Visualization of Urban Growth Pattern in Chennai Using Geoinformatics and Spatial Metrics

Abstract

Urban growth is the spatial pattern of land development to accommodate anthropogenic demand that influences other land uses (e.g.: open spaces, water bodies, etc.). Driven by population increase, urban growth alters the community’s social, political and economic institutions with changing land use and also affects the local ecology and environment. India’s urban population has increased by 91 million between 2001 and 2011, with migration, the inclusion of new/adjoining areas within urban limits, etc. Evidently, the percentage of urban population in India has increased tremendously: from 1901 (10.8 %) to 2011 (31.16 %). Chennai has an intensely developed urban core, which is surrounded by rural or peri-urban areas that lack basic amenities. Studying the growth pattern in the urban areas and its impact on the core and periphery are important for effective management of natural resources and provision of basic amenities to the population. Spatial metrics and the gradient approach were used to study the growth patterns and status of urban sprawl in Chennai city’s administrative boundary and areas within a 10 km buffer, for the past forty years. It is found that though Chennai experiences high sprawl at peri-urban regions, it also has the tendency to form a single patch, clumped and simple shaped growth at the core. During this transition, substantial agricultural and forest areas have vanished. Visualization of urban growth of Chennai for 2026 using cellular automata indicates about 36 % of the total area being converted to urban with rapid fragmented urban growth in the periphery and outskirts of the city. Such periodic land-use change analysis monitoring, visualization of growth pattern would help the urban planner to plan future developmental activities more sustainably and judiciously.

Full text

RESEARCH ARTICLE
Visualization of Urban Growth Pattern in Chennai Using
Geoinformatics and Spatial Metrics
Bharath H. Aithal
1
&T. V. Ramachandra
1,2,3
Received: 13 May 2014 /Accepted: 13 July 2015
#Indian Society of Remote Sensing 2016
Abstract Urban growth is the spatial pattern of land devel-
opment to accommodate anthropogenic demand that influ-
ences other land uses (e.g.: open spaces, water bodies, etc.).
Driven by population increase, urban growth alters the
community’s social, political and economic institutions with
changing land use and also affects the local ecology and en-
vironment. India’s urban population has increased by 91 mil-
lion between 2001 and 2011, with migration, the inclusion of
new/adjoining areas within urban limits, etc. Evidently, the
percentage of urban population in India has increased tre-
mendously: from 1901 (10.8 %) to 2011 (31.16 %). Chen-
nai has an intensely developed urban core, which is
surrounded by rural or peri-urban areas that lack basic
amenities. Studying the growth pattern in the urban areas
and its impact on the core and periphery are important for
effective management of natural resources and provision of
basic amenities to the population. Spatial metrics and the
gradient approach were used to study the growth patterns
and status of urban sprawl in Chennai city’s administrative
boundary and areas within a 10 km buffer, for the past forty
years. It is found that though Chennai experiences high
sprawl at peri-urban regions, it also has the tendency to
form a single patch, clumped and simple shaped growth
at the core. During this transition, substantial agricultural
and forest areas have vanished. Visualization of urban
growth of Chennai for 2026 using cellular automata indi-
cates about 36 % of the total area being converted to urban
with rapid fragmented urban growth in the periphery and
outskirts of the city. Such periodic land-use change analy-
sis monitoring, visualization of growth pattern would help
the urban planner to plan future developmental activities
more sustainably and judiciously.
Keywords Urban sprawl .Spatial patterns .Spatial metrics .
Cellular automata
Introduction
Urbanisation is the physical growth of urban areas or the ter-
ritorial progress of a region as a result of increase in popula-
tion due to migration or peri-urban concentration into cities.
The transition happens from rural to urban in terms of industry
structure, employment, living environment and social security
(Weber 2001; Bhatta 2009). Urbanisation may be planned
with basic infrastructure or organic it occurs as individual,
commercial establishment, and the government makes efforts
to improve the opportunities for jobs, education, housing, and
transportation. Only 14 % of the world’s population lived in
urban areas in 1900, which increased to 47 % by 2000
(Brockerhoff 2000; Ramachandra et al. 2015b). Specifically
if looked at Indian case in 2011, 31.16 % of India’s 1.2 billion
people lived in urban areas (http://censusindia.gov.in) and this
is projected to reach 60 % by 2030. Unplanned urban growth
has a considerable impact on the natural resources and has led
*T. V. Ramachandra
cestvr@ces.iisc.ernet.in
1
Energy & Wetlands Research Group, CES TE15, Centre for
Ecological Sciences, Indian Institute of Science, Bangalore 560012,
India
2
Centre for Sustainable Technolgies (astra), Indian Institute of
Science, Bangalore 560012, India
3
Centre for infrastructure, Sustainable Transportation and Urban
Planning [CiSTUP], Indian Institute of Science,
Bangalore, Karnataka 560 012, India
J Indian Soc Remote Sens
DOI 10.1007/s12524-015-0482-0

