Content uploaded by Elnazir Ramadan
Author content
All content in this area was uploaded by Elnazir Ramadan on Mar 16, 2015
Content may be subject to copyright.
Monitoring of Urban Growth based on Changes in NDVI and
Texture: A case of Shaoxing city, Zhejiang Province
∗
ELNAZIR Ramadan
*
, FENG Xuezhi , ZHANG Youshui, ZHAO Shuhe
GIS&RS Laboratory
Department of Urban and Resources Sciences
Nanjing University, 210093
Nanjing, P.R. China.
*Corresponding author’s e-mail address: elnazeerramadan@yahoo.com
ABSTRACT The present study illustrates a method for monitoring the urban growth of Shaoxing
city, which is experiencing an obvious trend of urbanization, with the help of two different TM images
acquired on different dates. The methodological approach followed was by creating vegetation index
using NDVI equation and then texturing the images to create a Normalized Difference Textural Index
for the years 1984 and 2000. This index represented the ratio of changes in the resultant image. The
method is found to be successful in showing the areas of change between the two consecutive images.
However, appropriate ground data is always necessary to reach a reliable conclusion. The results have
revealed a notable change in the vegetation cover in and around the City premises. The key
assumption of this study is to investigate the reciprocal relationship between landscape ecology
symbolized by the presence of vegetative cover and the land use represented by urbanization in
support with NDVI.
Key Words: Remote sensing, urban growth, Shaoxing, NDVI
基于
基于基于
基于 NDVI 和纹理变化的城镇扩展检测
和纹理变化的城镇扩展检测和纹理变化的城镇扩展检测
和纹理变化的城镇扩展检测——
————
——以浙江省绍兴市为例
以浙江省绍兴市为例以浙江省绍兴市为例
以浙江省绍兴市为例
*
GIS 210093
TM
NDVI NDVI
1984 2000
1. INTRODUCTION
Global urbanization represents most cities which are in a continuous state of growth. The impinge of
urbanization on natural land cover in the terms of farmland and forest is very observable in many
world cities. Urbanization, the conversion of other types of land into the uses associated with growth
∗
——
of population and economy is the main type of land use and land cover in human history. In china land
use/land cover patterns have undergone a fundamental change due to accelerated economic
development under its economic reform policies since 1978. Urban growth has been speeded up
putting more stress on the local environment. This is particularly true in the eastern coastal region
such as Zhejiang province where some agricultural lands are disappearing each year, converting to
urban or related uses (Ji, 1999). Monitoring and evaluating urban change is thus a major issue in
urban planning and management effort through a sound sustainable use of natural resources. Remote
sensing is continuously offering a rapid and cost effective opportunity to study and monitor changes.
It has proved to be the most significant source for collecting multispectral, multispatial and
multitemporal data and turn them into information, valuable for understanding and monitoring urban
land processes and building urban land cover datasets. Information about changes in the landscape
provides valuable facts and figures on processes at work. Therefore it is not surprising that significant
research has been undertaken to develop methods of obtaining change information from remotely
sensed data (Dahl, 1990; Estes, 1992; Jensen, 1996). Many change detection algorithms are commonly
used, where as the selection of appropriate change detection method should be depends upon adequate
understanding of landscape features, imaging systems, and information extraction methodology
employed in relation to the aims of analysis.
Provided the biophysical characteristics of the study area, a method based on NDVI was selected.
Since, the urbanization in non-arid regions replaces most of the vegetation (high NDVI) with building
materials (low NDVI), the sudden decrease in NDVI should indicate urban development. This method
was selected because the generation of a new composite image and the numerical differences between
the two images of the same sensor at different time facilitates information about change detections.
The objective of this study is to quantify the growth based on remote sensing data (land cover change)
in Shaoxing city of Zhejinag province. The study scope is on the urban expansion with particular
attention given to arable land consumption. The study area covers Shaoxing city as well as its
surrounding area as a part of its eastern coastal region, which characterized by high urban growth due
to its geographic location and investment environment. It’s a kind of small and medium size cities
which become a base for light industry production particularly the textile.
2. MATERIALS AND METHODS
The study area is a subsect image of Shaoxing city comprised of two TM images for the years 1984
and 2000. A method based on image transform is developed to allow the generation of new composite
image capable of detecting changes. This method also adopted aim for estimating the urban growth by
measuring differences in NDVI between subsequent images by means of estimating the amount of
vegetation available on both images. At first, NDVI was calculated for the each image where as in the
case of Landsat TM, NDVI is defined as (band4-band3) / (band4+band3). NDVI output for each
image was then generated and afterwards, a texture algorithm was run on both images. The texture
operation clumped pixels in the image together according to the value of the nearest pixel, which acts
as smoothing filter. The algorithm was run on ERDAS imagine, 8.4 using a 3x3 contextual window
and 7x7 filter to compare the differences. Then the images were compared to sort out the best filter. It
was found that the 3x3 is better in showing the differences in texture owing to the size of study area.
