2017).
So, these data were only enhanced for classification. Landsat 7 ETM+ inputs are not gap-filled in
Surface Reflectance production, and gapped areas are not processed for Surface Reflectance.
Therefore, 2011 satellite image was scanline error removed and correction were carried out to
convert the DN to surface reflectance. For specific time series, layers were stacked to form a
composite band using ArcMap 10.5.1.These composite layer formed were clipped to vector layer
of Chaurjahari Municipality. ArcMap 10.5.1. was used for the preprocessing of Variables
specified in Table 5.

3.3.2 Image classification

Training samples are representative of the desired land use classes (Magidi, 2010). These samples
were determined based on ground truthing, researcher’s personal experience and physiographical
knowledge of the study area (Jensen, 2007). On average, 25 to 30 training samples for each land
cover class were selected. All the images were analyzed with respect to their spectral and spatial
distribution in order to develop the training sites (Ahmed, 2011). The statistical signatures of each
land cover class (Table 4) were developed. Maximum likelihood classifiers (Wernick & Morris,
1988) was selected for supervised classification. This procedure has proven to be a robust and
consistent classifier for multidate classifications (Wu et al., 2006). Vegetation Indices and visual
interpretation were also used to improve the accuracy of the classification. Classification toolbar
inside ArcMap 10.5.1 was used for classification of satellite images of these three dates. Post-
classification smoothing was applied by a 3*3 grid-cell majority filter in the maps generated from
unsupervised, supervised and GIS post-processing approaches before the accuracy assessment
(Thapa and Murayama, 2009).

17
Table 4. Details about Land use land cover (LULC) classes.
Land Use Land Cover Class Land use and Land cover included in class
Forest Permanent tree covered area, Grassland with
scattered trees
Agriculture Vacant and cultivated Croplands
Water River and other water resources
Bare Area River banks, Bare ground, Landslide zones
Builtup Residential, industrial, commercial, airports,
roads/highways, Settlements
(Structure of all types related to urban)

Accuracy Assessment
Classification accuracy refers to the extent of correspondence between the remotely sensed data
and reference information (Congalton, 1991). One of the most common means of expressing
classification accuracy is the preparation of classification error matrix (Lillesand and Kiefer, 2007).
Assessment of classification accuracy of 2001, 2010 and 2018 LULC images was carried out to
determine the quality of information derived from the data. If the classification data are to be
useful in detection of change analysis, it is essential to perform accuracy assessment for individual
classification (Owojori and Xie, 2005). 150 Random points were created for 5 classes for each
dates in ArcMap 10.5.1.Ground truth data were collected for these points using high-resolution
google earth image and topographic map for 2011 and 2001 respectively.. For 2018, additional
field data along with high-resolution google earth image were used to collect ground truth of
classified LULC classes.
The comparison of reference data and classification results was carried out statistically using error
matrices. In addition, a non-parametric Kappa test was performed to measure the extent of
classification accuracy, as it not only accounts for diagonal elements but for all the elements in
the confusion matrix (Rosenfield and Fitzpatirck-Lins, 1986). Kappa is a measure of agreement
between predefined producer ratings and user assigned ratings. It is calculated by a formula:

K= P(A)-P(E)/1-P(E) -----(6)

where P(A) is the number of times the k raters agree, and P(E) is the number of times the k raters
are expected to agree only by chance (Gwet, 2002; Viera and Garrett, 2005).

18
3.3.3 Land cover change detection
Change detection analysis was carried out on Landsat images of different years (i.e. 2001, 2011
and 2018) to analyze the pattern and trend of change in the study area using Land Change Modeler
(LCM) (Eastman, 2006). Using LCM requires mainly two time categorical maps and so the
classified maps 2001, 2011 and 2018 were used as inputs for the change analysis at different pairs.
This enabled to understand the gains and losses, the net transition of areas and contributions
among the land use land cover classes; and to quantify the changes that occurred from 2001 to
2011,2011 to 2018 (Eastman, 2006). Also, Cross tabulate MODULE in TerrSet Geospatial
Monitoring and modelling software (Clarkslab.org) was used to generate pixel based land use
change matrix for (2001-2011) and (2011-2018).Change map from all five to five classes between
these periods were mapped using Change analysis Panel in LCM.

3.3.4. Change prediction

There exists several land use land cover change modeling techniques, but the right simulation
model depends on the scope of study, availability of datasets, objective of the research, and the
accuracy of the prediction (Pontius et al., 2008; Ahmed, 2011). Considering accuracy and wide
acceptability, MLP Markov modeling technique (Atkinson and Tatnall, 1997) was applied in this
study to project the future changes of land cover. Ahmed and Ahmed (2012) have comprehensively
described a detailed description on the MLP Markov modeling technique and concluded this
technique more accurate than CA Markov Model and Stochastic Markov Models.
Annual change rate was calculated using the equation adapted by Lamichhane, 2008.
1/n
Rate of change (%) = [(b/a) -1]*100 (UNDP, RFDTh and FAO)
Where, b = base year data, b = final year data, n = number of years

Variable Preparation
For building model, the potential driver variables for builtup area expansion were critically
analyzed. Maximum available variables were included in the model. These variables were
preprocessed to be accepted by LCM itself. Firstly, Euclidean distance was applied to the available
vector data formats. Fuzzy set membership Function MODULE (inside TerrSet) was used to
standardize the Driver variables to common scale and suitability, where 0= Low suitability for
transition to builtup, 255=Maxm suitability for transition to builtup. Further details regarding the
processing methodology of these potential driver variables for urban expansion are summarized
in Table 5.

19
Variables were Standardized and categorized based on three fuzzy standardization functions viz:
sigmoid, J-shaped and linear functions with adjustable settings described by Araya and Cabral
(2010) and followed by Sahalu (2014).

Table 5: Driver Variable Details

Data source Raw Data Fuzzy set membership Final Driver
data Processing Function variable
Osm Secondary Euclidean Linear, symmetric function Distance from
download.geofabrik roads distance a=0, b=25, c=25, d=maxm value Secondary Roads
.de/asia/nepal.html in raster
" Tertiary " Linear, symmetric function Distance from
roads a=0,b=10,c=10, d=maxm value tertiary roads and
in raster paths
" River and " Linear, symmetric Distance from
streams a=0, b=30, c=300,d=maxm Water sources
raster value
LULC Maps Builtup " Linear, monotonically Distance from
area decreasing builtup
C=0, d=maxm value in raster
Topo maps and Public " " Distance from
Field services Public services
USGs Website DEM Sigmoidal, monotonically Elevation
(https://earthexplor decreasing
er.usgs.gov/) C=700 d=maxm value in raster
" DEM Spatial Sigmoidal, symmetric Slope
Analyst- a=0, b=1, c=40 d=90
slope tool
LULC map, All to Map layer Evidence likelihood in Variable Evidence
builtup map transformation utility in LCM Likelihood

20
3.3.4.1 Model Building, Prediction and Validation
LCM (Land change Modeler) inside TerrSet Geospatial Monitoring and Modelling software was
used for model building, validation and Prediction.

3.3.4.1.1 Model Building

Transition Potentials tab inside LCM was used to build model. This tab is designed to create
transition potential maps of acceptable accuracy in order to run the actual modeling. It provides
group transitions in to a set of sub models and to explore the potential power of explanatory
variables. Variables can be added to the model either as static or dynamic components based on
their effect to urban expansion (Eastman, 2012; Shalu, 2016). Static variables express aspects of
basic suitability for the transition under consideration, and are not changing over time. Whereas
dynamic variables are time- dependent drivers such as proximity to existing development or
infrastructure and are recalculated over time during the course of a prediction.

(a) Transition Sub-Models Status

This panel lists all transitions that exist between the two land cover maps of time 1 and time
2. The researcher should specify which transitions to be considered for producing the
transition potentials (Eastman, 2012).For achieving the urban expansion in this study, Major
transitions from all landcover classes to Builtup from 2001 to 2011 were included into the
sub-model.

(b) Variable transformation Utility

The Variable Transformation Utility panel is an optional panel for transition potential
modeling that provides a selection of commonly used transformations. The transformations
available include natural log, exponential, logit, square root, power, and evidence likelihood
(Eastman, 2016). The evidence likelihood transformation is a very effective means of
incorporating categorical variables into the analysis. Categorical variables must either be
converted into a set of Boolean (dummy) variables, or transformed using the Evidence
Likelihood transformation option, which is highly recommended. Possible transition from
all landcover classes to Builtup and LULC map of 2001 were used as input for including
the output as driver variable.