to urban sprawl: a pressing issue in several metropolitan areas.
(Ji et al. 2006; Alsharif and Pradhan 2013; Arsanjani et al.
2013; Ramachandra et al. 2015a).
Urban sprawl refers to the uneven development along the
highways, surrounding the city or in the peri-urban region
resulting in the destruction of agricultural land and ecological
sensitive habitats (Chang 2003;Ramachandra,etal.2012a).
Rapid economic development during the last two decades has
resulted in high urban sprawl across India (Ramachandra et al.
2013a). The fragmented urban patches at the fringes originate
from multiple nuclei, resulting in urban growth with serious
environmental and social issues. Timely and accurate detec-
tion of changes to the Earth’s surface is vital to understand
relationships and interactions between anthropogenic and nat-
ural phenomena and thereby promote better decision-making
(Sudhira et al. 2004; Lu and Weng 2007;Bhatta2010;
Rahman et al. 2011; Setturu et al. 2012; Ramachandra et al.
2012a,b). This leads to better urban form and a positive rela-
tionship between a city and its surrounding areas. As conven-
tional methods of detecting changes are expensive, time con-
suming and lacks precision (Opeyemi 2008), geo-informatics
with temporal–spatial data acquired remotely through space
borne sensors have been adopted during recent years to map
and monitor specific regions (Jensen 1986;Singh1989;
Ramachandra et al. 2014).
Remote sensing provides vital data for monitoring land-
cover changes and its impacts on the environment at local,
regional and global scales (Johnson 2001; Kumar et al.
2011a). This aids in monitoring the changes in the region apart
from understanding the role of prominent causal factors. Re-
mote sensing is perhaps the only method for obtaining the
required data from inaccessible regions on a cost and time-
effective basis (Dessì and Niang 2008; Sharma and Joshi
2013). Remote sensing technology with geographic informa-
tion system (GIS) is ideal to understand the changes in the
landscape and help the planners to visualize likely implica-
tions with the future developments (Pathan et al. 1991;Ep-
stein et al. 2002; Civco et al. 2002;Heroldetal.2003a;
Matsuoka et al. 2004; Yorke and Margai 2007;Yangetal.
2008; Anindita et al. 2010; Kumar et al. 2011b). Spectral
and spatial details in the remote sensing data aid in delin-
eating land use categories to understand the surface land
characteristics (Ramachandra and Kumar 2008). Land-
scape dynamics have been understood by implementing
empirical studies focusing on monitoring, planning and
landscape design which failed to emphasize pattern of
growth specifically (O’Neill et al. 1999;Nassaueretal.
1999; Leitao and Ahern 2002; Ramachandra et al. 2012a;
Wen t z e t al. 2014). The land use dynamics can be under-
stood using temporal data acquired remotely through
space borne sensors. Land use changes reflect the varied
intensity and measure the spatial extent of the urban
growth.
The spatial characteristics of land use features are measured
using spatial metric, which explains the physical characteris-
tics of the land use (such as urban) forms and its pattern
(Herzog and Lausch 2001;Heroldetal.2002;Heroldetal.
2003a,b; Chang 2003). Further modelling based on these
changes would help in understanding future changes. Models
specific to urban growth have been used along with remote
sensing data and have proved to be important tools to measure
land-use change in peri-urban regions (Clarke and Gaydos
1998;Heroldetal.2003a,b; Mundia and Murayama 2010).
Torrens (Torrens 2000) suggests the use of cellular automata
(CA) for urban growth modelling and in simulating land use
changes as population migration and evolution can all be
modeled as automation, while the pixel and its neighbors
can account for various changes such as demographic data
etc., neighborhoods as part of the city can be simulated by
the cells on the lattice based on predefined site-specific rules
that represent the local current transitions that are raster-based
for modeling urban expansion for discrete time steps (Guan
et al. 2008). Further it can be noted that standalone CA
models lack the ability to account for the actual amount
of change since it cannot account for specific transitions
of change in the region. Eastman (Eastman 2009)sug-
gested coupling of Markov chains (MC) and CA. This
coupling helps in quantifying future likely changes based
on current and past changes which essentially addresses
the shortcoming of CA such as spatial allocation and the
location of change (Arsanjani et al. 2013). These studies
have failed to link agents of changes that are main driving
forces (He et al. 2008;Heetal.2013). Further, some
studies have used agents or drivers of changes that can
be transition potential using multi-criteria evaluation
(MCE) techniques. However, this approach failed due to
shortcoming in calibration techniques (Eastman 2009).
Hence it is necessary to calibrate the model and associate
the agents of changes and driving forces in order to un-
derstand and develop accurate transition potential maps.
Fuzzy-AHP technique of weighing agents was then pro-
posed to obtain such accurate calibrations (Ramachandra
et al. 2013a,b).First,fuzzyclusteringisusedtogroupthe
spatial units into clusters based on certain attribute data.
Analytical Hierarchal process (AHP) is then used to as-
sign weights to these spatial units thus based on various
inputs. Then once the weights are assigned Cellular
Markov models with the help of transition probability
matrix inherits past states of land use types to predict
future state (Praveen et al. 2013). Land use transitions is
simulated and validated for the year 2012. Further, pre-
diction for the year 2026 considering City Development
Plan [CDP] and without CDP were carried out from the
validated data.
The objective of the current communication is to vi-
sualize the urban growth patterns in Chennai. Chennai’s
J Indian Soc Remote Sens

