Industrialization The urban growth rate has risen up

Industrialization and development of technology have
caused rapid changes to land and contributed towards urbanization. Villages and
small cities are being converted to urban areas and into towns frequently these
days. Similarly, over the last few decades developed areas and urban sites in
cities and towns have been expanding in Bangladesh at a significant rate. Establishment
of industries and other landmarks have contributed significantly towards this
massive outburst and expansion of developed areas. To add to that there is also
better living standard and easier lifestyle in the cities. Therefore
urbanization is happening continuously. At the heart of this expansion in Bangladesh,
lies the city of Dhaka. Most of the outward expansion in terms of development
have happened centering Dhaka. The urban growth rate has risen up massively
since the last 15-20 years.

Bangladesh has seen a massive population outburst in the
last few decades. Dhaka is at the core of this rapid rise in population. Currently
about 2 billion people live in the city of Dhaka. To cope up with this
increasing number of population various other changes has been happening. To accommodate
a population this big the infrastructure and land use has also seen rapid
development. Social development, better financial chances and better lifestyle
have been the most intensive and leading factor towards change in land use. The
fast development has also caused changes in all the domains for this particular
town. Number of markets, industries, airports and different other establishments
have increased in Dhaka which has attracted even more people in this town and
as a result changes in land use is happening everyday by leaps and bounds. All
these influential factors have made rapid changes to Dhaka in geo level.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

In this context remote sensing imagery and concepts can
come in handy. Terabytes of satellite image data are generated everyday with
all the satellites that have been launched to monitor the earth from different
criterion. All these images illustrate the condition of built up areas,
vegetation, water bodies and of various other natural and man-made entities.
Using satellite based images for land cover analysis and classification provide
promising results and enables to perform detailed study on areas that otherwise,
would not have been possible.

To perform the analysis based on the satellite image a reliable
work flow and model had to be designed. This workflow consisted of multiple
measures and decisions based on which the information had to be obtained from
the image. This extraction of information from the data can sometimes be
difficult and time consuming. Therefore, the goal was to design a general pattern
for analysis of the images which can later on provide important details in
terms of the changes and development of the land use.

Thus. The objective of this study is to study and
interpret the satellite images of the study area. By performing this analysis
the various measures for different types of lands and can be generated. This
approach uses various remote sensing techniques and algorithms. The interpretation
was performed for various years spanning over 3 decades. This allows to have a
clear picture of the changes that are occurring and also provides a more
accurate representation. Finally, it generates a model representing the growth
patterns and the most densely urbanized areas for Dhaka.

The study area chosen for this study is the Dhaka district.
It consists of 6 smaller cities. They are Dhamrai, Savar, Dhaka, Keraniganj,
Nawabganj and Dohar. Among which Dhaka, Savar and a few places in Keraniganj
has experienced visible development and urbanization in recent times. Which was
not the case for the other 3 cities. Dhaka city is the capital of Bangladesh.
It is the largest city of Bangladesh and accommodate the lion’s share of this
huge population the country. After the 1990s the urbanization and industrialization
has begun enormously in this city. The rapid development of infrastructures and
multinational companies led to the overall development of the town.

All these factors played the role of catalysts for
attracting people from all over the country to migrate to Dhaka. All the socio
and economic factors played a key role too. As people started migrating to
Dhaka the land use change has been significant. The reduction of trees and
plantation is pretty much visible with naked eye. High raised buildings and skyscrapers
are a common scene now on every block of the city whereas there had been a
scarce number of high raised buildings only 20 years ago.

Besides construction of houses for this large number of
people the city needed to provide more land to the countless industries and mills
that have been established lately. They are more often located at the outskirts
of the city, near the rivers where it’s easier to commute. As a result, the
rivers have been overused and in some cases they are filled up. Not only the
rivers, many water bodies, such as small lakes and ponds are frequently filled
up to build houses and markets. Many trees have been cut down to build roads
for the rising number of cars in this town.

All these factors are influential for conducting a study
for analyzing the patterns of land use for the district of Dhaka. Another
objective is to monitor and figure out the changes across multiple decades and
where the changes have been maximum.

Remote sensing is commonly known to be the field that
deals with obtaining information about the surface of the Earth from a distant
place. The process is accomplished by sensing and recording reflected and
emitted energy. After that, that information can be processed analyzed and
applied for further use.

