Академический Документы
Профессиональный Документы
Культура Документы
To cite this article: Kuntal Ganguly, Rajiv Kumar, K. Mruthyunjaya Reddy, P. Jagadeeswara Rao,
Manoj Raj Saxena & G. Ravi Shankar (2016): Optimization of spatial statistical approaches to
identify land use/ land cover change hot spots of Pune region of Maharashtra using Remote
Sensing and GIS techniques, Geocarto International, DOI: 10.1080/10106049.2016.1178813
Article views: 13
Download by: [Library Services City University London] Date: 26 April 2016, At: 06:25
Publisher: Taylor & Francis
Journal: Geocarto International
DOI: http://dx.doi.org/10.1080/10106049.2016.1178813
Hyderabad, 500037
c
Department Of Geo-Engineering & Resource Development
Visakhapatnam-530 003
Full telephone, E-mail Senior Researcher, Land Use and Cover Monitoring Division,
Phone: 04023884278
Mobile: 7303916539
Optimization of spatial statistical approaches to identify land use/ land cover change hot spots of
Name of authors:
Kuntal Gangulya*
Rajiv Kumara
K. Mruthyunjaya Reddyb
Downloaded by [Library Services City University London] at 06:25 26 April 2016
P. Jagadeeswara Raoc
G. Ravi Shankara
Affiliations:
a
Land Use and Land Cover Monitoring Division, National Remote Sensing Centre, ISRO,
Senior Researcher, Land Use and Cover Monitoring Division, National Remote Sensing
Phone: 04023884278
Mobile: 7303916539
Key words
Land use / land cover; LULCC; hot spot; spatial statistics; remote sensing; Getis-Ord Gi*
Downloaded by [Library Services City University London] at 06:25 26 April 2016
No of Table: 15
No of Figure: 13
Abstract
This study investigated land use / land cover change (LULCC) dynamics using temporal satellite
images and spatial statistical cluster analysis approaches in order to identify potential LULCC hot
spots in the Pune region. LULCC hot spot classes defined as new, progressive and non-progressive
were derived from Gi* scores. Results indicate that progressive hot spots have experienced high
growth in terms of urban built-up areas (20.67% in 1972-92 and 19.44% in 1992-12), industrial areas
Downloaded by [Library Services City University London] at 06:25 26 April 2016
(0.73% in 1972-92 and 3.46% in 1992-12) and fallow lands (4.35% in 1972-1992 and -6.38% in 1992-
12). It was also noticed that about 28.26% of areas near the city were identified as new hot spots after
1992. Hence, non-significant change areas were identified as non-progressive after 1992. The study
demonstrated that LULCC hot spot mapping through the integrated spatial statistical approach was an
effective approach for analysing the direction, rate, spatial pattern and spatial relationship of LULCC.
Keywords: Land use / land cover; LULCC; hot spot; spatial statistics; remote sensing; Gi* score
1. Introduction
There is unequal urban growth, which is taking place all over the world, but the rate of urbanization is
very fast in developing countries, especially in Asia. In 1800 A.D., only 3% of the world’s population
lived in urban centres, but this figure reached 14% in 1900 and in 2000, about 47% (2.8 billion) people
were living in urban areas. India no longer lives in villages and 79 million people were living in urban
areas in 1961, but it went up to 285 million in 2001. In India and China alone, there are more than 170
urban areas with populations of over 750,000 inhabitants (United Nations Population Division 2001).
Statistics show that India’s urban population is the second largest in the world after China, and is
higher than the total urban population of all countries put together barring China, USA and Russia. In
1991, there were 23 metropolitan cities in India, which increased to 35 in 2001 (Census of India 2001).