rapid urbanization has resulted in increased population
density, traffic congestion and poor environmental qual-
ity, within and surrounding the city. Thus, planners need
to understand and visualize future plans to address these
problems and ensure that basic infrastructure and ame-
nities are available in the city. The multi-temporal re-
mote sensing data have been used to study the urban
structure and its dynamics. The spatial characteristics of
the urban pattern is analysed through gradient approach
using spatial metrics.
Study Area
Chennai, previously known as BMadras^is the capital city
of the Indian state of Tamilnadu and the fourth metropoli-
tan area in India (Dowall and Monkkonen 2008). It lies at
12
0
9′to13
0
9′Nand80
0
12′to 80
0
19′E at the eastern
coast and southern peak of India. The average elevation of
the city is about 8 m above the mean sea level. The average
temperature ranges from 38
0
to 42
0
C in summer and 18
0
to
20 °C in winter. As per the 1971 census, Chennai’s popu-
lation was 0.3 million, which has increased to more than 8
million (as per census 2011). The population density has
increased steadily from 769 persons per sq.km to 1041
(1981), 1315 (1991), 1558 (2001) and 2109 (2011) persons
per sq.km. Chennai is one of the more industrialised and
economically developed cities in India. Major industries
include automobile, software, textiles, and post the
1900s, information technology. Urban growth patterns
have been assessed considering the administrative bound-
ary of Chennai with a10 km buffer (Fig. 1)toaccountfor
dispersedgrowthinperi-urbanareas.
Materials and Methods
Chennai’s urban growth patterns have been assessed using
temporal remote sensing data of Landsat satellite
downloaded from public domains at Global Land Cover Fa-
cility (GLCF) (http://www.glcf.umd.edu/index.shtml)and
United States Geological Survey (USGS) Earth Explorer
(http://edcsns17.cr.usgs.gov/NewEarthExplorer/). Table 1
describes the data used, including remote sensing data and
collateral data. The Survey of India (SOI) topographic maps
of 1:50,000 and 1:250,000 scales were used to generate base
layers such as the city boundary. Chennai’s administrative
boundary is digitized from the city administration map ob-
tained from the municipality. Ground control points to reg-
ister and geo-correct remote sensing data were also collected
using hand held pre-calibrated GPS (Global Positioning Sys-
tem) devices, Survey of India topographic maps and Google
earth (http://earth.google.com,http://bhuvan.nrsc.gov.in).
The method adopted to assess the urban growth patterns
includes preprocessing, generation of land cover and land
use and a gradient-wise zonal analysis of Chennai, is repre-
sented in Fig. 2.
Preprocessing
The remote sensing data obtained was geo-referenced, geo-
corrected, rectified and cropped pertaining to the study area.
Remote sensing data from different sensors (with different
spatial resolutions) was resampled to 30 m in order to maintain
uniformity in spatial resolution. The study area includes the
Chennai administrative area and 10 km buffer from the ad-
ministrative boundary.
Fig. 1 Study area - Chennai ad-
ministrative boundary and a
10 km buffer
J Indian Soc Remote Sens

Land Cover Analysis
Land Cover analysis was performed to understand the changes
in the vegetation cover during the study period in the study
region. Normalized Difference Vegetation Index (NDVI) was
found suitable and was used for measuring vegetation cover
(Ramachandra et al. 2012a). NDVI values range from −1to+
1. Very low values of NDVI (−0.1 and below) correspond to
Tabl e 1 Data used in the
Analysis Data Year Purpose
Landsat Series Thematic mapper
(28.5 m)
1991, 2000 Land use, Land cover [LULC] analysis, landscape
dynamics, urban growth patterns
IRS P6 –LISS III MSS data
(23.5 m)
2012 LULC analysis, landscape dynamics, urban growth
patterns
Survey of India (SOI) topographic maps
of 1:50,000 and 1:250,000 scales
To Generate boundary and Base layer maps.
Field visit data –captured using GPS For geo-correcting and generating validation dataset
Aster GDEM of 1 arc-second (30 m) grid 2010 Extraction of Drainage lines, Slope analysis.
City developmental plans, location of
various agents
2005, 2015 Extraction of various agents of growth using
Google earth and ancillary data
Fig. 2 the procedure adopted for
classifying the landscape,
computation of metrics and
modelling
J Indian Soc Remote Sens

soil or barren areas of rock, sand, or urban built up. Zero
indicates the water cover. Moderate values (0.1 to 0.3) repre-
sent low-density vegetation, while high values (0.6 to 0.8)
indicate thick canopy vegetation.
Land Use Analysis
Land use categories were classified using supervised tech-
nique with Gaussian Maximum Likelihood classifier
(GMLC). The spatial data pertaining to different time frame
were classified, using signatures from training sites for the
land use types listed in Table 2. The training polygons were
compiled from collateral data of corresponding time period.
Latest data were classified using signatures (training poly-
gons) digitized with the help of Google earth. False color
composite of remote sensing data (bands –green, red and
NIR), was generated to visualise the heterogeneous patches
in the landscape. 60 % of the training data was used for clas-
sifying remote sensing data while the balance has been used
for validation or accuracy assessment.
Data is classified on the basis of training data through
GMLC a superior method that uses probability and cost func-
tions in its classification decisions (Duda et al. 2000;
Ramachandra et al. 2012a). Mean and covariance matrices
are computed using the maximum likelihood estimator. Land
use was analysed using the temporal data retrieved from the
open source program GRASS - Geographic Resource Analy-
sis Support System (http://ces.iisc.ernet.in/foss). Signatures
were collected from field visit and Google earth. Classes of
the resulting image were reclassified and recoded to form four
land-use classes (Table 2).
Accuracy Assessment
Accuracy assessment has been done for the classified data to
evaluate the performance of classifiers (Ramachandra et al.
2012a). This is done through using kappa coefficients
(Congalton et al. 1983). Overall (producer and user) accura-
cies were computed through a confusion matrix. Assessing
overall accuracy and computing Kappa coefficient are widely
accepted methods to test the effectiveness of classifications
(Lillesand and Kiefer 2002; Congalton and Green 2009).
Zonal Analysis
The study area (city boundary with 10 km buffer region) is
divided into4 geographic zones based on direction–Northeast
(NE), Southwest (SW), Northwest (NW) and Southeast (SE)–
as a city or its growth is usually defined directionally. Zones
were sub-divided using centroid as a reference (Central Busi-
ness District). The growth of the urban areas in each zone was
studied and understood separately, by computing urban den-
sity for different periods.
Gradient Analysis
(Division of zones into concentric circles): To visualise the
process of urban growth at local levels and to understand the
agents responsible for changes, each zone was divided into
concentric circles that are 1 km apart and radiate from the
city-center. The analysis of urban growth patterns at local
levels help the city administrators and planners identify the
causal factors of urbanization in response to the economic,
social and political forces and visualizing the forms of urban
growth with sprawl.
Urbanisation Analysis
To understand the growth of the urban area in specific zone
and determine whether it is compact or divergent, the
Shannon’s entropy (Sudhira et al. 2004; Ramachandra et al.
2012a) was computed for each zone. Shannon’s entropy (Hn),
given in Eq. 1, explains the development process and its char-
acteristics over a period of time and indicates whether the
growth was concentrated or aggregated.
Hn ¼−Xn
i¼1Pi log PiðÞ ð1Þ
Where, Pi is the proportion of the built-up in the i
th
con-
centric circle. The lowest value of zero reflects the distribution
is maximally concentrated. Conversely, the maximum value
equivalent to log n indicates of sprawl with even distribution
among the concentric circles.
Computation of Spatial Metrics
Spatial metrics have been used to quantify spatial characteris-
tics of the landscape. Selected spatial metrics were used to
anlayse and understand the urban dynamics. FRAGSTATS
(McGarigal and Marks in 1995) was used to compute metrics
at three levels: patch level, class level and landscape level.
Tab le 3. below gives the list of the metrics along with their
description considered for the study.
Tabl e 2 Land use classification categories adopted
Land use
Class
Land uses included in the class
Urban This category includes residential area, industrial area,
all paved surfaces (road, etc.) and mixed pixels
having major share of built up area.
Water bodies Tanks, lakes, reservoirs.
Vegetation Forest, nurseries.
Others Rocks, cropland, quarry pits, open ground at building
sites, un-metaled roads.
J Indian Soc Remote Sens