21
(c) Test and Selection of site and driver variables
This optional panel provides a quick test of the potential explanatory power of a variable.
Both quantitative and qualitative variables can be tested. However, qualitative variables
need either to be broken out to a set of separate Boolean layers or transformed with the
Evidence Likelihood transformation tool before use. The explanatory variable test
procedure is based on a contingency table analysis. The quantitative measure of association
used is Cramer’s V. It is found that variables that have a Cramer’s V of about 0.15 or higher
are useful while those with values of 0.4 or higher are good (Eastman, 2016). All the
potential variables available were tested and added to model. Variable specific Cramer's V
is presented in Annex-IV.

(d) Transition sub model structure

In this panel, the selected and tested driver variables for a specific sub-model were specified
to create transition potential maps for each sub-model in turn. Variables can be added as
static and dynamic based on their influence. Dynamic variables are time-dependent drivers
such as proximity to existing development or infrastructure and recalculated over time
during the course of prediction (Eastman, 2016).Dynamic variables in this study are:
distance from Builtup area and roads.

(e) Run Transition sub model

Multi Layer perceptron (MLP) is a feedforward artificial neural network (ANN) with one
or more layers between input and output layers. The term feedforward indicates the flow of
data is in one direction from input to output layer (forward). MLPs are widely applicable
for classification and prediction of land use changes. It can also solve non-linear separable
problems (Ahmed and Ahmed, 2012). Due to the capacity of MLP to run multiple transitions,
up to 9 at a time, it was chosen for building model for Builtup area (Eastman, 2016). After
getting satisfactory results, transitional map for the potential transition between two times
were obtained.

22
3.3.4.1.2 Model prediction and validation
(a) Model Prediction
Markov chain method under change demand modelling panel under change prediction tab
was used to generate the transition probability matrix for the year 2018 using the base years
(2001 and 2011). A transition probability matrix was obtained by cross tabulation of two
images of different time and it determines the probability of a pixel in a land-use class to
change into another class during that time (Subedi and Thapa, 2013).
For change prediction, change allocation panel inside the same tab was used. LCM offers
two modes of change prediction: hard and soft. A hard prediction is a commitment to a
specific scenario and results a landcover map with the same categories as the inputs. The
hard prediction procedure used by LCM is based on TerrSet’s multi-objective land
allocation (MOLA) module (Eastman, 2016). In contrast, the soft output is a continuous
mapping of vulnerability to change. It is simply an aggregation of the transition potentials
of all selected transitions. Logical OR by default was selected for the aggregation type for
soft prediction.

(b) Validation
Predicted map of 2018 was compared with LULC map 2018 by running a 3-way cross
tabulation inside validation panel (Eastman, 2016). The output illustrate the accuracy of the
model results where:
A | B | B = Hits (green) – Model predicted change and it changed
A | A | B = Misses (red) – Model predicted persistence and it changed
A | B | A = False Alarms (yellow) – Model predicted change and it persisted.

3.3.4.1.3. Future Prediction

Prediction of the future dates 2030 and 2050 were carried out using the same process and variables
used in model building, prediction and validation in LCM (Figure 5).

23
Figure 5: Conceptual Framework for Future Prediction using LCM

24
 CHAPTER FOUR
RESULTS AND DISCUSSIONS

4.1 Land Use Land cover Change

4.1.1 LULC 2001
For LULC 2001, Landsat 7 of the date 2001 (Table 3) was used. Maximum likelihood
classification was used for the classification of the image preprocessed.

Figure 6: LULC Map of 2001

The results showed that Agriculture (46.91%) was the major land use at this date followed by
Forest (44.41%), Bare Area (6.36%), Builtup (1.57%) and water (0.75%).
Table 6: Class specific LULC of Chaurjahari Municipality in 2001
Land use class Area in Hectares Percentage
Forest 4751.91 44.41
Agriculture 5019.08 46.91
Water 80.5258 0.75
Bare Area 680.353 6.36
Builtup 168.119 1.57
Total 10700 100

25
4.1.2 LULC 2011
2011 LULC map (Figure 7) was obtained after hybrid classification of the satellite image of the
date (Table 3) similar to land use land cover map for 2001 date.

Figure 7: LULC map of 2011

The results showed that Forest (47.94%) was the major Landuse class followed by Agriculture
(43.42%), Bare area (4.26%), Builtup (3.61%) and water (0.77%) (Table 7).

Table 7: Class specific LULC of Chaurjahari Municipality in 2011

Land use class Area in Hectares Percentage
Forest 5129.87 47.94
Agriculture 4645.49 43.42
Water 82.5204 0.77
Bare Area 455.832 4.26
Builtup 386.288 3.61
Total 10700 100

26
4.1.3 LULC 2018
Landsat 8 satellite image was classified using Isocluster followed by Maximum Likelihood
classification method. LULC Map of the Municipality obtained from the classification is figured
below.

Figure 8: LULC map of 2018

The results showed that Forest (43.85%) was the major Landuse class followed by Agriculture
(42.88%), Bare area (6.99%), Builtup (5.85%) and water (0.84%) (Table 8).

Table 8: Class specific LULC of Chaurjahari Municipality in 2018

Land use class Area in Hectares Percentage
Forest 4648.7 43.45
Agriculture 4588.04 42.88
Water 89.558 0.84
Bare Area 748.048 6.99
Builtup 625.654 5.85
Total 10700 100

27
4.1.4 Accuracy assessment
The accuracy of the classified image is very crucial part in any image analysis. The limitation of
utilizing remote sensing data in mountainous terrain are topographic shadow, steep slopes, cloud
cover etc. (Shrestha and zinck, 2001) which reduces the accuracy of the image classification.
Spatial resolution of the image and cost is also an important limiting factor in developing nation
like Nepal. The accuracy report is presented in Annex-II.

Table 9: Accuracy assessment results for different classified images

LULC 2001 LULC 2011 LULC 2018
Landuse class
UA PA KC UA PA KC UA PA KC
Forest 79.7101 77.4648 86.8421 84.6154 90.625 82
0.640697283

0.682395644
Agriculture 75.7576 79.3651 75.8621 78.5714 78.7879 90

0.74217575
Water 100 75 75 100 66.6667 67
Bare Area 71.4286 83.3333 75 66.6667 100 88
Builtup 80 66.6667 75 75 70 70
Overall Accuracy 78.66667 81.33333 84
UA=Users Accuracy, PA= Producers Accuracy, KC= Kappa Coefficient

Maximum classification was observed in 2018 followed by 2011 and 2001 Land use land cover
maps (Table 9).

4.1.5 Land use land cover change

4.1.5.1. Change between 2001 and 2011
Change analysis was done using TerrSet Geospatial Monitoring and modelling software
(Clarkslabs.org). The results showed that forest followed by builtup area and water has increased
while agriculture and bare area decreased at this period. From table 10, highest change is seen in
forest (increased by 3.53%) and lowest in water (increased by .02%). Change matrix was
computed for the period using Cross Tabulate MODULE and presented in Annex-III. This matrix
gave the transition between all classes from 2001 to 2011. Map representation of the change
between each Land use classes and the matrix is presented in Annex-V.

28
Table 10: Land use Land cover Change between 2001 and 2011
2001 2011 Change
Land use Change
Area in % Area in %
class rate (%)
Hectares cover Hectares cover Ha %
Forest 4751.91 44.41 5129.87 47.94 377.96 3.53 0.77
Agriculture 5019.08 46.91 4645.49 43.42 -373.59 -3.49 -0.77
Water 80.5258 0.75 82.5204 0.77 1.99 0.02 0.26
Bare Area 680.353 6.36 455.832 4.26 -224.52 -2.10 -3.93
Builtup 168.119 1.57 386.288 3.61 218.17 2.04 8.68

The overall change rate per category yearly is maximum for Builtup area that is 8.68% per year.
This may be due to rapid in migration of peoples from villages to sub urban places for their
subsistence then. Bare area has gone decreasing at a rate of -3.93% per year, followed by
Agriculture with rate -0.77% per year. Forest has gone increasing at a rate of 0.77% per year,
while water seems to be somewhat constant.

4.1.5.2 Change between 2011 and 2018

Change between 2011 and 2018 was analyzed using Normal Ms. Excel Table and Change analysis
panel in TerrSet. It showed that highest change is experienced by forest (-4.49%) followed by
Bare area (2.73%), Builtup (2.24%), Agriculture (-0.54%) and water (0.07%).Forest has decreased
for this date compared to previous period. Class based change in this time period is presented in
Annex-V alongside with Change map for the period.