With rapid urbanization, there has been tremendous growth in population and buildings in cities,
which lead to the drastic reduction in the greenery area and is opposite to the increase in impervious
area. In India, Pune city and its surrounding villages are now home to more than 3.1 million people
and growing at a rapid pace. It has experienced rapid land use / land cover dynamics in the last four
decades. The galloping process of industrialization, urbanization and globalization has brought
mounting environmental problems, including climate change, water shortage and pollution (Ganguly
Remote sensing data have been an attractive source in the determination of land cover thematic
mapping, providing valuable information for delineating the extent of land cover classes, as well as for
performing temporal land cover change analysis and risk analysis at various scales (Kavzoglu &
Colkesen 2009). Remotely sensed data, as a result, is often highly spatially auto correlated and
various techniques have been developed to assess the spatial dependence characteristics of
Downloaded by [Library Services City University London] at 06:25 26 April 2016
remotely sensed imagery. The Getis statistic provides a measure of spatial dependence for each
pixel while also indicating the relative magnitudes of the digital numbers in the neighbourhood
of the pixel (Wulder & Boots 2010). A hot spot, as the name suggests, is a state of indicating some
arrangements of clusters in a spatial distribution. One of the most popular approaches for the detection
of LULCC hot spots is cluster analysis. Cluster analysis can be an effective method for determining
particularly challenging task to detect hot spots using clustering techniques. It is mainly because of the
uncertainty associated with the appropriate number of LULCC clusters to be generated, as well as
establishing the significance of the individual clusters identified. Not surprisingly, research indicates
that certain areas are more prone to higher concentrations of land use and land cover change and
widely labelled as ‘hot spots’, such areas are often observed as most dynamic in nature. Where
resources are constrained, the identification of hot spots is helpful because the significance of change
and identification of an area cannot be estimated through operational approaches. Operationally, the
delineation of hot spot boundaries is somewhat arbitrary. LULCC density was measured over a
continuous area. Therefore, the boundaries separating hot spots of the change areas from areas without
enough change to merit the label hot spot are perceptual constructs. Moreover, depending on the scale
of geographic analysis, a hot spot can mean very different things (Harries 1999). An essential need in
the use of non-hierarchical approaches for the detection of hot spots is to better understand what
makes certain clusters more significant in the context of hot spot analysis. In other words, what makes
a hot spot ‘hot’ and what makes other clusters ‘cold’. A body of research does exist for approaching
this issue, though its use in the context of hot spot detection is uncultivated. The results of the CMRC
(1998) study suggest that a good approach for detecting hot spots is tests of spatial autocorrelation. As
demonstrated by Jefferis (1999), the implementation of the Getis-Ord statistic (Gi statistic) in
SpaceStat provided very good measures of crime hot spots for Baltimore County. The utility of spatial
autocorrelation and the Gi statistic for hot spot analysis is further supported in the work of Craglia et
al. (2000). Ghimire et al. (2010) have also demonstrated that the use of the Getis statistic with
different neighbourhood sizes leads to substantial increase in per class classification accuracy of
2. Study Area
The study area makes up a part of Pune district, which includes four taluks, namely, Haveli, Khed,
Marvel and Mulshitaluks, which cover an area about 4840 sq. km and lies between 73 o 18’ 50” E to
74o 12’ 03” E and 18o 10’ 51” N to 19o 05’ 53” N (Figure 1). River Pavana, Mula and Mutha are major
river systems flowing through the study area. The geomorphological setting of the study area shows a
backdrop of hills on the western side, with steeper slopes and rocky red soils. The area is underlaid by
basaltic lava flows of upper cretaceous Eocene age associated with basic intrusive. The soil texture
contains alluvial deposits of sand, gravels, fine silts and clays along the bank of the rivers.
INSERT FIGURE 1
3. Methods
The methodology adopted for this research ranges from data collection, processing, analysis and
generation of other geospatial datasets for the result and analysis. The conceptualized framework of
INSERT FIGURE 2
Landsat MSS of December 21, 1972 and Landsat TM image of February 04, 1992 and IRS
Resourcesat-2 LISS-3 of March 17, 2012 were selected for use in this study. Images were
geometrically corrected and georeferenced to the Universal Transverse Mercator zone 43N coordinate
system by using WGS-84 datum. The specification of the satellite images is given in Table 1.
INSERT TABLE 1
Village boundaries were created using a mosaic of the cadastral boundary for each village and were
used to extract the information on land use / land cover change for a particular village.
Experimental and operational classifications are two different approaches for large area land cover
mapping and monitoring (Defries & Townshend 1999; Hansen & Loveland 2012). From the point of
view of land use / land cover (LULC) mapping at multi resolution (23m-56m), the selection of
appropriate classes suitable for complex landscapes, as well as quality of the resulting data products
Downloaded by [Library Services City University London] at 06:25 26 April 2016
needs to be addressed. To identify dominant LULC classes in the study area, visual interpretation has
been carried out using on screen digitization (Roy et al. 2015). Manual or visual classification of
remotely sensed data is an effective method of classifying land cover, especially when the analyst is
familiar with the area being classified. It relies on the interpreter employing visual cues such as tone,
texture, shape, pattern and relationship to other objects to identify the different land cover classes
(NRSC/ISRO 2010). Our classification scheme, with level 2 classes (Table 2), was based on the land