Visualisation of Urban Growth in Chennai by 2026
Agents of urbanisation and constraints (listed in Table 4)with
temporal land uses were taken as base layers for modelling
and visualisation. Data values were normalized through
fuzzification wherein the new values ranged between 0 and
255, 255 indicating maximum probability of change in land
Tabl e 4 Agents and constraints considered for modelling
Agents Industries, proximities to roads, railway stations, metro
stations, educational institutes, religious places etc.
Constraints Drainage lines, slope, water bodies, costal regulated
areas, catchment areas etc.
Tabl e 3 Landscapemetrics used in the current analysis(McGarigal and Marks in 1995; Aguilera et al. 2011; Ramachandra et al. 2012a; Ramachandra
et al. 2015b)
Indicator Formula
PLAND
PLAND ¼Pi¼∑n
j¼1aij
=A
P
i
=proportion of the landscape occupied by patch type i.
a
ij
=area (m
2
) of patch ij, A=total landscape area (m
2
).
Number of patches (Built-up) - NP N= n
i
,Range:NP≥1
Patch Density (PD) PD ¼ni
A10;000ðÞ100ðÞ,Range:PD>0
Largest patch Index (Built-up) (LPI)
LPI ¼maxn
j¼1aij
ðÞ
A100
ðÞ
,Range:0<LPI≤100
Normalised landscape shape Index (NLSI) NLSI ¼ei−minei
max ei−minei
Range: 0 to 1
Landscape shape Index (LSI) LSI ¼ei
mnei,Range:LSI≥1, without limit.
Interspersion and Juxtaposition Index (IJI)
IJI ¼
−∑m
K¼1eik
∑m
k¼1eik

ln eik
∑m
k¼1eik

ln m−1ðÞ 100ðÞ
Range: 0<IJI ≤100
eik= total length (m) of edge in landscape between patch types I and k.
m= number of patch type present in landscape.
Clumpiness Index (Clumpy)
CLUMPY ¼
Gi–Pi
Pi

forGi<PiPi<5;else
Gi–Pi
1P
i
0
B
B
@1
C
C
A
Gi
¼gii
∑m
k¼1gik

−minei
"#
Range: Clumpiness ranges from −1to1
Aggregation Index (AI)
AI ¼gii
maxgii
hi
100ðÞ
g
ii
=number of like adjacencies (joins) between pixels of patch type (class) i based
on the single-count method. max-g
ii
=maximum number of like adjacencies
(joins) between pixels of patch type (class) i based on the single-count method.
Percentage of Like Adjacencies (PLADJ)
PLADJ ¼100ðÞ*gij
∑m
k¼1gik

g
ii
=number of like adjacencies (joins) between pixels of patch type
(class) i based on the double-count method.
g
ik
=number of adjacencies (joins) between pixels of patch types
(classes) i and k based on the double-count method.
Patch Cohesion index
Cohesion ¼1−∑n
j¼1Pij
∑n
j¼1PIJ ffiffiffiffi
aij
2
p

1−1
ffiffiffi
A
2
p
hi
−1*100
J Indian Soc Remote Sens

use in contrast with 0 for no changes. Fuzzy outputs thus
derived are then taken as inputs to AHP for different factors
into a matrix form to assign weights. Each factor is compared
with another in pair wise comparison followed by enumera-
tion of consistency ratio which should be preferably less than
0.1 for the consistency matrix to be acceptable. Once weights
are determined MCE was used to determine the site suitability
considering two scenarios i). Restrictions based on City De-
velopment Plan (CDP); ii). As usual scenario without CDP.
These suitability change maps were considered in the MC-CA
model. Considering earlier land uses, transition potentials
were computed using a Markovian process. Using and hexag-
onal CA Filter of 5 × 5 neighborhood with variable iteration at
every step until a threshold is reached. Careful model valida-
tion through kappa statistics was conducted to assure accuracy
in prediction and simulation. Built-up areas were predicted for
2012 were cross-compared with the actual built-up areas in
2012 using classified data. The kappa index of 0.9 shows a
good agreement accuracy of the model. Future patterns of
urban expansion were then simulated for the years 2026.
Results and Discussion
Land cover (LC) of the region is assessed using the
preprocessed temporal remote sensing data using
Normalized Difference Vegetation Index (NDVI).
Figure 3depicts the vegetation cover based on NDVI
for Chennai region during 1991–2012. Table 5tabulates
the vegetation cover dynamics, which show that the
percentage of vegetative cover has drastically reduced
by 22 % during the past two decades, with the increase
in non-vegetative area (buildings, open space, water,
etc.). As per the LC analysis, the current vegetation
cover is about 48.18 %. To understand the status of
various land use (LU) classes in the region classification
of temporal remote sensing data were done using
GMLC. LU analysis helps to demarcate the different
classes and LC provides only the details of vegetative
cover (which might be vegetation and also cropland).
figure 4depicts land use changes, based on the analysis of
temporal remote sensing data using Gaussian Maximum Like-
lihood Classifier (GMLC). Table 6provides land use statistics,
which reveals that the area under vegetation declined from 70
to 48 % with an increase in urban (paved) surfaces from
1.46 % to about 18.5 %. Table 7tabulates overall accuracy
Fig. 4 Output of land use analysis in the study region
Tabl e 5 Vegetation cover
changes in Chennai Year Vegetation
(%)
Non-
vegetation
(%)
1991 70.47 29.53
2000 56.7 43.27
2012 48.18 51.82
Fig. 3 Temporal land cover analysis using NDVI
Tabl e 6 Land use in the study region
Land Use category Built-up Vegetation Water body Others
Area (%) Area (%) Area (%) Area (%)
1991 1.46 1.38 27.64 69.52
2000 2.52 0.8 27.25 68.35
2012 18.55 1.51 28.15 51.38
J Indian Soc Remote Sens