Table 11: Land use Land cover change between 2011 and 2018
2011 2018 Change
Land use
Area in % Area in % Change
class
Hectares cover Hectares cover Ha % rate (%)
Forest 5129.87 47.94 4648.7 43.45 -481.17 -4.49 -1.40
Agriculture 4645.49 43.42 4588.04 42.88 -57.45 -0.54 -0.18
Water 82.5204 0.77 89.558 0.84 7.04 0.07 1.25
Bare Area 455.832 4.26 748.048 6.99 292.22 2.73 7.34
Builtup 386.288 3.61 625.654 5.85 239.37 2.24 7.15

29
For this period, change rate for Builtup area has slightly decreased compared to previous time
(2001-2011) with a rate of 7.15% per year. Agriculture is decreasing at rate of -0.18% per year.
Forest decreased in this time with a rate of -1.4%, conversion to agricultural land, and Builtup due
to expansion of roads and different developmental activities may be the reasons for this. Bare area
has gained its area of 2001 with an increasing rate of 7.34% at this time.

4.1.5.3. Overall Change between 2001, 2011 and 2018

Per category change

4 3.53
3 2.73
2.04 2.24
2
1
changed Percentage

-0.54 0.02 0.07

0
Forest Agriculture Water Bare Area Builtup
-1
-2
-2.1
-3
-4 -3.49
-5 -4.49
LULC Classes

Change (2001-2011) Change (2011-2018)

Figure 9: Overall change in Percentage between 2001, 2011 and 2018.

Increase in first period in forest may be due to Maoist influence in the area. Migration to safe
places caused their land fallow for years and converted to grassland/shrub land, which are
classified as Forest. Decrease in later period may be due to encroachment, urban expansion
activities like roads and other development activities. Similar results for Forest cover is concluded
by Rijal et al., 2018. Agriculture is gradually decreasing in both periods contributing to urban
expansion and bare area. Lack of adequate manpower for cultivation due to foreign and urban
opportunities for employment may be major causes for the declining trend of Agriculture. Water
is somewhat constant in the area, slice changes may be due to shifting of Bheri River at the
boundary of the area. Bare area follows decreasing and then increasing trend, due to contribution

30
of area to builtup in 2011.Builtup is increasing in both periods by 2.04% (2001-11) and 2.24%
(2011-18).This increasing pattern is contributed by the Agriculture, Bare area and Forest. Mid hill
Puspalal highway may be the major cause of expansion in urban areas. Change map between all
to all categories obtained from change analysis panel of TerrSet is presented in Annex-V. Change
matrix showing the transition between classes for these periods were obtained by using Cross
Tabulate MODULE in TerrSet presented in Annex-III.

Contribution to Builtup Expansion

250
210.71 206.67
200
Area in hectares

150

100

50 42.14
27.80 21.25
11.39
0.18 0.36
0
Forest to Builtup Agriculture to Water to Builtup Bare Area to
Builtup Builtup
2001-2011 2011-2018

Figure 10: Class wise contribution to Builtup area Expansion

Agricultural land contributes maximum to expansion in builtup land in area (Rimal, 2011;
Ishtiaque, 2017; Rijal et al., 2018; Zhang et al., 2018). Provision of government for using private
lands by owner themselves without government restriction might be the reason behind maximum
conversion of Agriculture land to builtup (Ishtiaque, 2017).

31
4.2 LULC Change Prediction
4.2.1 Model Building, Prediction and Validation
MLP Markov model has been found the most appropriate method for future prediction of urban
areas (Ahmed and Ahmed, 2012).so, this model present inside LCM was used in this study for
model building and future prediction. Potential transition to builtup from all other classes were;
Forest to builtup, Agriculture to Builtup, Bare Area to Builtup and water to Builtup. Due to very
small area (>1 hectares) from water to builtup for 2001-2011, this sub-model was not included in
the model. Testing of explanatory variables resulted in Cramer's V (Annex-IV) for all and
individual classes showed that categorical variable of all possible transition to builtup obtained
from Evidence Likelihood in Variable transform utility (Eastman, 2016) have highest Cramer's V,
projecting it's importance to Model Building and Prediction.

Figure 11: Transition Potential maps.

In figure, 1=Forest, 2=Agriculture, 4=Bare area and 5=Builtup.

32
Static and dynamic explanatory variables (Table 5) were added to sub-model. Leaving all the
parameters default in Run-Transition Sub-model panel illustrated in figure. Following statistics
were obtained.

Iterations: 10,000 Learning rate: 0.001

Training RMS: 0.2235 Testing RMS: 0.2403
Accuracy Rate: 73.44% Skill Measure: 0.6813
Sample Size per class: 127 (50%Training/50%Testing)

These statistics were satisfactory for model building. Therefore, further sub-model was not run in
the panel. Same Model with change in base year data was used to predict the Future scenarios.

Change Demand Modelling

The default procedure, Markov Chain, determines the amount of change using the earlier and later
landcover maps along with the date specified. The procedure determines exactly how much land
would be expected to transition from the later date to the prediction date based on a projection of
the transition potentials into the future and creates a transition probabilities file (Eastman, 2016).
Markov chain was used to create transition probability matrix for the prediction year 2018 (Table
12).

Table 12: Transition probability matrix for 2018

LULC class Forest Agriculture Water Bare Area Builtup
Forest 0.919 0.0756 0.0025 0.002 0.0009
Agriculture 0.1181 0.8146 0 0.0352 0.0321
Water 0.089 0.0292 0.6826 0.1992 0
Bare Area 0.0752 0.4455 0.0282 0.4201 0.031
Builtup 0.0635 0.0635 0 0.0096 0.8634

33
Change Prediction
Change Allocation Panel Predicts the future scenarios (Eastman, 2016) based on the predetermined
model and matrix called MLP Markov Model (Ahmed and Ahmed, 2012).Finally,
Simulated/Predicted map for 2018 was predicted using both hard and soft prediction.

Figure 12: Predicted LULC map 2018

The hard prediction procedure is based on a multi-objective land allocation algorithm (MOLA)
that looks through all transitions and creates a list of host classes (classes that will lose some
amount of land) and a list of claimant classes (classes that will acquire land) for each host. The
quantities are determined from a run of the Markov chain procedure. A multi-objective allocation
is then run to allocate land for all claimants of a host class. The results of the reallocation of each
host class are then overlaid to produce the result (Eastman et al., 1995; Eastman, 2016). In
contrast, the soft output is a continuous mapping of vulnerability to change. It is simply an
aggregation of the transition potentials of all selected transitions. Logical OR by default was
selected for the aggregation type for soft prediction. Hard predicted map (figure 13) and soft
predicted map (Annex V) for 2018 were obtained in a single step.

34
Validation
The Validation panel allows to determine the quality of the prediction land use map in relation to
a map of reality. It does this by running a 3-way cross tabulation between the later landcover map,
the prediction map, and a map of reality (Eastman, 2016). No. of Hits, No. of Misses. False alarm
and Null scores were analyzed and Model validated. The results showed higher percentage in
agreement (78.48%) compared to disagreement or Error (21.52%) (Table 13). So, the model was
adopted for the future prediction. In addition, Cross Tabulate MODULE in TerrSet gave same
overall accuracy for these two LULC maps (Annex IV). This accuracy is greater than compared in
the study by Ahmed and Ahmed, 2012 in Dhaka; Basyal, 2013 in Turkey; Sahalu, 2014, in Ethiopia
and Gala et al., 2015 in Illinois. MLP Markov was found more accurate than CA Markov and
Stochastic Markov models (Ahmed and Ahmed, 2012).

Table 13: Agreement/Disagreement between Predicted and actual LULC of 2018.

Name of Component Percentage

Persistence simulated correctly (Null success) 78.22588081
Change simulated Correctly (Hits) 0.256511279
Total Agreement 78.48239209
Change simulated as persistence (Misses) 20.20571199
Persistence simulated as Change (False alarms) 1.31189592
Total Disagreement 21.51760791
Total Percentage 100

35
4.2.2. Future Prediction
4.2.2.1. Prediction 2030
Hard Prediction was carried out taking reference layers as 2001 and 2018. Dynamic variables were
updated. All the steps of 4.2.1. were carried out except the base layers changed. Transition
probability matrix (Table 14) was obtained using MARKOV chain Module in LCM.

Table 14: Transition probability matrix for 2030 and 2050

LULC class Forest Agriculture Water Bare Area Builtup
Forest * 0.8688 0.1187 0.0037 0.0075 0.0013
** 0.7158 0.2299 0.0068 0.0277 0.0198
Agriculture * 0.1012 0.7542 0 0.0791 0.0654
** 0.2047 0.5721 0.0032 0.0917 0.1283
Water * 0.0493 0 0.681 0.2679 0.0018
** 0.119 0.1849 0.382 0.2846 0.0295
Bare Area * 0.0448 0.462 0.0276 0.4205 0.0452
** 0.1471 0.5357 0.0276 0.1796 0.1101
Builtup * 0.0545 0.0784 0 0.0098 0.8572
** 0.1306 0.1672 0.0007 0.0244 0.6771
* = 2030 and ** =2050

36
Figure 13: Predicted LULC Map of 2030

The predicted map of 2030 showed Forest (43.41%) as the major component in the LULC system,
followed by Agriculture (40.06%), Builtup (9.02%), Bare area (6.63%) and water (0.83%) (Figure
15).