use / land cover classification system developed by NRSC/ISRO (2011) for interpretation of remote
INSERT TABLE 2
Depending on the availability of LULC of the study area, ten dominant LULC classes were extracted
in the region. Urban, rural, industries, cropland, fallow land, forest, tree clad, wasteland, river and
Accuracy implies a comparison between the map and reference information and this comparison
requires rules to carry it out consistently (Strahler et al. 2006). An independent sample of 300 random
points was collected from ground verification and same points were used to find out class accuracy of
the predicted (classified) image using ArcGIS 10.2.1. Error matrices as cross-tabulations of the
mapped class vs. the reference class were used to assess classification accuracy (Congalton & Green
1999). Class accuracy, observed accuracy, expected accuracy and the Kappa statistic were then
derived from the error matrices. The Kappa statistics incorporates the off diagonal elements of the
error matrices (i.e. classification errors) and represents agreement obtained after removing the
Following the classification of imagery from the individual years, a multi-date post
classification comparison change detection field was used to determine changes in land use /
land cover in two intervals i.e. 1972–1992 and 1992-2012. This is perhaps the most common
approach to change detection (Jensen 2004) and has been successfully used by NRSC/ISRO
Downloaded by [Library Services City University London] at 06:25 26 April 2016
(2011) to monitor land use / land cover changes at 1: 50000 scale all over India. The post-
classification approach provides ‘‘from–to’’ change information and the kind of landscape
transformations that have occurred can be easily calculated and mapped (Yuan et al. 2005). A
change detection map with 13 and 28 combinations of ‘‘from–to’’ change information were
4. Spatial Statistics
The basic principle of cluster analysis in spatial statistics is the task of grouping a set of objects in
such a way that aims to collect or cluster objects into the same group that is more similar to each other
than to those in other groups. It is defined as the correlation of a variable with itself through space. It
evaluates the effectiveness of spatial correlation and tests the assumption of independence (or
entropy). If there are any systematic patterns in the spatial distribution of a variable, that variable is
said to be spatially correlated, if nearby areas are alike, the correlation is positive. Negative correlation
applies to neighbouring areas that are unlike and random patterns exhibit no significant spatial cluster.
In spatial statistics, cluster analysis does not explain why locations that bear clusters of statistically
significant land use / land cover change have a higher incidence of changes than other positions; thus,
this method cannot identify factors that cause land use / land cover change.
& Getis 1995). They represent a global spatial autocorrelation index, which is a local spatial
autocorrelation index also. It is more suitable for discerning cluster structures of high or low
(1)
Where, Gi* is the SA statistic of an event i over n events and wij is the measure of the spatial
proximity between regions i and j. The term, xj, characterizes the magnitude of the variable x at events
Downloaded by [Library Services City University London] at 06:25 26 April 2016
j over all n. The distribution of the Gi* statistic is normal when normality is observed in the
underlying distribution of the variable x. The threshold distance (i.e., the proximity of one land
transformation to another) in this subject field was set to zero to designate that all characteristics were
considered neighbours of all other features. This threshold was applied over the entire region of the
study. The standardized Gi* is essentially a Z-value and can be associated with statistical significance
as:
(2)
Positive and negative Gi* statistics with high absolute values correspond to clusters of transforming
land with high and low-value events, respectively. A Gi* close to zero implies a random distribution
GIS database design and utilization has a great advantage to utilize the data repeatedly without
disturbing the original dataset. In the present study, the classified data in GIS vector format were used
to find out the land use / land cover change over the study area using a spatial query based technique.
The same dataset containing land use / land cover class fields was utilized to derive the changes for
each of the two ten-class maps. The database for change layers contained combinations of land
transformation and the combinations were achieved through querying each land use / land cover class
at temporal scale. As the statistical analysis deals with numeric values, therefore, all the LULCC
combinations were assigned a unique value. In the present study, the Getis-Ord Gi* test was carried
out using the ArcGIS 10.2.1 spatial statistical tool to calculate the LULCC hot spots. The resultant Z-
scores were identified in the cluster spatially with high and low values. Therefore, the feature with a
high Z-score value may not be a statistically significant hot spot if not beleaguered by other features
with high values as well. For simplicity of data analysis and ease of implementation, we decided to
transform LULCC data at 100m x 100m fixed grid cell which was further utilized for zonation of the
hot spot. The Gi* statistic returned for each polygon in the dataset with a Gi* score, was further
converted into point information for each grid. The points were used with ordinary kriging to establish
Downloaded by [Library Services City University London] at 06:25 26 April 2016
a contiguous map listing all other prediction points through the chosen model. The method of
interpolation or kriging estimated a value of an unknown variable in regions not sampled. Hence, the
spatial variation of all points was mapped using the Gi* score, which categorized the study area into
cold spots (minimum spatial change area) and hot spots (maximum spatial change area). Depending on
the dimensionality and dynamics of the change region over time, the hot spots were further
categorized into new, progressive and non-progressive. Simultaneously, the number of villages of the
study area was also depicted under different categories of the hot spot and to identify villages prone to
urbanization.