Mostly in remote sensing, the process involves interaction
between incident radiation and targets of interest. Remote sensing also uses
non imaging sensors and involves sensing emitted energy through different
techniques. The steps involving remote sensing can be broken down into different
phases.

The first requirement for remote sensing is to have a
constant energy emitting source. In our case, it’s the sun. Sunlight emitted
from the sun directly falls on the lands and water bodies. The latter phases
work after based on the reactance of the objects to this energy coming from the
sun.

As the energy from the sun travels to reach earth it has
to travel through the atmosphere and the different levels of it. As it goes
through it it, it interacts and radiates energy through it. The same process is
repeated while it returns after being reflected from the target object from
earth before reaching the satellite.

After reaching the area of interest the energy interacts
with the target object. After that the target object reacts to it, or emits
energy as well. This varies from material to material and depends on the
properties of the object. For example, soil and water don’t react the same way
to the radiation.

After the object reacts the procedure is reversed again. This
time the signal travels through the atmosphere whence it came through. To capture
this energy or radiation there are sensors in the satellite placed to capture
these signals. These signals are then later collected in the satellite and
converted to electrical form so that they can passed on to a different place to
perform analysis. Hence, they are sent to ground stations or to earth where the
can be processed and interpreted

In a nutshell, the light emitted from the sun falls on the
land and other infrastructures that are present. Based on their own
characteristics and properties they have their own reflectance and absorption patterns.
This is captured from the remote satellites. These imagery are used as raw
elements for performing our study.

The solar radiation that any object can reflect is usually
dependent on wavelength and the properties of the observed object. This enables
us to differentiate various land cover types based on their response values for
a specific wavelength. Therefore it can be a means of identifying a particular
land use pattern. However, this spectral signature varies for the same object
or same geolocation time to time during a year. It happens because at different
times along the year the land can have vegetation that are about to grow, or it
can have fully grown ones. In short, it may vary depending on the season which
can give an illusion that the land use pattern has changed significantly. Various
other feature that contribute towards the particular characteristic of soil are
moisture amount, structure and color. This specific spectral characteristics
help to classify the objects based on their reactance and behavior.

For vegetation, this is somewhat a little different. They
can easily be identified among all the other object as they possess a unique
character. Their spectral response pattern can be used to characterize and
identify green vegetation and crops from other surface features. Green
vegetation are solely dependent on photosynthesis for their food production.
And photosynthesis needs sunlight of visible wavelength. Therefore the chlorophyll
from leaves of the plant tend to absorb light of visible wavelength. Therefore
to plants can easily be distinguished my analyzing the dominant reflectance towards
near infrared as plants don’t absorb much of it. On the other hand it will not
show much reflectance towards visible wavelength of light. Light from near infrared
emitted from the source will be mostly reflected after being emitted on the
plants. This rule applies to both crops and other vegetation or trees. But
there is a small distinction in it as well. For crops the image is mostly solid
and uniform all over. Whereas, non-vegetation and trees display a noisy and
textured response.

Water shows varying response value based on various
properties such as suspended sediments or dissolved matters. Turbidity and water
depth also play significant role here. Clear water offers minimal response
values. Whereas for turbid water and closed water the response value varies
according to the level of turbidity and sediments.

All these factors were taken into account for performing the
spectral analysis. Using the different spectral response of various surface features
they were identified and classified. Images of varying time spanning over 3
decades were chosen to accomplish this initial task.

For conducting this study the software used was Erdas
Imagine. It is a professional software of performing analysis of remote sensing
with ablities of raster and vector image study and also modeling tool
geo-spatial application. It mainly aims to process geospatial data and allows
the user to preprocess, enhance and process the satellite images for further
study in GIS or any other spatial analysis sector. It enables the user to
perform clustering. It provides satellite image classifcation techniques, both
supervised and unsupervised which is a tedious task when it comes to
classifying features from remote sensing satellite images. It also enables the
user to combine different categories of geospatial data and provides unique
facilities to develop operational programs for performing numerical, logical
operations through model or functions. In a nutshell Erdas Image to perform
various different operations on images and obtains various characteristics and
profiles of a specific geographical location.

For conducting this study images had to be obtained across
different time frames. This would allow to detect the changes that occurred
within that timeframe. The timeframe shouldn’t be too large on the other hand
picking a small gap between two images might not extract useful information
after analysis. Therefore our study used 4 images after the 2000s. They were
from 2000, 2010, 2014 and 2017.