and kappa statistics obtained for classified images–these re-
sults show that the classified and ground truth data are closely
related. Overall accuracy and kappa statistics showed a good
relation of the classified data with ground truth data.
figure 5illustrates the urban growth pattern during 1991 to
2012, derived from the classified data.
Shannon’s Entropy (Hn)
Shannon’s entropy aids in assessing the urbanisation
pattern and was calculated for each zone considering
the gradients and as shown in Fig. 6. The values closer
to 0 indicates of a compact growth, as in 1991. All
zones show a compact growth or more concentrated
growth towards the central business district as even seen
in classified data. The Shannon’s entropy values have
increased temporally to 0.4 indicating the tendency of
dispersed fragmented growth or sprawl in all directions.
This indicates a fragmented outgrowth and mostly in
North west direction. Increased patches in all zones in-
dicate that the city is heading towards higher
fragmented outgrowth and more clumped inner core.
Further this pattern is analysed using landscape metrics
for better understanding at each gradient and zone.
Analysis of Spatial Metrics
Results of all metrics are given in Fig. 7a to i; the X-axis
represents the gradients considered and Y-axis, the Metric
value.
Percentage of Landscape (PLAND)
PLAND equals the proportion of landscape comprised of the
corresponding class patches. Built up proportion was comput-
ed to understand the ratio of built up in the landscape. The
results given in Fig. 7a, reveal that the proportion has in-
creased during the three decades with the tremendous growth
in the central circles. The outer gradients show a slight but
significant increase, which indicate sprawling development in
the periphery and buffer. The NW shows a relatively higher
increase as compared to that in the NE and SW.
Fig. 5 Urban growth in the study region
Tabl e 7 Accuracy assessment from confusion matrix and kappa statistics
Category 1991 2000 2012
Producer’s Accuracy User’s Accuracy Producer’s Accuracy User’s Accuracy Producer’s Accuracy User’s Accuracy
Built-up 92.04 93.02 90.42 97.38 78.23 93.02
Vegetation 99.31 92.49 97.75 99.42 98.98 79.49
Water body 91.50 93.84 91.65 89.52 99.47 93.84
Others 96.28 92.45 92.70 98.82 39.50 76.45
Overall Accuracy 92 91 97
Kappa 0.92 0.9 0.93
Fig. 6 Shannon entropy
J Indian Soc Remote Sens

Fig. 7 a Percentage of landscape (PLAND). b: Number of Patches (NP).
c: Patch Density (PD). d: Largest Patch Index (LPI). e: Landscape shape
index (LSI) - directionwsie. f: Normalized Landscape shape index (LSI).
g. Clumpiness index (CLUMPY). h. Aggregation Index. i.Patchcohe-
sion index. j. Percentage of like adjacencies metric
J Indian Soc Remote Sens