Error matrices were used to assess classification accuracy and are summarized in Table 3. LULC
classes like rural built up, river and water bodies have achieved 97% of overall accuracy throughout
the image; whereas, the rest of the classes were having above 80% of class accuracy except cropland
(77%), tree-clad (77%) and wasteland (73%), with a Kappa statistic of 83.7 %. As the study area has
heterogeneous characteristics of LULC, medium resolution data was not sufficient to some extent to
classify the objects, thus the three classes, i.e. cropland, tree-clad and wasteland has achieved lower
INSERT TABLE 3
change statistics for the three years are summarized in Table 4. From 1972 to 2012 built-up urban and
industrial areas have increased by 7.13% and 1.64% of the total geographic area (TGA) respectively,
with the greatest increase occurring from 1992 to 2012, while a constant an overall decrease in
croplands have also been observed by 6.92% of TGA from 1972 to 2012. A substantial alteration was
observed in wastelands which show a considerable decrease of 3.85% in terms of TGA. It has also
been noted a significant increase in the water body class due to the construction of new reservoirs
from 1972-2012 (Figure 3). Although the extent of water bodies may change from year to year due to
Downloaded by [Library Services City University London] at 06:25 26 April 2016
INSERT TABLE 4
INSERT FIGURE 3, 4
To further evaluate the results of LULC conversions, matrices of LULCC from 1972 to 1992 and 1992
to 2012 were created. In the table, unchanged areas are located along the major diagonal of the matrix.
Table 5 shows the land use / land cover change matrix from 1972 to 1992. From the table, it is
observed that there was a considerable interclass change during the 20 year period from 1972 to 1992
when major land use / land cover e.g. croplands and wastelands have been won over to urban built-up.
It was also noted that due to the construction of new reservoirs some part of agricultural lands were
transformed into water bodies. In comparison with the change matrix of 1972-1992, more significant
interclass changes have been observed from 1992 to 2012. Table 6 demonstrates the impact of
urbanization and industrialization on land use / land cover. It was observed that the land shifts in urban
areas have been led from almost all the classes, where cropland, fallow lands, and wastelands
contributed 113.38 sq. km, 56.67 sq. km. and 46.99 sq. km. to urban area, respectively. Along the
similar lines of urban expansion, major land transformations have also been observed in every year.
Major contributions to industries were observed from cropland (22.92 sq. km), fallow land (16.61 sq.
km) and wasteland (40.62 sq. km). The cross-tabulation (Table 7) of the 1972 and 2012 land use/ land
cover maps revealed that most of the losses were converted to urban or built-up land (394.48 sq. km),
water bodies (111.98 sq. km) and industrial (90.81 sq. km).
INSERT TABLE 5, 6, 7
The dominance of LULCC combinations was examined with respect to its total number of occurrences
during 1972-1992 and 1992-2012. Table 8 shows increases of LULCC class combinations from 13 to
28 during 1972-1992 and 1992-2012, respectively. It was also observed that in 14.93% of the cases the
change was ‘‘croplands to urban’’ during 1972-1992, which increased up to 19.07% during 1992-
2012. Fallow lands also contributed about 8.19% area to urban expansion during 1992-2012. Whereas,
the change from “croplands to water bodies” has decreased from 20.02% to 12.35% during 1972 to
2012. Table 8 also reveals considerable increases in industrial areas by 8.7% utilizing both cropland
Downloaded by [Library Services City University London] at 06:25 26 April 2016
and wastelands.
INSERT TABLE 8
5.3. Detecting land use / land cover change hot spots and zonation
Hot spot detection aims at identifying subgroups in the observations that are unexpected and are very
significant with respect to baseline information. In the present study the land use / land cover change
hot spots have been calculated using the Z-score, which distinguishes the region with higher
concentration of LULCC clusters. On the other hand, the changes with very low Z-scores or very
lower changes were identified as a cold spot. Similarly, the range of Z-scores was further categorized
on the basis of the Gi* value which varies from +3 to-3. Understanding the nature of clustering
overtime, the output Gi* values were gridded into 100m x 100m cells (Figure 5 and Figure 6). On the
other hand, to assess the severity of hot spots from 1972 to 2012 a GIS query based analysis was
performed. Zonation of the hot spots was also being performed using a kriging interpolation technique.
The results showed the entire study area has been divided into three zones as, new, progressive and
non-progressive hot spots (Figure 7). It was noted from the analysis of the LULCC hot spot map of
1972-1992 and 1992-2012, 190 new villages of the study area have experienced very significant
LULC changes of about 28.26% of the total village area during the last 20 years (Figure 8). These new
villages were categorized under new hot spots. Likewise, it was also observed that 27 villages of the
study area which were defined as progressive hot spots (covering an area about 54% of the villages)
continuing their growth since the last 40 years (Figure 9). Significant land transformations have not
been perceived for non-progressive hot spots during last two tens. Although 18 villages (80% area)
within non-progressive hot spots have shown industrial expansion during 1972 to 1992, however, land
transformation became more stable after 1992. Therefore, hot spots were not identified during 1992-
INSERT TABLE 9
INSERT FIGURE 5, 6, 7, 8, 9, 10
5.4. Land use / land cover change characteristics of hot spot regions
The geometry of land development and land use patterns was achieved through the identification of
Downloaded by [Library Services City University London] at 06:25 26 April 2016
hot spots that further provided detailed spatial information for land use planning. In the present study,
GIS-based cluster analysis provides a utilitarian instrument to distinguish hot spots and to enforce the
land transformation morphological approach to identify dynamic villages. Hence, three major classes-
new, progressive and non-progressive hot spots were obtained from the analysis, which was useful to
Analysis of hot spots at temporal scale has identified few clusters which had higher and lower
concentration over the study area. It was found from the analysis of the last 40 years (1972-1992 and
1992-2012) of land transformation and hot spot delineation, about 190 new villages of Pune city and
surrounding regions have had a higher concentration of land use / land cover changes after 1992.