Forest Agriculture
43.43% 40.08%

Builtup
9.02%
Bare Area Water
6.63% 0.83%

Figure 14: Pie-chart showing the Percentage coverage of each LULC Classes in Predicted map
2030.

37
4.2.2.2. Prediction 2050
Similar steps to 4.2.2.1 were carried out for prediction of Land use land cover change 2050. Only
Hard Prediction was done to predict only the LULC Classes. Major change was Transition
probability matrix (Table 14) for 2050 obtained from same MARKOV Module attached in LCM.

Figure 15: Predicted LULC map of 2050

The predicted map of 2030 showed Forest (42.61%) as the major component in the LULC system,
followed by Agriculture (37.37%), Builtup (12.97%), Bare area (6.18%) and water (0.83%)
(Figure 16).
Forest
42.63%

Agriculture
37.38%
Builtup
12.98%
Bare Area Water
6.18% 0.83%
Figure 16: Pie chart showing the coverage of each LULC Classes in Predicted map 2050.

38
Builtup Scenario

Percentage
16

14 12.97

12

10 9.02
Percentage (%)

8
5.85
6
3.61
4
1.57
2

0
1990 2000 2010 2020 2030 2040 2050 2060
Year

Figure 17: Trend of Expansion in Builtup Area (2001-2050)

Change rate
10 8.68
8 7.15
Change rate (%)

6
3.66
4
1.83
2

0
2001-11 2011-18 2018-2030 2030-2050
Period

Figure 18: Annual change rate (2001-2050)

Figure shows that the trend of increasing with time. However, the rate of change per year is
decreasing with increase in time. This might be due to saturation of the urban suitable areas as
inputted to the model based on past trends.

39
0verall change 2001 to 2050
In this period of 50 years, Maximum change is predicted to be occurring in Builtup (11.40%)
followed by Agriculture (-9.54%), Forest (-1.80%), Bare area (-0.18%) and water (0.08%).The
expansion in Builtup area will be provided by agriculture land. Maximum change is seen in those
private lands because there is no policy regarding use of private lands. People can use in their own
wish (Ishtiaque et al., 2017). Forest in Nepal are under very integrated system of conservation,
which shows the low decrease in its cover. In addition, another reason for slight change in Forest
may be due to classification of all green tree covered area along with Shrub lands, grasslands as
forest. There may be interclass change, which is not seen in this study. Study that is more intensive
can be helpful to determine the intensive changes.

Table 15: Overall change 2001 to 2050

Year Forest Agriculture Water Bare Area Builtup
2001 44.41 46.91 0.75 6.36 1.57
2011 47.94 43.42 0.77 4.26 3.61
2018 43.45 42.88 0.84 6.99 5.85
2030 43.41 40.06 0.83 6.63 9.02
2050 42.61 37.37 0.83 6.18 12.97

40
 CHAPTER FIVE
CONCLUSION AND RECOMMENDATIONS

5.1 Conclusion
Landsat's satellite images can be used to obtain land use land cover maps with acceptable
accuracy. Hybrid classification method further increases the accuracy of the classified images.
The obtained land use land cover maps summarized the highest coverage in the area by forest and
agriculture followed by other three classes in all three dates.

For the classified dates 2001, 2011 and 2018, it is observed that forest increased in the first period
while slightly decreased in later period coming to its original level. Forest has decreased slightly
in the period with an annual rate of 0.13%. Agriculture/cultivation land is decreasing with a rate
of -0.53% per year in the period 2001 to 2018. Agriculture is major factor contributing to the
Builtup expansion in the area. The rate of Builtup expansion in this period is maximum with an
annual increasing rate of 8.045%. Water in the area is somewhat constant in the period with an
increasing rate of 0.67% per year. This change in water coverage is due to the shifting of the Bheri
River in the boundary along with time. Bare area has slightly increased at this period 0.56% per
year. Bare area is low compared to other studies. The reason behind this is that all the playgrounds
around the settlements and other zones are classified as Builtup area in this study. Bare area in
this study mostly contains sandy riverbanks and areas where landslide has been occurred.

MLP Markov Model predicts the land use land cover map of 2018 with an accuracy of 78.48%,
which is more accurate, compared to Ahmed and Ahmed, 2012 in Dhaka. Same accuracy was
obtained when crosschecking was done using cross-tabulate module in Tersset. The model
calibration was also found satisfactory with an accuracy rate of 73.44%. Predicted land cover was
underestimated compared to Classified LULC map of 2018. The sub- model were focused only in
Builtup area transition. This may have caused some low accuracy to other classes' prediction.
Prediction map of 2030 and 2050 shows increasing pattern of Builtup area with 9.02% and 12.97%
respectively. The prediction shows Forest area will remain change with very small amount. Other
land use land cover classes will decrease contributing their area to Builtup area expansion in the
area.

41
Comparing 2001 and 2050 Land use land cover classes composition in the map, maximum rate of
change is observed in Builtup area with 11.40% per year followed by Agriculture land (-9.54%),
Forest (-1.80%), bare area (-0.18%) and water (0.08%).

5.2 Recommendations
 National if not Major cities land cover data should be updated frequently, and made
available in digital format to users.
 There is a serious shortage of digital data in all sectors. All the local, regional and national
governmental and non-governmental institutions should provide digital data to the users
and must update these data regularly.
 More studies regarding the land cover change assessment using GIS and Remote sending
should be carried out.
 Government of Nepal should provide high-resolution satellite and other data to researchers
studying on local level/intensive level.
 Studies that are more intensive should be carried out to document the existing drivers
behind Land use land cover changes in the microclimate level.
 Consistent multi-temporal Landsat satellite data for each year provides detail comparison
of images, change analysis and modeling. Especially for Markov chain model,
 Incorporating socio-economic data, land policy, biophysical and human factors (population
density, technology, political) could improve the performance of land use models for future
predictions. Hence, it is important to the planners, decision makers and stakeholders for
efficient utilization of land.

42
References
Acharya, K., and Dangi, R. (2009). Case studies on measuring and assessing forest degradation:
forest degradation in Nepal: review of data and methods. Rome, Italy. p 163.

Agarwal, C., Green, G.M., Grove, J.M., Evans, T.P., and Schweik, C.M. (2002). A Review and
Assessment of Land-Use Change Models: Dynamics of Space, Time, and Human Choice,
Technical Report, US Department of Agriculture, Forest Service, Northeastern Research Station,
Pennsylvania, USA

Ahmad, W., Jupp, L.B., and Nunez, M. (1992). Land cover mapping in a rugged terrain area using
Landsat MSS data. Int. J. Remote Sensing, 13, 673–683.

Ahmed, B. (2011). Urban land cover change detection analysis and modeling spatio-temporal
Growth dynamics using Remote Sensing and GIS Techniques: A case study of Dhaka,
Bangladesh. Master’s Thesis, Erasmus Mundus Program, Universidade Nova de Lisboa (UNL),
Instituto Superior de Estatística e Gestão de Informação (ISEGI), Lisbon, Portugal.

Ahmed, B., and Ahmed, R. (2012). Modeling urban land cover growth dynamics using multi-
temporal satellite images: A case study of Dhaka, Bangladesh. ISPRS Int. J. Geo-Infographics, 1,
3–31.

Anderson, J.R., Hardy, E.E., Roach, J.T., Witmer, R.E. (1976). A Land Use and Land Cover
Classification System for Use with Remote Sensor Data. Geological Survey Professional Paper
964. United States Government Printing Office Washington.

Atkinson, P.M. and Tatnall, A.R.L. (1997). Introduction Neural networks in remote sensing. Int. J.
Remote Sens., 18, 699–709.

Awasthi, K. D. Sitaula, B. K. Singh, B. R. and Bajrachrya, R. M. (2002) . Land-use change in two

Nepalese watersheds: GIS and geomorphometric analysis. Land Degradation and Development,
13, 495–513.

Araya, Y. H and Cabral, P. (2010). Analysis and Modeling of Urban Land Cover Change in Setúbal
and Sesimbra, Portugal. Remote Sensing, 2(6), 1549–1563.doi:10.3390/rs2061549.

43
Bajracharya, S., Maharjan, S., and Shrestha, F. (2014). Glacier Status in Nepal and Decadal
Change from 1980 to 2010 Based on Landsat Data. ICIMOD, Kathmandu, Nepal. p 88.