Fig. 7 (continued)
J Indian Soc Remote Sens

Fig. 7 (continued)
J Indian Soc Remote Sens

Fig. 7 (continued)
J Indian Soc Remote Sens

Fig. 7 (continued)
J Indian Soc Remote Sens

Number of Patches (NP)
NP equals the number of built up patches in a landscape. It
indicates the level of fragmentation in built up landscape.
Figure 7b depicts the significant increase in number of urban
patches during two decades in all directions in all gradients,
indicating the fragmentation of landscape during post 2000.
Patches at the city centre are combining to for single clumped
dominant patch, whereas in outer gradients is more
fragmented.
Patch Density (PD)
Patch density was computed as an indicator of urban frag-
mentation. As the number of patches increase, patch den-
sity increases representing higher fragmentation. Figure 7c
shows the trends in patch density and comparable to the
number of urban patches i.e., fragmented urban patches
towards the fringes in all directions and a clumped
patched growth at the center.
Largest Patch Index (LPI)
Largest Patch Index (LPI) was computed to represent the per-
centage of landscape that contain largest patch. Figure 7d il-
lustrates that gradients at the center are with higher values due
to the larger size urban patches indicating compact growth.
The declining trends at the gradients away from the city center
at fringes or peri urban areas indicates of fragmented urban
patches.
Landscape Shape Index (LSI) and Normalized Landscape
Shape Index (NLSI)
Landscape shape index provides a simple measure of class
aggregation or disaggregation. Aggregation is measured via
class edge. Normalized landscape shape index is similar to
landscape shape index but this represents the normalized val-
ue. figure 7e and f depicts LSI and NLSI respectively indicate
that lower values in the gradients at the city center during post
2000 highlights of simpler shapes with compact urban patches
compared to the fringes which show complex shape with the
fragmented outgrowth.
Clumpiness (CLUMPY) and Aggregation Index (AI)
Clumpiness index represents the status of urban landscape and
is the measure of patch aggregation. It measures the
clumpiness of the overall urban patches. Clumpiness ranges
from −1 to 1 where Clumpy is −1, when the urban patch type
is maximally disaggregated, Clumpy is 0 when the patch is
distributed randomly and approaches 1 when urban patch type
is maximally aggregated. Aggregation Index gives the similar
indication as clumpiness i.e., it measures the aggregation of
the urban patches, Aggregation index ranges from 0 to 100.
An analysis of clumpiness index for the year 1991, explained
that the city, except a few patches in the outskirts and buffer,
had a simple clumped growth at the center business district
with the values close to 0. Analysis for the year 2012 indicated
a high fragmented growth or non-clumped growth at outskirts
and more clumped growth in the center. The aggregation in-
dex showed the similar trend as the clumpiness index.
Figure 7h, also revealed that the center part of the city is
becoming single homogeneous patch whereas the outskirts
Tabl e 8 Predicted land use statistics for the year 2026
Categories / Year Predicted 2026
with CDP
Predicted 2026
without CDP
Builtup 36.6 36.5
Vegetation 2.4 2.4
Water 27.9 27.8
Others 33.1 33.3
Predicted 2026 with CDP Predicted 2026 without CDP
Fig. 8 Predicted land use
categories for the year 2026
J Indian Soc Remote Sens

are growing as fragmented patches of urban area destroying
other land uses and forming a single large urban patch.
Patch Cohesion Index
figure 7i measures the physical connectedness of the urban
patches. Higher cohesion values in 2012 indicate of higher
urban patches while lower values (in 1991) illustrate of fewer
urban patches in the landscape.
Percentage of Like Adjacencies (PLADJ)
This indicates the percentage of cell adjacencies of the corre-
sponding patch type. Figure 7j depicts that the urban patches
in the landscape are most adjacent to each other in 2012, and
were fragmented in 1991 and 2000. This illustrates that urban
land use is dominant in 2012 and is in the process of forming a
single patch.
Visualisation Using Fuzzy AHP CA-Markov
Prediction for the year 2026 was performed considering both
scenarios and agents as specified earlier. Wherein water bod-
ies and coastal regulation areas were kept constant insisting no
development at these areas. Subsequently prediction was done
without considering CDP. Table 8shows % land use category
change. Built-up areas have been increased two fold, decrease
in vegetation and increase in other categories can be observed
from 2013 to 2026. Significant changes can be seen in areas
which falls within the CMDA boundary such as Korathur and
Cholavaram lake bed, Redhills catchment area, Perungalathur
forest area, Sholinganallur wetland area etc. as visualized in
Fig. 8.
Conclusions
Urban growth is the spatial pattern of land development to
accommodate anthropogenic demand, influencing other land
uses (open spaces, etc.). Urban growth patterns of Chennai
and the sprawl in the surrounding 10 km buffer have been
assessed using temporal remote sensing data of four decades
and spatial metrics. The spatial characteristics of land use fea-
tures were measured using spatial metrics to explain the phys-
ical characteristics, forms and pattern of the land use. The area
under vegetation has declined from 70 % to 48 % with urban-
ization has increased from 1.46 % to about 18.5 %. The study
has demonstrated that urban growth processes, patterns and
structures can be quantified and monitored using a combina-
tion of remote sensing data, geo-informatics and spatial met-
rics through gradient approaches. The combination of remote
sensing, landscape metrics and gradient analysis provides in-
sights into the multidimensional phenomenon of urbanisation
and of dispersed growth–factors essential for urban planning.
The spatial metrics reveal a significant increase in number of
urban patches in all directions and gradients during two de-
cades, indicating landscape fragmentation post 2000. Patches
at the city center are combining to form a single clumped
dominant patch, whereas in outer gradients is more
fragmented. The analysis of urban growth patterns emphasize
the need for judicious land-use transformation, as well as the
formulation of urban development policy with an emphasis on
the sustainable utilization of natural resources.
Acknowledgments We are grateful to NRDMS Division, The
Ministry of Science and Technology, Government of India;
ISRO-IISc Space Technology Cell, Indian Institute of Science;
Centre for infrastructure, Sustainable Transportation and Urban
Planning (CiSTUP), Indian Institute of Science for the financial
and infrastructure support. Remote sensing data were downloaded
from public domain (http://glcf.umiacs.umd.edu/data). We are also
thankful to National Remote Sensing Centre, Hyderabad (http://
nrsc.gov.in) for providing the latest data of IRS 1D.
References
Aguilera, F., Valenzuela, L., & Leitao, A. (2011). Landscape metrics in
the analysis of urban land use patterns: a case study in a Spanish
metropolitan area. Landscape and Urban Planning, 99,226–238.
Alsharif, A. A. A., & Pradhan, B. (2013). Urban sprawl analysis of
Tripoli metropolitan city (Libya) using remote sensing data and
multivariate logistic regression model. Journal of the Indian
Society of Remote Sensing, 42(1), 149–163.
Anindita, D., Kumar, U., & Ramachandra, T. V. (2010). Urban Landscape
analysis through spatial metrics, In the proceedings of the
Conference on Infrastructure, Sustainable Transportation and
Urban Planning, CISTUP@CISTUP2010, JN Tata Auditorium,
Indian Institute of Science, Bangalore. 18–20 October, 2010.
Arsanjani, J., Helbich, M. W., Kainz, & Darvishi, A. (2013). Integration
of logistic regression, Markov chain and cellular automata models to
simulate urban expansion - the case of Tehran. International Journal
of Applied Earth Observation and Geoinformation, 21,265–275.
Bhatta, B. (2009). Analysis of urban growth pattern using remote sensing
and GIS: a case study of Kolkata, India. International Journal of
Remote Sensing, 30,4733–4746.
Bhatta, B. (2010). Analysis of urban growth and sprawl from remote
sensing data. Berlin: Springer.
Brockerhoff, M. P. (2000). An urbanizing world. Population Bulletin, 55,
3–44.
Chang, C. Y. (2003). Landscape structure and bird’s diversity in the rural
areas of Taiwan. Journal of Environmental Sciences, 15,241–248.
Civco, D. L., Hurd, J. D., Wilson, E. H., Arnold, C. L., & Prisloe, S.
(2002). Quantifying and describing urbanizing landscapes in the
northeast United States. Photogrammetric Engineering and
Remote Sensing, 68,1083–1090.
Clarke, K. C., & Gaydos, L. (1998). Loose-coupling a cellular automaton
model and GIS: long-term urban growth prediction for San
Francisco and Washington Baltimore. International Journal of
Geographical Information Science, 12,699–714.
Congalton, R. G., & Green,K. (2009). Assessingthe accuracy of remotely
sensed data: Principles and practices. Boca Raton: CRC Press
Taylor & Francis Group.
Congalton, R. G., Oderwald, R. G., & Mead, R. A. (1983). Assessing
Landsat classification accuracy using discrete multivariate analysis
J Indian Soc Remote Sens