INSERT FIGURE 11
Figure11 shows adjoining villages of the Pune and Pimpri-Chinchwad city areas which have been
debated as new hot spots have continuous and higher rates of LULC changes since 1992. Hence, a
higher pace of land transformation has been taken place in the 190 villages of surrounding Pune and
Pimpri-Chinchwad. Rapid industrialization was also noted after 1992 along with urbanization reducing
croplands and wastelands at a higher pace. According to the Pune Business Directory & Information
Portal, the industrial development of Pune and Pimpri-Chinchwad and its environs covers a span of the
last forty years or so, though there had certainly been a few isolated attempts to start industries in these
cities earlier. The emergence of industrial Pune and Pimpri-Chindwad began with mechanical
engineering industries forming the base. Pune’s proximity to Mumbai, good climate and availability of
talent made it a preferred destination for large firms like Tata motors (TELCO then), Buckau Wolf
(Thyssen Krupp now), KSB Pumps, Hindustan Antibiotics, and several others.
Table 10 shows the inter-class transformation of land use / land cover from 1972 to 1992. Very
nominal changes were observed during 1972 to 1992, although considerable land transformations from
cropland to urban areas and reservoir were reported during this time period. Some amounts of
wastelands also decreased due to urban expansion. Table 11 shows the continuation of urban
Downloaded by [Library Services City University London] at 06:25 26 April 2016
expansion and remarkable industrial growth during 1992 to 2012. It was noticed about 29.78 sq. km
and 11.44 sq. km of croplands was converted to urban built-up and industries since 1992, respectively.
The agricultural reduction was also identified due to the setup of new reservoirs in new hot spot
regions, where 4.31 sq. km of croplands was converted to water bodies. Aside from agricultural
reduction, it was also found a substantial reduction of fallow lands (10.40 sq. km) and wastelands
The Getis-Ord Gi* statistical test has already shown some certain regions of the study area which were
having a gradual increase of the Z-score from 1972 to 2012. It was also identified from the statistical
analysis that, about 27 villages of the study region experienced an increasing pace of land
transformation which accelerated the growth from 95% to 99% confidence level during 1992-2012.
INSERT FIGURE 12
Figure 12 shows an overall higher rate of urbanization (above 35% of the hot spot village area),
whereas marginal changes (below 5%) were noted in the case of industrial expansion. On the other
hand, gradual declination of croplands and wastelands were also observed from 1972 to 2012. It caters
for the twin city of Pimpri-Chinchwad and surrounding regions, developed primarily due to rapid
urbanization and industrial development. Although, the new special township policy act of the
Maharashtra government has put down an integrated access to the development of townships to
decongest Pune Municipal Corporation areas and encourage new settlements in the fringe, thus most
of the renowned developers in Pune are developing integrated townships (Source: Project Gutenberg
Self-Publishing Press).
In Table 12 the land use / land cover change matrix of 1972-1992 clearly reveals that about 37.15 sq.
km of the area has been converted from cropland to urban built-up, which was 35.86 sq. km during
1992-2012. Likewise, a substantial decrease of wastelands (57.86 sq. km) and conversion into urban
area has also been observed during 1972-1992, whereas comparatively, the contribution to urban area
was less (23.32 sq. km) during 1992-2012 (Table 13). It also exhibits a considerable reduction of
Downloaded by [Library Services City University London] at 06:25 26 April 2016
fallow lands (27.61 sq. km) which were transformed into urban built-up during last two decades.