Bajracharya, S.R., Shrestha, F., and Maharjan, S.B. (2011). Glaciers shrinking in Nepal Himalaya.
INTECH Open Access Publisher 458.

Balzter, H. (2000). Markov chain models for vegetation dynamics. Ecology Model, 126, 139–154.

Baker, W. (1989). A Review of models of landscape change. Landscape Ecology, 2 (2), 111-133.
Bayes, A. (2012). Modeling urban land cover growth dynamics using multi temporal satellite
images: A case study of Dhaka, Bangladesh, ISPRS International Journal of Geo-Information, 1
(1), 3-31.

Baysal, G. (2013). Urban land use and land cover change analysis and modeling: A case study area
Malatya, Turkey. Dissertation submitted in partial fulfillment of the requirements for the Degree
of Master of Science in Geospatial Technologies.

Becker, A., and Bugmann, H. (2001). Global change and mountain regions - an IGBP initiative for
collaborative research. In: Visconti G et al. (eds.), Global Change and Protected Areas. Advance
in Global Change Research. Laquila, Italy. pp 3-9.

Bhatta, B. (2010). Analysis of Urban Growth and Sprawl from Remote Sensing data; Advances in
Geographic Information Science, Series Edition, Shivan and Balram, Canada. Springer-Verlag,
80–82.

Bisht, B.S., and Kothyari, B.P. (2001). Land Cover Change Analysis of Garur Ganga Watershed
Using GIS/Remote Sensing Technique. Indian Society of Remote Sensing, 29(3), 165-174.

Breuer, J.A., Huisman, A., and Frede, H.G. (2006). Monte Carlo assessment of uncertainty in the
simulated hydrological response to land use change. Journal of Environmental Model
Assessments, 11, 209-218.

44
Brown, D.G., Pijanowski, B. C., and Duh J. D. (2000). Modelling the relationships between land use
and land cover on private lands in the Upper Midwest, USA. Journal of Environmental
Management, 59 (4), 247–263.

Campbell, J.B., and Wynne, R.H. (2011). Introduction to Remote Sensing; Guilford Press: New
York, NY, USA, Volume 5.

Carmelo, R.F., Giuseppe, M., and Maurizio, P. (2012). Land Cover Classification and Change
Detection Analysis Using Multi-Temporal Remote Sensed Imagery and Landscape Metrics.
European Journal of Remote Sensing, 45, 1-18.

Castella, J.C., Boissau, S., Trung, T.N., and Quang, D.D. (2005). Agrarian transition and lowland-
upland interactions in mountain areas in northern Vietnam: Application of a multi-agent
simulation model. Agricultural systems, 86, 312–332.

Central Bureau of Statistics (CBS). (2014). Population Monograph of Nepal; National Planning
Commission Secretariat; CBS: Kathmandu, Nepal.

Chaurasia, R., Closhali, D.C., Dhaliwal, S.S., Minakshi, Sharma, P.K., Kudrat, M., and Tiwari, A.K.
(1996). Landuse change analysis for agricultural management – a case study of Tehsil Talwandi
Sabo, Punjab. Indian Society of Remote Sensing, 24 (2), 115–123.
Chaurjahari Municipality. (2018). Brief introduction of Chaurjahari Municipality. Retrieved from
http://Chaurjaharimun.gov.np

Chen, J., Liao, A. et al. (2015). Global land cover mapping at 30m resolution: A POK-based
operational approach. ISPRS Journal of Photogrammetry and Remote Sensing, 103, 7-27.DOI:
10.1016/j.isprsjprs.2014.09.002

Chhatkuli, R.R. (2004). National Geographic Information Infrastructure in Nepal for Strengthening
Planning and Resource Management. Nepalese Journal on Geoinformatics, 3, 1-9.

Chilar, J. (2000). Land cover mapping of large areas from satellites: status and research priorities.
International Journal of Remote Sensing, 21(67), 1093–1114.

45
Civco, D.L. (1993). Artificial neural networks for landcover classification and mapping. Int. J.
Geogr. Inf. Sci., 7, 173–186.

Clarke, K. C., Gaydos, L., and Hoppen, S. (1997). A self-modifying cellular automaton model of
historical urbanization in the San Francisco Bay area, Environment and Planning B: Planning and
Design, 24, 247–261.

Congalton, R.G. (1991). A review of assessing the accuracy of classifications of remotely sensed
data. Remote Sens. Environ, 37, 35–46.

Coppin P. L. and Bauer M. E. (1996). Change detection in forest ecosystems with remote sensing
digital imagery, Remote Sensing Reviews, 130, 207–234.

Cramer, H. (1999). Methods of estimation. In Mathematical Methods of Statistics, 19th ed.;

Chapter 33; Princeton University Press: Princeton, NJ, USA, pp. 497-506.

Dadhich, P.N., and Hanaoka, S. (2010). Remote sensing, GIS and Markov’s method for land use
change detection and prediction of Jaipur district. Geomatics, 4, 9–15.

Daryei, J. (2003). Digital Change Detection Using Multi-Scale Wavelet Transformation and Neural
Network, M.Sc. Thesis, International Institute for Geoinformation Science and Earth Observation,
Enschede, The Netherlands.

Dhinwa, P.S., Pathak, S.K., Sastry, S.V.C., Rao, M. Majundar, K.L., Chotani, M.L. Singh, J.P. and
Sinha, R.L.P. (1992). Land use change analysis of Bharatpur District using GIS. Indian Society of
Remote Sensing, 20 (4), 237-250.

District Development Committee (DDC). (2010). District Profile, Surkhet; Government of Nepal,
Ed.; DDC: Surkhet, Nepal.

Dongjie, G., Weijun, G., Kazuyuki, W., and Hidetoshi, F. (2008). Land use change of Kitakyushu
based on landscape ecology and Markov model. Geographical Science, 18, 455–468.

46
Deadman, P., Brown, R. D., and Gimblett, H. R. (1993). Modelling rural residential settlement
patterns with cellular automata, Journal of Environmental Management, 37, 147–160.

De Cola, L. (1989). Fractal Analysis of a Classified Landsat Scene, Photogrammetric Engineering

and Remote Sensing, 55 (5), 601-610.

De Jong, S. M., and Burrough, P.A. (1995). A Fractal Approach to the Classification of
Mediterranean Vegetation Types in Remotely Sensed Images, Photogrammetric Engineering and
Remote Sensing, 61 (8), 1041-1053.

Eastman, J.R. (2006). IDRISI Andes Guide to GIS and Image Processing, Worcester, Clark Labs.

Eastman, J.R. (2009). IDRISI Taiga Guide to GIS and Image Processing; Manual Version 16.02;
Clark Labs: Worcester, MA, USA,

Eastman, J.R. (2016). TerrSet Manual. Clark lab. Worcester, Clarks Labs.

Gala, T.S., Friehat, T., and Mulugeta, G. (2015). Modeling Urban Sprawls in Northeastern Illinois.
Journal of Geosciences and Geomatics, Vol. 3, No. 5, 133-141. DOI: 10.12691/jgg-3-5-4

Gao, J. (2012). Land cover change and its relationship with climate change in Koshi River Basin
of central Himalaya. Institute of Geographic Sciences and Natural Resources Research, Graduate
University of Chinese Academy of Sciences. p 160.

Gautam, A., Shivakoti, G., and Webb, E. (2004). A review of forest policies, institutions, and
changes in the resource condition in Nepal. International forestry review., 6(2), 136-148.DOI:
10.1505/ifor.6.2.136.38397

Gibson, P., and Power, C. (2000). Introductory Remote Sensing: Digital Image Processing and
Applications. Routledge, London.

Gwet, K. (2002). Kappa statistic is not satisfactory for assessing the extent of agreement between
raters. Stat. Methods Inter-Rater Reliab. Assess, 76, 378–382.

47
Halil, I. C., Siamak, K., and Stacy, A. C. N. (2006). Correspondence analysis for detecting land
cover change, Remote Sensing of Environment, 102 (3-4), 306-317.

Hegazy, I.R., and Kaloop, M.R. (2015). Monitoring urban growth and land use change detection
with GIS and remote sensing techniques in Daqahlia governorate Egypt. International Journal of
Sustainable Built Environment, 4, 117-124
Hilferink, M., and Rietveld, P. (1999). An integrated GIS based model for long-term projections of
land use in urban and rural areas. J. Geogr. Syst., 1, 155–177.

Huang, W., Liu, H., Luan, Q., Bai M, and Mu, X. (2008). Monitoring urban expansion in Beijing,
China by multi temporal TM and spot images. IEEE Proc, IGARSS. 4, 695–698.

Ishtiaque, A., Shrestha, M., and Chhetri, N. (2017). Rapid Urban Growth in the Kathmandu Valley,
Nepal: Monitoring Land Use Land Cover Dynamics of a Himalayan City with Landsat Imageries.
Environments, 4, 72-88. DOI: 10.3390/environments4040072.