statistical techniques. Photogrammetric Engineering and Remote
Sensing, 49,1671–1678.
Dessì, F. G., & Niang, A. J. (2008) Thematic mapping using quickbird
multispectral imagery in Oung El-Jemel area, Tozeur (SW Tunisia).
In Desertification and risk analysis using high and medium resolu-
tion satellite data (207–212) by Springer.
Dowall, D. E., & Monkkonen, P. (2008). Chennai urban land market
assessment. International Real Estate Review, 11(2), 142–165.
Duda, R. O., Hart P. E., & Stork, D. G. (2000). Pattern classification,A
Wiley-Interscience Publication, Second Edition, ISBN 9814-12-
602-0.
Eastman, J. R. (2009). IDRISITaiga, guide to GIS and remote processing.
Guide to GIS and remote processing. Worcester: Clark University.
Epstein, J., Payne, K., & Kramer, E. (2002). Techniques for mapping
suburban sprawl. Photogrammetric Engineering and Remote
Sensing, 68,913–918.
Guan, D. J., Gao, W. J., Watari, K., & Fukahori, H. (2008). Land use
change of Kitakyushu based on landscape ecology and Markov
model. Journal of Geographical Sciences, 18,455–468.
He, C. Y., Okada, N., Zhang, Q. F., Shi, P. J., & Li, J. G. (2008).
Modelling dynamic urban expansion processes incorporating a po-
tential model with cellular automata. Landscape and Urban
Planning, 86,79–91.
He, J., Liu, Y., Yu, Y., Tang, W., Xiang, W., & Liu, D. (2013). A coun-
terfactual scenario simulation approach for assessing the impact of
farmland preservation policies on urban sprawl and food security in
a major grain-producing area of China. Applied Geography, 37,
127–138.
Herold, M., Scepan, J., & Clarke, K. C. (2002). The use of remote sensing
and landscape-metrics to describe structures and changes in urban
land uses. Environment and Planning, 34,1443–1458.
Herold, M., Goldstein, N., & Clarke, K. (2003a). The spatio-temporal
form of urban growth: measurement, analysis and modeling.
Remote Sensing of Environment, 85,95–105.
Herold, M., Goldstein, N. C., & Clarke, K. C. (2003b). The spatiotempo-
ral form of urban growth: measurement, analysis and modeling.
Remote Sensing of Environment, 86,286–302.
Herzog, F., & Lausch, A. (2001). Supplementing land-use statistics with
landscape metrics: some methodological considerations.
Environmental Monitoring and Assessment, 72,37–50.
Jensen, J. R. (1986). Introductory digital image processing, a remote
sensing perspective. Englewood Cliffs: Prentice-Hall.
Ji, W., Ma, J., Twibell, R. W., & Underhill, K. (2006). Characterizing
urban sprawl using multi-stage remote sensing images and land-
scape metrics. Computers, Environment and Urban Systems, 30,
861–879.
Johnson, M. P. (2001). Environmental impacts of urban sprawl: a survey
of the literature and proposed research agenda. Environment &
Planning A, 33,717–735.
Kumar, A., Pandey, A. C., Hoda, N., & Jeyaseelan, A. T. (2011a).
Evaluating the long-term urban expansion of Ranchi Township.
India using Geospatial Technology. Journal of Indian Society of
Remote Sensing, 39(2), 213–124.
Kumar, A., Pandey, A. C., Hoda, N., & Jeyaseelan, A. T. (2011b).
Evaluation of urban sprawl pattern in tribal dominated cities of
Jharkhand State, India. International Journal of Remote Sensing,
32, 7651–7675.
Leitao, A. B., & Ahern, J. (2002). Applying landscape ecological con-
cepts and metrics in sustainable landscape planning. Landscape and
Urban Planning, 59,65–93.
Lillesand, T. M., & Kiefer, R. W. (2002). Remote sensing and image
interpretation, Fourth Edition (215–216). Wiley ISBN 9971-51-
427-3.
Lu, D., & Weng, Q. (2007). A survey ofimage classification methods and
techniques for improving classification performance. International
Journal of Remote Sensing, 28,823–870.
Matsuoka, M., Hayasaka, T., Fukushima, Y., & Honda, Y. (2004).
Analysis of the land cover and its change over yellow river basin
using satellite data. The International Archives of the
Photogrammetry Remote Sensing and Spatial Information
Sciences, 35,836–839.
McGarigal, K., & Marks, B. J. (1995). FRAGSTATS: spatial pattern
analysis program for quantifying landscape structure. USDA
Forest Service General Technical Report PNW-351.
Mundia, C. N., & Murayama, Y. (2010). Modeling spatial processes of
urban growth in African cities: a case study of Nairobi city. Urban
Geography, 31,259–272.
Nassauer, J. I., Corry, R. C., & Cruse, R. M. (1999). The landscape in
2025 alternative, future landscape scenarios: a means to consider
agricultural policy. Journal of Soil and Water Conservation, 57,
44–53.
O’Neill, R. V., Ritters, K. H., Wichham, J. D., & Jones, K. B. (1999).
Landscape pattern metrics and regional assessment. Ecosystem
Health, 5,225–233.
Opeyemi, Z. (2008). Monitoring the growth of settlements in Ilorin,
Nigeria (a GIS and remote sensing approach). The International
Archives of the Photogrammetry, Remote Sensing and Spatial
Information Sciences, 37,225–232.
Pathan, S. K., Shukla, V. K., Patel, R. G., & Mehta, K. S. (1991). Urban
land use mapping - A case study of Ahmedabad city and its envi-
rons. Journal of the Indian Society of Remote Sensing, 19(2), 95. I
12.
Praveen, S., Kabiraj, S., & Bina, T. (2013). Application of a hybrid cel-
lular 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.
Rahman, Y., Kumar, S., Fazal, S., & Bhaskaran. (2011). Urbanization and
quality of urban environment using remote sensing and GIS tech-
niques in East Delhi-India. Journal of Geographic Information
System, 3,61–83.
Ramachandra, T. V., & Kumar, U. (2008). Wetlands of Greater
Bangalore, India: automatic delineation through pattern classifiers.
Electronic Green Journal, 26,https://escholarship.org/uc/item/
3dp0q8f2#page-3
Ramachandra, T. V., Bharath, A. H., & Durgappa, D. S.(2012a). Insights
to urban dynamics through landscape spatial pattern analysis.
International Journal of Applied Earth Observation and
Geoinformation, 18,329–343.
Ramachandra, T. V., Bharath, A. H., & Sreekantha, S. (2012b). Spatial
metrics based landscape structure and dynamics assessment for an
emerging Indian Megalopolis. International Journal of Advanced
Research in Artificial Intelligence, 1(1), 48–57.
Ramachandra, T. V., Bharath, H. A., Vinay, S., Kumar, U., & Joshi, N. V.
(2013a).Modelling urban revolution in Greater Bangalore, India, In
the proceedings of 30th Annual In-house Symposium on Space
Science and Technology, 7–8 November 2013, ISRO-IISc Space
Technology cell, Indian Institute of Science, Bangalore
Ramachandra, T. V., Bharath, A. H., & Sowmyashree, M. V. (2013b).
Analysis of spatial patterns of urbanisation using geoinformatics and
spatial metrics. Theoretical and Empirical Researches in Urban
Management, 8(4), 5–24.
Ramachandra, T. V., Bharath, H. A., & Sowmyashree, M. V. (2014).
Urban structure in Kolkata: metrics and modeling through geo-in-
formatics. Applied Geomatics, 6(4), 229–244.
Ramachandra, T. V., Bharath, H. A., & Shreejith, K. (2015a). GHG foot-
print of major cities in India. Renewable and Sustainable Energy
Reviews, 44,473–495.
Ramachandra, T. V., Bharath, H. A., & Sowmyashree, M. V. (2015b).
Monitoring urbanization and its implications in a mega city from
space: spatiotemporal patterns and its indicators. Journal of
Environmental Management, 148,67–91.
J Indian Soc Remote Sens