The cluster analysis through the Getis-Ord Gi* statistical test has revealed there was a parallel growth
of urban areas along with the rapid industrialization near Pune and Pimpri-Chinchwad in the recent
past. It was noticed that the less changed areas have low clusters which were considered as hot spots
during 1972-92, but the same places were not considered during 1992-12. It was also noticed that
about 18 villages near the Pune and Pimpri-Chinchwad region of Pune district showed less growth
after 1992. Figure13 shows the villages have experienced an overall 1.5 % increase of urban area and
industrial area from 1972 to 2012. In comparison with progressive and new hot spots, the amount of
change was much less. Reduction of agricultural lands and wastelands were continued like other hot
INSERT FIGURE 13
Distribution of LULC classes and their interclass transformation has been analysed using a change
matrix. Comparative analysis of the change matrix of 1972-1992 and 1992-2012 (Table 14 and Table
15) depicts that most of the land transformations occurred during 1992-2012. It was also noticed that
the region experienced urbanization for the first time after 1992. Although, during the 1972-1992 time
period there was a negligible amount of conversion of wastelands (0.04 sq. km) into industrial land
which was comparatively more (0.82 sq. km) during 1992-12. It was also noticed that reductions in
agricultural lands had an immense contribution to urban built-up (14.21 sq. km) and industrial area
6. Conclusion
In this paper, an attempt has been made to assess the long-term land use / land cover dynamics for part
of the Pune district of India. The study has also been extended to establish the geospatial model to
identify land use / land cover change hot spots. A methodology was developed and Getis-Ord Gi*
spatial statistical cluster analysis approach was employed to identify LULC changed hot spots. Last 40
years of land use / land cover change analysis and the hot spot map obtained from the cluster analysis
Downloaded by [Library Services City University London] at 06:25 26 April 2016
has identified potential areas of future LULC change. The high priority hot spots of LULCC have been
identified within the suburban zone, whereas northern and western parts of the study area, which are
relatively less accessible due to the rugged terrain, are less prone to change and hence fall under low-
priority change areas. The proposed LULC change hot spot identification and the application of Getis-
Ord Gi* statistical test in LULC change analysis differed from earlier models and this technique can
be used in the same way to identify changes to hot spots of other natural resources.
On the other hand, the combined application of high-resolution historical imageries like
CORONA along with existing high-resolution satellite imageries can provide more accurate
Acknowledgements
Authors gratefully acknowledge the support of Land Use and Cover Monitoring Division of
NRSC/ISRO to complete the work. Authors place on record their deep sense of gratitude to Dr. V. K.
Dadhwal, Director, NRSC/ISRO and Dr. P. G. Diwakar, Dy. Director, NRSC/ISRO for their support
and motivation to do the work. Authors are also grateful to Dr. T. Ravi Shankar, Group Director,
References
District census handbook: Census of India [Internet]. 2001. New Delhi. Available from:
http://www.censusindia.net/.
Congalton RG, Green K. 1999. Assessing the accuracy of remotely sensed data: Principles and
Crime Mapping Research Center. 1998. Hot spot, Mapping Crime and Geographic
Defries RS, Townshend JRG. 1999. GLC characterization from satellite data: from research to
environment and its associated biophysical parameters using Remote Sensing and GIS
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Getis A, Ord JK. 1992. The analysis of Spatial Association by use of Distance Statistics. Geogr Anal.
24(3):189-206.
spatial dependence in land-cover classification models using random forests and the
Harries K. 1999. Mapping Crime: Principle and Practice. National Institute of Justice:
Jensen JR. 2004. Digital change detection. Introductory digital image processing: A remote
Kavzoglu T, Colkesen I. 2009. A kernel functions analysis for support vector machines for land cover
Manepalli URR., Bham GH, Kandada S. 2011. Evaluation of Hot spot identification using kernel
density estimation and Getis-Ord (Gi*) on I-630. In: 3rd International Conference on
NRSC, ISRO: Ebook on Remote Sensing Applications [Internet]. 2010. Hyderabad: National
Ord JK, Getis A. 1995. Local Spatial Autocorrelation Statistics: Distributional Issues and an
Roy PS, Roy A, Joshi PK, Kale MP, Srivastava VK, Srivastava SK, Dwevidi RS, Joshi C, Behera
MD, Meiyappan P, et al. 2015. Development of Decadal (1985-1995-2005) Land Use and
Songchitruksa P, Zeng X. 2010. Getis-Ord Spatial Statistics for Identifying Hot spots Using
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Incident Management Data. In: Transportation Research Board (TRB) 89th Annual
Strahler AH, Boschetti L, Foody GM, Friedl MA, Hansen MC, Herold M, Mayaux P, Jeffrey T,
Stehman SV, Woodcock CE. 2006. Global Land Cover Validation: Recommendations for
World Population Prospects: The 2000 Revision [Internet]. 2001. New York: United Nations
wpp2000.
imagery assessed with the Getis statistic. Int J Remote Sens. 11(19): 2223-2231.
Yuan F, Sawaya KE, Loeffelholz BC, Bauer ME. 2005. Land cover classification and change analysis
of the Twin Cities (Minnesota) Metropolitan Area by multitemporal Landsat remote sensing.
Table 1
Land use /
land cover Descriptions
classes
Urban Residential areas, mixed built‐up, recreational places, public / semi‐public utilities,
communications, public utilizes/facility, commercial areas, reclaimed areas, vegetated areas,
transportation.
Rural Built‐up areas, smaller in size, mainly associated with agriculture and allied sectors and
non‐commercial activities. They can be seen in clusters non‐ contiguous or scattered.