Iqbal, M.F. and khan, I.A. (2014). Spatiotemporal Land Use Land Cover change analysis and
erosion risk mapping of Azad Jammu and Kashmir, Pakistan, The Egyptian Journal of Remote
Sensing and Space Sciences 17, 209–229.

Jensen, J.R. (2007). Remote Sensing of the Environment: An Earth Resource Perspective. 2nd
Edition, Pearson Prentice Hall, Upper Saddle River.

Joshi, B., and Tiwari, P.C. (2014). Land-use changes and their impact on water resources in
Himalaya. In Environmental Deterioration and Human Health. Springer: Dordrecht, The
Netherlands; 389–399.

Kachhwala, T.S. (1985). Temporal monitoring of forestland for change detection and forest cover
mapping through satellite remote sensing. In: Proceedings of the 6th Asian Conf. On Remote
Sensing. Hyderabad, 77–83.

Karul, C., and Soyupak, S. (2006). A comparison between neural network based and multiple
regression models for chlorophyll-a estimation. In Ecological Informatics; Recknagel, F., Ed.;
Springer: Berlin, Germany, pp. 309-323.

48
Khanal, N. (2002). Land Use and Land Cover Dynamics in the Himalaya: A Case Study of the
Madi Watershed, Western Development Region, Nepal. Tribhuvan University, Kirtipur, Nepal. p
297.

Kok, K. and Veldkamp, T.A. (2001). Evaluating impact of spatial scales on land use pattern analysis
in Central America. Agric. Ecosyst. Environment, 85, 205–221.

Lambin, E. F. (1997). Modelling and monitoring land-cover change processes in tropical regions,
Progress in Physical Geography, 21 (3), 375–393.

Lambin, E.F., Geist, H.J., Lepers, E. (2003). Dynamics of land use/cover change in tropical regions.
Annu. Rev. Environ. Resources, 28,205–241.

Latham, J. (2014). Initiative Brings Global Land covers Data under One Roof for the First Time,
FAO’s Land and Water Division. FAO, United Nations.
Li, X., and Yeh, A. G. (2002). Neural network based cellular automata for simulating multiple land
use changes using GIS, International Journal of Geographical Information Science, 16 (4), 323-
343.

Lillesand, T.M., and Kiefer, R.W. (2007). Remote sensing and image interpretation, fifth ed. Wiley,
New York.

LRMP, (1986). Land Utilization Report. Land Resource Mapping Project, Kenting Earth Science
Canada and Department of Topography, Government of Nepal, Kathmandu, Nepal. p 112.

Lu, D., and Weng, Q. (2007). A survey of image classification methods and techniques for
improving classification performance. Int. J. Remote Sens., 28, 823–870.

Magidi, J.T. (2010). Spatio-Temporal Dynamics in Land Use and Habitat Fragmentation in the
Sandveld, South Africa. MSc Thesis, Department of Biodiversity and Conservation Biology,
University of the Western Cape, Western Cape.

49
Maier, H. R., and Dandy, G. C. (2000). Neural network for the prediction and forecasting of water
resources variables: a review of modelling issues and applications, Environmental Modelling and
Software, 15 (1), 101-124.

Mas, J. F. (1999). Monitoring land cover changes: a comparison of change detection techniques,
International Journal of Remote Sensing, 20 (1), 139-152.

Mas J. F., Puig, H., Palacio, J. L. and Sosa-Lopez, A. (2004). Modelling deforestation using GIS and
artificial neural network, Environmental Modelling and software, 19 (5), 461-471.

Masek, J.G., Vermote, E.F., Saleous N.E., Wolfe, R., Hall, F.G., Huemmrich, K.F., Gao, F., Kutler,
J., and Lim, T-K. (2006). A Landsat surface reflectance dataset for North America, 1990–2000.
IEEE Geoscience and Remote Sensing Letters, 3(1), 68-72.
http://dx.doi.org/10.1109/LGRS.2005.857030.

McConnell, W., Sweeney, S.P., and Mulley, B. (2004). Physical and social access to land: Spatio-
temporal patterns of agricultural expansion in Madagascar. Agric. Ecosyst. Environment, 101,
171–184.

McDermid, G.J., Franklin, S.E., and LeDrew, E.F. (2005). Remote Sensing for Large-Area Habitat
Mapping. Progress in Physical Geography, 29, 449-474.
http://dx.doi.org/10.1191/0309133305pp455ra

Miller, J., and Rogan, J. (2007). Using GIS and Remote Sensing for Ecological Mapping and
Monitoring. In: Mesev, V., Ed., Integration of GIS and Remote Sensing, John Wiley and Sons,
Ltd., 233-268.
Moghadam, H.S., Helbich, M. (2013). Spatiotemporal urbanization processes in the megacity of
Mumbai India: A Markov chains-cellular automata urban growth model. Appl. Geogr., 40, 140–
149.

MoFSC, (2009). Asia-Pacific Forestry Sector Outlook Study II: Nepal Forestry Outlook Study.
Ministry of Forests and Soil Conservation. Singha Durbar, Kathmandu, Nepal. p 83.

50
Mouat, D.A., Glenda, G.M., and Lancaster, J. (1993). Remote Sensing Techniques in the Analysis
of Change Detection. Geocarto International, 8 (2), 39-50.

MPFS, (1989a). Master Plan for Forestry Sector, Main Report. Ministry of Forestry and Soil
Conservation, Kathmandu, Nepal.

MPFS, (1989b). Master Plan for Forestry Sector, Forestry Resource Information and Planning
Report. Ministry of Forestry and Soil Conservation, Kathmandu, Nepal.

Nield, R. (1985). Fuelwood and fodder-problems and policy. Water and Energy Commission
Secretariat. Kathmandu, Nepal. p 44.

Owojori, A., and Xie, H. (2005). Landsat image-based LULC changes of San Antonio, Texas using
advanced atmospheric correction and object-oriented image analysis approaches. Paper Presented
at the 5th International Symposium on Remote Sensing of Urban Areas, Tempe, AZ.

Paudel, B., Zhang, Y.L., and Li, S.C. (2016). Review of studies of land use land cover change in
Nepal. Journal of Mountain Sciences, 13(4).DOI:10.1007/s11629-015-3604-9

Phiri, D., and Morgenroth, J. (2017). Developments in Landsat Land Cover Classification
Methods: A Review. Remote Sensing, 9, 967; doi: 10.3390/rs9090967.

Phong, L. T. (2004). Analysis of Forest Cover Dynamics and their Driving Forces in Bach Ma
National Park and Buffer Zone using Remote Sensing and GIS. MSc Thesis submitted to the
International Institute for Geo-information Sciences and Earth Observation, Enschede, The
Netherlands.

Pijanowski, B.C., Gage, S.H., and Long, D.T. (2000). A Land Transformation Model: Integrating
Policy, Socioeconomics and Environmental Drivers using a Geographic Information System. In
Landscape Ecology: A Top down Approach; Harris, L.; Sanderson, J., Eds.; CRC Press: Boca
Raton, FL,USA, pp. 183–198.

51
Pijanowski, B. C., Gage, S. H., Long, D. T. and Cooper, W. E. (2000). A land transformation model
for the Saginaw bay watershed, In, Landscape Ecology: A Top down Approach (Sanderson and
Harris eds.), 183–198, Boca Raton, FL: Lewis Publishers.

Pontius, R.G., and Spencer, J. (2005). Uncertainty in extrapolations of predictive land change
models. Environ. Plan. B, 32, 211–230.

Pontius, R.G., Jr., Boersma, W., Castella, J.-C., Clarke, K., de Nijs, T., Dietzel, C., Duan, Z., Fotsing,
E., Goldstein, N., and Kok, K. (2008). Comparing the input, output, and validation maps for several
models of land change. Ann. Reg. Sci., 42, 11–47.

Poudel, K.R. (2008). “Urban Growth and Land Use Change in the Himalaya Region” A Case Study
of Pokhara Sub-Metropolitan City, Nepal, GIS Ostrava, 1-11.

Ramankutty, N., and Foley J.A. (1999). Estimating historical changes in global land cover:
Croplands from 1700 to 1992.Glob. Biogeochem. Cycles, 13, 997-1027.

Regmi, R.R., Saha, S.K., and Subedi, D.S. (2017). Geospatial Analysis of Land Use Land Cover
Change Modeling in Phewa Lake Watershed of Nepal by Using GEOMOD Model. The
Himalayan Physics, 65-72.

Rijal, S., Rimal, B., and Sloan, S. (2018). Flood Hazard Mapping of a Rapidly Urbanizing City in
the Foothills (Birendranagar, Surkhet) of Nepal. Land, 7, 60-73. DOI: 10.3390/land7020060.