Setturu, B., Bharath, H. A., Durgappa, D. S., & Ramachandra, T. V.
(2012). Landscape dynamics through spatial metrics, Proceedings
of India GeoSpatial Conference, Epicentre, Gurgaon, India, 7–9
February, 2012.
Sharma, R., & Joshi, P. K. (2013). Monitoring urban landscape dynamics
over Delhi (India) using remote sensing (1998–2011) inputs.
JournaloftheIndianSocietyofRemoteSensing,41(3), 641–650.
Singh, A. (1989). Digital change detection techniques using remotely
sensed data. International Journal of Remote Sensing, 10, 989–
1003.
Sudhira, H. S., Ramachandra, T. V., & Jagdish, K. S. (2004). Urban
sprawl: metrics, dynamics and modelling using GIS. International
Journal of Applied Earth Observation and Geoinformation, 5(1),
29–39.
Torrens, P. M. (2000). How cellular models of urban systems work,
CASA Centre for Advanced Spatial Analysis, University
College London, Working paper 28, URL http://www.casa.
ucl.ac.uk/working_papers/paper28.pdf (accessed 15th
May 2015)
Weber, C. (2001). Remote sensing data used for urban agglomeration
delimitation. In J. P. Donnay, M. J. Barnsley, & P. A. Longley
(Eds.), Remote sensing and urban analysis (pp. 155–167).
London: Taylor and Francis.
Wentz, E. A., Anderson, S., Fragkias, M., Netzband, M., Mesev, V.,
Myint, S. W., Quattrochi, D., Rahman, A., & Seto, K. C. (2014).
Supporting global environmental change research: a review of
trends and knowledge gaps in urban remote sensing. Remote
Sensing, 6(5), 3879–3905.
Yang, X., Ren, L., Yong, B., & Jiao, D. (2008). Remote sensing-based
land use and land cover change in Shalamulun catchment. IEEE
Computer Society, 2,504–507.
Yorke, C., & Margai, F. (2007). Monitoring land use change in the Densu
river basin, Ghana using GIS and remote sensing. African
Geographical Review, 26,87–110.
J Indian Soc Remote Sens