Industrial Industrial areas and their dumps, and ash/cooling ponds. surface rocks and stone quarries, sand
Downloaded by [Library Services City University London] at 06:25 26 April 2016
and gravel pits, brick kilns, etc. Areas of stockpile of storage dump of industrial raw material or
slag/effluents or waste material or quarried/mixed debris from earth's surface.
Crop lands Areas with standing crop, under agricultural tree crops planted adopting agricultural
management techniques as on the date of Satellite overpass.
Fallow lands Lands, which are taken up for cultivation but are temporarily allowed to rest, un‐cropped for one
or more season, but not less than one year.
Forests Land with a tree canopy cover of more than 10 percent and area of more than 0.5 ha, Forest
Plantations, Scrub forest.
Tree clads Trees outside forest.
Wastelands Degraded lands which can be brought under vegetative cover with reasonable effort and which
is currently underutilized and land which is deteriorating for lack of appropriate water and soil
management or on account of natural causes. It consists of salt‐affected land, gullied / ravinous
land, scrub land, sandy area, barren rocky/stony waste, rann area.
Rivers Rivers/streams are natural course of water flowing on the land surface along a definite
channel/slope regularly or intermittently towards a sea in most cases or in to a lake or an inland
basin in desert areas or a marsh or another river.
Water bodies Areas with surface water in the form of ponds, lakes, tanks and reservoirs.
Table 2
Crop Fallow Tree Waste Water
Classes Urban Rural Industries Forests River Accuracy
lands lands clads lands bodies
Urban 25 1 1 2 0 0 0 1 0 0 83%
Rural 1 29 0 0 0 0 0 0 0 0 97%
Industries 2 2 26 0 0 0 0 0 0 0 87%
Crop lands 0 0 0 23 2 0 2 2 1 0 77%
Fallow lands 0 4 0 0 25 0 0 1 0 0 83%
Forests 0 0 0 0 0 25 4 0 1 0 83%
Tree clads 0 5 2 0 0 0 23 0 0 0 77%
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Wastelands 0 4 2 1 0 0 1 22 0 0 73%
River 0 0 0 0 0 0 0 0 29 1 97%
Waterbodies 0 0 0 0 0 0 0 0 1 29 97%
Table 3
% of % of % of
1972 1992 2012 Change Change Change
1972-92 1992-12 1972-12
LULC Classes (Area in (Area in (Area in with with with
change change change
sq. km) sq. km) sq. km) respect to respect respect
TGA to TGA to TGA
Urban 58.71 231.02 453.07 172.30 3.11 222.05 4.01 394.36 7.13
Rural 22.22 23.87 35.99 1.66 0.03 12.11 0.22 13.77 0.25
Industries 16.11 23.45 106.84 7.34 0.13 83.39 1.51 90.73 1.64
Crop lands 2555.28 2386.38 2172.07 -168.90 -3.05 -214.31 -3.87 -383.22 -6.92
Fallow lands 174.25 239.22 181.21 64.97 1.17 -58.02 -1.05 6.95 0.13
Forests 905.98 905.98 898.58 0.00 0.00 -7.40 -0.13 -7.40 -0.13
Tree clads 604.33 603.18 598.02 -1.15 -0.02 -5.16 -0.09 -6.31 -0.11
Wastelands 1018.46 904.86 805.14 -113.61 -2.05 -99.72 -1.80 -213.32 -3.85
River 57.67 55.94 50.13 -1.73 -0.03 -5.81 -0.11 -7.54 -0.14
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Water bodies 121.14 160.26 233.11 39.12 0.71 72.86 1.32 111.98 2.02
Table 4
1972-1992
1972 1992 1972 Total
Crop Fallow Tree Waste Water
Urban Rural Industries Forests River
lands lands clads lands bodies
Urban 58.71 58.71
Rural 22.22 22.22
Industries 16.11 16.11
Crop lands 71.04 1.66 1.74 2386.38 65.93 28.54 2555.28
Fallow lands 173.29 0.96 174.25
Forests 905.98 905.98
Tree clads 601.69 2.64 604.33
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Table 5
1992-2012
1992 2012 1992 Total
Crop Fallow Tree Waste Water
Urban Rural Industries Forests River
lands lands clads lands bodies
Urban 230.79 0.22 231.02
Rural 3.71 19.99 0.01 0.17 23.87
Industries 0.03 23.37 0.05 23.45
Crop lands 113.38 5.18 22.92 2163.28 31.13 1.71 0.05 48.73 2386.38
Fallow lands 56.67 4.14 16.61 5.79 149.88 6.13 239.22
Forests 1.39 2.72 898.18 3.52 0.01 0.15 905.98
Tree clads 0.10 0.39 0.37 1.40 596.79 0.10 4.03 603.18
Wastelands 46.99 6.28 40.62 1.34 0.03 0.40 1.23 799.79 8.17 904.86
River 0.20 0.02 50.06 5.65 55.94