Rohan Chandralal Wickramasuriya, Arnold K. Bregt, Hedwig van Delden, et al. (2009). The
dynamics of shifting cultivation captured in an extended constrained cellular automata land use
model, Ecological Modelling, 220, 2302–2309.

Rokaya, M.B., Muenzbergova, Z., and Shrestha, M.R. (2012). Distribution patterns of medicinal
plants along an elevational gradient in central Himalaya, Nepal. Journal of Mountain Science. 9
(2) 201-213. DOI: 10.1007/s11629-012-2144-9.

Rosenfield, G.H. and Fitzpatirck-Lins, K. (1986). A coefficient of agreement as a measure of

thematic classification accuracy. Photogramm. Eng. Remote Sensing, 52 (2), 223–227.

52
Sahalu, A.G. (2014). Analysis of urban Land use land cover changes: A case study from Bahir Dar,
Ethiopia. Dissertation submitted in partial fulfillment of the requirements for the Degree of Master
of Sciences in Geospatial Technologies.

Sang, L., Zhang, C., Yang, J., Zhu, D, and Yun, W. (2011). “Simulation of land use spatial pattern
of towns and villages based on CA-Markov model”, Mathematical and Computer Modelling, 54,
938-943.

Sankhala, S., and Singh, B. (2014). Evaluation of urban sprawl and land use land cover change
using remote sensing and GIS techniques: a case study of Jaipur City, India. Int. J. Emerging
Technol. Adv. Eng, 4 (1), 66–72

Santé, I., Garcia, A.M., Miranda, D., and Crecente, R. (2010). Cellular automata models for the
simulation of real-world urban processes: A review and analysis. Landscape Urban Planning, 96,
108-122.

Schowengerdt, R.A. (2012).Techniques for Image Processing and Classifications in Remote

Sensing; Academic Press: Cambridge, MA, USA.

Schulz, J.J., Cayuela, C., Echeverria, C., Salas, J., and Rey Benayas, J.M. (2010). Monitoring Land
Cover Change of the Dryland Forest Landscape of Central Chile (1975-2008). Applied
Geography, 30,436-447. http://dx.doi.org/10.1016/j.apgeog.2009.12.003.

Sharma, K.P., Vorosmarty, C.J., and Moore Iii, B. (2000). Sensitivity of the Himalayan hydrology
to land-use and climatic changes. Climatic Change. 47(1-2), 117-139. DOI: 10.1023/a:
1005668724203.

Shrestha, A.B., and Joshi, S.P. (2009). Snow cover and glacier change study in Nepalese Himalaya
using Remote Sensing and Geographic Information System. Journal of Hydrology and
Meteorology, 6(1), 26-36.

Shrestha, D. P., and Zinck, J.A. (2001). Land Use Classifications in mountainous areas: Integration
of image processing, digital elevation data and field knowledge (application to Nepal).
International Journal of Applied Earth Observation and Geoformation, 3(1), 78-85.

53
Silva, E.A., and Clarke, K.C. (2002). Calibration of the SLEUTH urban growth model for Lisbon
and Porto, Portugal. Comput. Environ. Urban Syst., 26, 525–552.

Singh, A., Singh, S., Garga, P.K., and Khanduri, K. (2013). Land Use and Land Cover Change
Detection: A Comparative Approach using Post Classification Change Matrix Function Change
Detection Methodology of Allahabad City. International Journal of Current Engineering and
Technology, 3 (1), 142-148.

Singha, A., Khanduric, K., Singhb, S., and Garga, P.K., (2013). Land Use and Land Cover Change
Detection: A Comparative Approach Using Post Classification Change Matrix and Discriminate
Function Change Detection Methodology of Allahabad City. International Journal of Current
Engineering and Technology, 3:1

Singh, A. (1989). Digital change detection techniques using remotely sensed data, International
Journal of Remote Sensing, 10 (6), 989-1003.

Singh, A. K. (2003). Modelling Land Use/ Land Cover changes using Cellular Automata in a
geospatial environment. M.Sc. Thesis. Indian Institute of Remote Sensing (IIRS), Dehradun, India
and International Institute for Geoinformation Science and Earth Observation (ITC), Enschede,
The Netherlands.
Steiner, D. (1970). Automation in photo interpretation. Geoforum, 1, 75–88.

Star, J.L., Estes, J.E., and McGwire, K.C. (1997). Integration of geographic information systems
and remote sensing. New York, NY: Cambridge University Press.

Subedi, P., Subedi, K., and Thapa, B. (2013). “Application of a Hybrid Cellular Automaton –
Markov (CA-Markov) Model in Land-Use Change Prediction: A Case Study of Saddle Creek
Drainage Basin, Florida.” Applied Ecology and Environmental Sciences, 1 (6), 126-132. Doi:
10.12691/aees-1-6-5.

Sudhira H. S. (2004). Integration of agent based and cellular automata based models for simulating
urban growth, M.Sc. Thesis, Indian Institute of Remote Sensing (IIRS), Dehradun, India and

54
International Institute for Geoinformation Science and Earth Observation (ITC), Enschede, The
Netherlands.

Thapa, R.B., and Murayama, Y. (2009). Urban mapping, accuracy, & image classification: A
comparison of multiple approaches in Tsukuba City, Japan. Applied Geography, 29, 135–144.

Theobald, D. M., and Hobbs, N. T. (1998). Forecasting rural land-use change: a comparison of
regression and spatial transition-based models, Geographical and Environmental Modelling. 2,
65–82.

Thompson, M.M., and Mikhail, E.M. (1976). Automation in photogrammetry: Recent

developments and applications (1972–1976). Photogrammetria, 32, 111–145.

Turner, M.G. (1987). Spatial simulation landscape changes in Georgia: A Comparison of 3

transition models, Landscape Ecology, 1 (1), 29-36.

Uddin, K., and Gurung, D.R. (2010). Land cover change in Bangladesh- a knowledge based
classification approach. Area, 41–46.

Uddin, K., Shrestha, H.L., and Murthy MSR. (2015). Development of 2010 national land cover
database for the Nepal. Journal of Environmental Management, 148, 82-90.DOI:
10.1016/j.jenvman.2014.07.047

UNDESA (United Nations, Department of Economic and Social Affairs). (2015). World Urbanization
Prospects. The 2014 Revision; UNDESA: New York, NY, USA.

Venkataratnam, L. (1980). Use of remotely sensed data for soil mapping. J. Ind Soc. Photo-
Interpret. Remote Sensing, 8, 19–25.

Vescovi, F.D. Park, S.J. and Vlek, P.L. (2002). Detection of human-induced land cover changes in
a savannah landscape in Ghana: I. Change detection and quantification. In 2nd Workshop of the
EARSeL Special Interest Group on Remote Sensing for Developing Countries. Bonn, Germany.

55
Viera, A.J. and Garrett, J.M. (2005). Understanding inter-observer agreement: the kappa statistic.
Family Med. 37, 360–363.

Vink, A. P. (1975). Land use in Advancing Agriculture. New York Spring Verlage Heibdeberg,
Berlin

Wang, X.L., and Bao, Y.H. (1999). Study on the methods of land use dynamic change research,
Progress in Geography. 18 (1), 81–87.

Wernick, M.N. & Morris, G.M. (1988). Maximum- Likelihood Image Classification. Proc. SPIE
0938, Digital and Optical Shape Representation and Pattern Recognition, 317.

Wood, E.C., Lewis, J.E., Tappan, G.G., and Lietzow, R.W. (1997). “The development of a land
cover change model for Southern Senegal”. [Online] Available online at:
http://www.ncgia.ucsb.edu/conf/landuse97/papers/wood_eric/pecdoc.html [Accessed Mar 3,
2012].

Wu, Q., Li, H., Wang, R., Paulussen, J., Hec, Y., Wang, M., Wang, B. and Wang, Z. (2006).
Monitoring and Predicting Land Use Change in Beijing Using Remote Sensing and GIS.
LandscapeandUrban Planning, 78,322-333.http://dx.doi.org/10.1016/j.landurbplan.2005.10.2.

Yoshitsugu Hayashi, H. Ji., and Xin, C. (2009). Application of an integrated system dynamics and
cellular automata model for urban growth assessment: A case study of Shanghai, China,
Landscape and Urban Planning, 91,133–141.

Young, N.E., Anderson, R.S., Chignell, M., Vorster, G., Lawrence, R., and Evangelista, P.H. (2017).
A survival guide to Landsat preprocessing. Concepts and Synthesis: Emphasizing new ideas to
stimulate research in Ecology, 98(4), 920–932

Yuan, D., Elvidge, C.D., and Lunetta, R.S. (1998). Survey of Multispectral Methods for Land Cover
Change Analysis. In: Lunetta, R.S., Elvidge, C.D. (Eds.), Remote Sensing Change Detection.
Environmental Monitoring Methods and Applications. Ann Arbor Press, Chelsea, MI, 21–39.