The two textural images were then subtracted (2000-1984) and the resultant image being of a pixel
difference between the two textural images. Then again, arithmetic functions such as addition and
division were carried out for better comparisons and the reference is a classified image, which enables
to detect the changes. The subtraction operation was carried out on the pairs of co-registered images to
assess the changes taken place between the two images. The maximum negative difference is 0-255 =
-255 and the maximum positive difference is 255-0 = +255, if the value 255 is added to the difference
then the dynamic range is shifted to 0-510. Next divide this range by 2 to give 0-255. Scaling the
result of image substraction on to 0-255 pixel range, no change is recorded at 0 value and the change
increases towards 255, where the threshold is represented by a color. White color indicates the
changes while black represents no change
3. RESULTS AND DISCUSSION
The final result is a Normalized Difference Texture Index or a texture NDVI of both years. The
analysis of the image value would show the difference between vegetative and non-vegetative areas
on both images, which believed to be associated with urbanization. Owing to the limited field data, it
is difficult to confirm the textured image showing changes over the time but however, it can be an
indicative of the difference. The comparison made with the classified images showed that the areas in
the vicinity of Shaoxing city has undergone obvious transformations towards urbanization since the
area is shown as a white color in the divided texture image. The farming areas around the city also
shown as white in the divided texture image that matches with the changes in TM image, whereas the
green areas inside city are shown as gray to black in the divided texture image, indicating only a
meager change. The visual interpretation carried out on the original image when compared with the
resultant image to check the accuracy of the method, have shown an accuracy of 83%.
Many studies have proposed change detection techniques for monitoring urban growth based on
changes in Normalized Difference Vegetation Index (NDVI). Defined as normalized difference
between near-infrared and red reflectance, NDVI can be directly related to the amount of
photosynthetic (green) biomass within a pixel (Rouse et al., 1973, Tucher et al., 1981). Since the
urbanization in non-arid regions mostly replaces vegetation (high NDVI) with building materials (low
NDVI), the sudden decreased in NDVI clearly indicates the urban development. Howarth and Boassan
(1993) have found that the changes in vegetation indices strongly correlate with urban growth.
Nicoloyanni (1990) used a change-vector technique with axes for MSS NDVI and brightness to
produce an urban growth map for Athens, Greece for periods 1975-1981. NDVI variability was also
used as a basis for preliminary urban growth study of Washington DC by Johnson and Watters (1996),
with the help of 11-year records. Landsat TM observed that the Washington metropolitan area was
growing at a rate of less than 1% per year during that period.
Subtracted image Added image
Fig. 1 Raster grayscale outputs: Subtracted (2000-1984 NDVI images),
Added (2000+1984 NDVI images) and Divided (Subtracted/Added images)
Applying the NDVI change technique to southeast England, Griffths (1988) has found that NDVI
differencing alone tended to include area of agricultural change (e.g. crop rotation) in addition to the
real urban growth, causing extremely high errors of commission. And hence, he suggested that NDVI-
change map be filtered to remove unwanted agricultural noise. Using a proximity measure (in which
urban growth was preferentially identified if it occurred near other urban areas) as a filter, he managed
to improve the results although urban development was still overestimated by a factor or two.
4. CONCLUSIONS
During the present study, a methodology based on NDVI and texture was developed for investigating
the urban sprawl in Shaoxing city of Zhejiang Province. Results have revealed a notable change in the
vegetation cover based on NDVI. Since the key assumption of this study is to investigate the
reciprocal relationship between landscape ecology symbolized by the presence of vegetative cover and
land use represented by urbanization, the conclusion drawn is supporting that NDVI is quite good in
detecting changes. However, more ground data is required to reach a final conclusion to say whether
the decreased vegetation is an indicative of urbanization or not.
REFERENCES:
1. Dahl, T.W., 1990. Wetland losses in the United States1780’. U.S. Dept Interior, U.S Fish &Wildlife
Service, Washington, D.C., p. 21
2. Estes, J. E., 1992 Technology and Policy issues impact global monitoring. GIS world, 5 (10): 52-
55.
3. Gordon, S. I 1980, Utilizing Landsat imagery to monitor land use change: a case study in Ohio.
Remote Sensing for Environment, 9, 189-196.
4. Griffith, G., H 1988, Monitoring urban change from Landsat TM and SPOT satellite imagery by
image differencing. Proceedings of IGARSS ’88 Symposium Edinburgh”(Piscataway,
NJ:IEEE), pp 493- 497
Divided image
5. Howarth, P.Jand Boasson, E., 1983, Landsat digital enhancement for change detection in urban
environments. Remote Sensing of the Environment, 13, 149-160.
6. Jensen, R.J. (1996) Introductory Digital Processing: A Remote Sensing Perspective, Prentice Hall
Series in Geographic Information Science, Prentice Hall, Inc.
7. Ji, C.Y.,Liy,Q.,Sun,D and Wang, S (2001) Monitoring urban expansion with remote sensing in
China. International Journal of Remote Sensing.22 1440-1455
8. Nicoloyanni, E., 1990, a diachronic change index applied to two Landsat MSS images of Athens,
Greece. International Journal of Remote Sensing, 11, 1617-1623
9. Rouse, J.W.,Haas, R.S., Scheli., J.A., and Deering, D. W 1973 , Monitoring vegetation systems in
the great plain with ERTS. Proceedings, 3rd ERTS Symposium, 1, 48-62.
10. Tucher, C.J., Holben, B, N., Elgin, J., H and McMurtry, J. E 1981, Remote Sensing of total dry
matter accumulation in winter wheat. Remote Sensing of The environment, 11, 171-189.