Waterbodies 160.26 160.26
Downloaded by [Library Services City University London] at 06:25 26 April 2016
2012 Total 453.07 35.99 106.84 2172.07 181.21 898.58 598.02 805.14 50.13 233.11 5534.15
Table 6
1972-2012
1972 2012 1972 Total
Crop Fallow Tree Waste Water
Urban Rural Industries Forests River
lands lands clads lands bodies
Urban 58.59 0.13 58.71
Rural 2.21 19.84 0.17 22.22
Industries 0.03 16.03 0.05 16.11
Crop lands 219.18 6.46 29.18 2168.97 49.55 1.71 0.05 80.18 2555.28
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Table 7
Feature Feature Feature
Change Type (1972-1992) % Change Type (1992- 2012) % Change Type (1992-2012) %
counts counts counts
Croplands to Fallow lands 295 34.95 Croplands to Fallow lands 206 4.54 Forests to Waterbodies 13 0.29
Croplands to Industries 17 2.01 Croplands to Forests 25 0.55 River to Croplands 5 0.11
Croplands to Rural 10 1.18 Croplands to Industries 395 8.70 River to Waterbodies 54 1.19
Croplands to Urban 126 14.93 Croplands to Rural 168 3.70 Rural to Urban 20 0.44
Croplands to Waterbodies 169 20.02 Croplands to Urban 866 19.07 Tree clads to Forests 397 8.74
Fallow lands to Waterbodies 21 2.49 Croplands to Wastelands 11 0.24 Tree clads to Industries 15 0.33
River to Waterbodies 6 0.71 Croplands to Waterbodies 561 12.35 Tree clads to Rural 13 0.29
Tree clads to Waterbodies 10 1.18 Fallow lands to Croplands 19 0.42 Tree clads to Urban 11 0.24
Wastelands to Industries 36 4.27 Fallow lands to Industries 216 4.76 Tree clads to Wastelands 4 0.09
Wastelands to River 1 0.12 Fallow lands to Rural 77 1.70 Tree clads to Waterbodies 50 1.10
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Wastelands to Tree clads 1 0.12 Fallow lands to Urban 372 8.19 Wastelands to Industries 396 8.72
Wastelands to Urban 95 11.26 Fallow lands to Waterbodies 74 1.63 Wastelands to Rural 112 2.47
Wastelands to Waterbodies 57 6.75 Forests to Industries 32 0.70 Wastelands to Urban 274 6.03
Forests to Urban 18 0.40 Wastelands to Waterbodies 138 3.04
Table 8
Non-progressive
New hot spots Progressive hot spots
Confidence Level hot spots
(Area in %) (Area in %)
(Area in %)
Cold Spot - 99% Confidence 0.21 0.00 0.00
Cold Spot - 95% Confidence 4.24 0.80 1.32
Cold Spot - 90% Confidence 4.45 0.98 4.23
Not Significant 52.22 43.83 79.87
Hot Spot - 90% Confidence 4.01 4.10 2.78
Hot Spot - 95% Confidence 6.61 6.15 3.57
Hot Spot - 99% Confidence 28.26 44.14 8.23
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Table 9
1972-1992
Table 10
1992-2012
1992 2012 1992
Total
Rura Industrie Crop Fallow Forest Tree Waste Rive Water
Urban
l s lands lands s clads lands r bodies
Urban 35.31 0.32 35.63
Rural 1.13 4.61 0.01 0.17 5.92
Industries 0.03 3.69 0.05 3.77
Crop lands 35.77 2.09 9.48 620.55 13.43 0.22 0.55 0.04 18.84 700.96
Fallow
16.96 1.6 9.29 3.6 47.98 2.1 81.54
lands
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Table 11
1972-1992
1972 1992 1972
Total
Crop Fallow Tree Waste Water
Urban Rural Industries Forests River
lands lands clads lands bodies
Urban 43.85 43.85
Rural 1.99 1.99
Industries 5.15 5.15
Crop lands 37.15 146.45 19.97 0.10 203.67
Fallow lands 14.77 14.77
Forests 17.63 17.63
Tree clads 19.19 19.19
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Table 12
1992-2012
Table 13
Table 14
1972-1992
Table 15
Figure Captions
map (b)
Figure 7. Hot spot zone map of the of the study area
Figure 8. Area coverage of villages under new hot spots
Figure 9. Area coverage of villages under progressive hot spots
Figure 10. Area coverage of villages under non-progressive hot spots
Figure 11. Land use / land cover change statistics of villages under new hot spots
Figure 12. Land use / land cover change statistics of villages under progressive hot spots
Figure 13. Land use and land cover change statistics of villages under non-progressive hot spots
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016
Downloaded by [Library Services City University London] at 06:25 26 April 2016