56
Zhang R, Tang C, Ma S, Yuan H, Gao L, and Fan W. (2011). Using Markov chains to analyze
changes in wetland trends in arid Yinchuan plain, China. Math. Computer Modelling, 54, 924–
930.

Zubair, A.O. (2006). Change detection in land use and Land cover using remote sensing data and
GIS: a case study of Ilorin and its environs in Kwara State, Msc Thesis, University of Ibadan,
Nigeria.

57
 ANNEXES

Annex-I Description of Topographic Map used in the study

Sheet No. Sheet Title Copyright

2882 05B Jajarkot-Khalanga Department of Survey, HMG
2882 05D Swikot Department of Survey, HMG
2882 06A Simli Department of Survey, HMG
2882 06C Pharulachaur Department of Survey, HMG

For all four sheets, Following details were same

Scale: 1:25000 (Map compiled from 1:50 000 scale aerial photography
of 1996.Field verification done in 1999)
Contour Interval: 20m (Supplementary contours at 10m)
Reference: Mean sea level (India)
Horizontal Datum
Spheroid Everest 1830
Projection Modified UTM
Origin Longitude 81˚E,
Latitude 0˚N
False co-ordinates of origin 500000m Easting
0m Northing
Scale factor at central meridian 0.9999

58
 Annex-II Accuracy Assessment

a. Error matrix for classified image of 2001

Reference
Class
F A W BA B Row Total UA (%)
LULC 2001

F 55 12 0 0 2 69 79.71
A 15 50 0 1 66 75.76
W 0 0 3 0 0 3 100.00
BA 1 1 0 5 0 7 71.43
B 0 0 1 0 4 5 80.00
Column total 71 63 4 6 6 150
PA (%) 77.46 79.37 75.00 83.33 66.67 OA= 0.78
Kappa coefficient = 0.640697283

b. Error matrix for classified image of 2011

Reference
Class
F A W BA B Row Total UA (%)
LULC 2011

F 66 10 0 0 0 76 86.84
A 11 44 0 2 1 58 75.86
W 0 0 3 1 0 4 75.00
BA 1 1 0 6 0 8 75.00
B 0 1 0 0 3 4 75.00
Column Total 78 56 3 9 4 150
PA (%) 84.62 78.57 100.00 66.67 75.00 OA= 81.33
Kappa coefficient = 0.682395644

c. Error matrix for classified image of 2018

Reference
Class F A W BA B Row Total UA (%)
LULC 2018

F 58 5 0 1 0 64 90.625
A 10 52 1 0 3 66 78.7879
W 1 0 2 0 0 3 66.6667
BA 0 0 0 7 0 7 100
B 2 1 0 0 7 10 70
Grand Total 71 58 3 8 10 150
PA (%) 82 90 67 88 70 OA=.84
Kappa coefficient = 0.74217575
F=Forest, A=Agriculture, W=Water, BA=Bare Area, B=Builtup, UA=User's Accuracy,
PA=Producer's Accuracy and OA= Overall Accuracy

59
 Annex-III Land change matrix

a. Land use land cover change matrix (2001-2011) (Pixel based cross Tabulation)
LULC 2011
Class
F A W BA B Total KIA
LULC 2001

F 47237 8831 104 848 167 57188 0.8548

A 5296 42628 69 3660 162 51817 0.6927
W 165 13 511 217 0 909 0.5919
BA 201 2138 175 2533 22 5073 0.3197
B 127 2350 2 310 1527 4321 0.8095
Total 53027 55962 864 7572 1883 119308
Overall Kappa: 0.8353
KIA= Kappa index of agreement using 2001 image as Reference image.

b. Land use land cover change matrix (2011-2018) (Pixel based cross Tabulation)
LULC 2018
Category F A W BA B Total KIA
LULC 2011

F 46992 4639 47 115 83 51877 0.77

A 8977 40183 24 1776 198 51160 0.7115
W 171 62 592 167 0 995 0.6519
BA 577 4626 239 2774 73 8293 0.5304
B 470 2305 4 237 3962 6983 0.9154
Total 57188 51817 909 5073 4321 119308
Overall Kappa: 0.8353
KIA= Kappa index of agreement using 2011 image as Reference image.

F=Forest, A=Agriculture, W=Water, BA=Bare Area, B=Builtup

60
 Annex-IV

a. Cramer's V Table
Cramer's V Value
Variable
Overall V Builtup Agriculture Bare Area Water Forest
Water Resources 0.1183 0.1712 0.1371 0.1197 0.0661 0.0000
Builtup 2001 0.1838 0.283 0.1919 0.1629 0.1178 0.1029
Distance from

Public services 0.106 0.1393 0.1176 0.1073 0.0971 0.0827

Secondary Road 0.1698 0.2133 0.2051 0.1845 0.1277 0.0000
Tertiary roads and
paths 0.1113 0.2279 0.169 0.0585 0.0477 0.0000
Elevation 0.1988 0.2439 0.2272 0.1852 0.1494 0.0000
Slope 0.1024 0.148 0.1247 0.1096 0.0788 0.0000
Evidence Likelihood 0.4374 0.7372 0.6474 0.3876 0.3254 0.0000

b. Cross Tabulation table for Predicted and Classified LULC maps of 2018
LULC 2018
LULC Class Forest Agriculture Water Bare area Builtup Total Error C UA
Forest 46968 8955 171 572 470 57136 0.17796 82.2039
Agriculture 4527 39148 61 4405 2011 50152 0.21941 78.0587
Predicted 2018

Water 47 24 592 239 4 906 0.34658 65.3422

Bare area 115 1757 167 2648 225 4912 0.46091 53.9088
Builtup 219 1274 1 425 4268 6187 0.31017 68.9834
Total 51876 51158 992 8289 6978 119293
Error 0 0.09461 0.23476 0.40323 0.68054 0.38836 0.21518
PA 90.539 76.5237 59.6774 31.946 61.1637 78.4824

61
 Annex-V MAPS
Note: The standard details of all the maps in Report is: Map Resolution: 300 dpi, Width: 3509
Pixels and Height: 2481 Pixels (A4 size either Potrait or Landscape)
a. LULC Change map (2001-2011)

Per category change in Hectares (2001-2011)

Change from to Hectares Change from to Hectares
Forest to Forest 4235.37 Water to Water 45.8174
Agriculture to Forest 791.807 Bare Area to Water 19.4567
Water to Forest 9.32487 Forest to Bare Area 18.0221
Bare Area to Forest 76.0336 Agriculture to Bare Area 191.698
Builtup to Forest 14.9736 Water to Bare Area 15.6909
Forest to Agriculture 474.851 Bare Area to Bare Area 227.114
Agriculture to Agriculture 3822.12 Builtup to Bare Area 1.97257
Water to Agriculture 6.18669 Forest to Builtup 11.3871
Bare Area to Agriculture 328.164 Agriculture to Builtup 210.706
Builtup to Agriculture 14.5253 Water to Builtup 0.17932
Forest to Water 14.7943 Bare Area to Builtup 27.7953
Agriculture to Water 1.16561 Builtup to Builtup 136.914

62
b. LULC change map (2011-2018)

Per category change in Hectares (2011 to 2018)

change from to Hectares Change from to Hectares
Forest to Forest 4213.406727 Water to Water 53.08003
Agriculture to Forest 415.943007 Bare Area to Water 14.97359
Water to Forest 4.214124 Forest to Bare Area 51.735097
Bare Area to Forest 10.311155 Agriculture to Bare Area 414.777399
Builtup to Forest 7.441964 Water to Bare Area 21.429269
Forest to Agriculture 804.897689 Bare Area to Bare Area 248.72298
Agriculture to Agriculture 3602.896717 Builtup to Bare Area 6.545342
Water to Agriculture 2.151893 Forest to Builtup 42.14124
Bare Area to Agriculture 159.240091 Agriculture to Builtup 206.671402
Builtup to Agriculture 17.753118 Water to Builtup 0.358649
Forest to Water 15.332238 Bare Area to Builtup 21.249945
Agriculture to Water 5.559057 Builtup to Builtup 355.241689

c. Driver Variable map

(a)Distance from water resources, (b)Evidence Likelihood for classification, (c)Distance from
builtup 2001, (d)Distance from public services, (e)Distance from secondary roads, (f)Distance
from tertiary roads, (g)Elevation, (h)Slope

63
64
d. Predicted map 2018

e. Soft Predicted map 2018

65
66