 

Classification is a methodology of analysis in which
decision for identification made with the help of the image which represents
data of the region of interest. It is achieved by forming groups of homogeneous
region across the image. This homogeneous regions are called classes and each
of this sub groups signify different land cover types based on their
characteristics. Classification can be of two types, supervised classification
and unsupervised classification. Unsupervised classification method has a
common intent to expose the major land cover classes that is present in the
image without any knowledge beforehand of the images actually represent.

This form of analysis which is also used in our study
falls under domain of cluster analysis. They are called so for they try to look
for cluster of sub regions of pixels in the image which show similar reaction
and response characteristics in a multispectral image. Clustering also provides
a general sense of the land use pattern since it tries to analyze the entire
image and find any repetitive pattern that is common between neighboring
groups. It tries to identify the clusters of data from the image, calculates
the mean with respect to every image channel. After that each and every pixel
is assigned to one specific cluster on the basis of the minimum distance to
mean rule in.

The clustering process starts by assigning first an
arbitrary initial cluster vector. Next it assigns all the other pixels to the
cluster that is the nearest from that particular point. Afterwards, the new
mean vectors for all the clusters are calculated based on all the pixels in one
cluster. These last two steps are repeated a number of times before the
variation has reached a minimal threshold. This specific cutoff in error or
minimum threshold can be calculated by measuring the change of the distances of
the mean cluster vectors from one iteration to another. On the other hand it
can be calculated using the proportion of pixels that have changed with
different iterations. The basic fundamental target of K Means clustering is to
maximize inter cluster distance means and to minimize intra cluster distance
means. The objective is to minimize a specific cost function which calculates
and provides the approximate error in the clustering. It is called the sums of
squares distances and is calculated using the squared distances between each
pixel and its assigned cluster center. The goal of a good classification is to
minimize this MSE function.

To conduct this study, based on the geological surface and
land use habits, 5 categories were chosen to be the target classes. They are
water bodies, crops, other vegetation, built up and others. This choice of
categories were made by keeping the surface property and land type of Dhaka in
mind. There are numerous lakes, bogs, marshes and ponds all over Dhaka.
Although they are not much visible inside the city nowadays on account of huge
built up and building constructions, they are visible throughout the other
cities and also at the outskirts of the main Dhaka city. All these fall under
this category. There are many rivers that run around the Dhaka district like
the Buriganga River, Turag River etc. These rivers fall under this class as
well. Using the turbidity, depth and movement pattern knowledge derived from
the satellite images they are classified under water bodies.

Bangladesh is general is a fertile land. It has crop
fields all over its region. Dhaka is no different. Although in recent times due
to the large number of migration inside the Dhaka city the amount of crop lands
and vegetation fields has reduced significantly yet, they are very much visible
in all the other cities. These crops are mainly rice fields, wheat fields,
potato and pulse cultivation lands and many more. Huge acres of lands are
invested and bread in order to grow these products. These huge patches of lands
are the most convenient to classify as they show their special reflectance
characteristic in the satellite images. As crops attract light of visible range
and absorb it for photosynthesis it doesn’t help much while analysis. Their
high reflectance towards light of near infrared range is used to differentiate
them from rest of the surface objects. They produce a relatively smooth and
strong pattern in the satellite images depending on the spectral band. The
other category chosen for our study is other vegetation. All the other non-crops,
trees, patches of grassland, orchards and gardens fall under this. They
somewhat display similar spectral characteristics like the crops. But the
important different is that their response to multispectral signal is not quite
strong like the crops and they display textured properties while for crops it’s
relatively smooth and strong. Not only are the signals weaker than crops but
also they span for smaller regions than crops. For crops it can be across acres
of lands that we might find the same signal level. Whereas, for non-crop areas,
as they don’t span for longer areas the regions are more distributed and not
continuous for a large section of area.

The forth category is the built up areas. This is the most
important criteria of analysis in this study. The goal was to identify all the buildings,
cottages, human settlements and other developed sites from the satellite
images. Concrete and asphalt has their specific level of reactance and
reflectance level to multispectral signal. They exhibit different
characteristics than water or crops. This was used while classifying them. The
urban areas are not uniformly distributed across the district. They rather are
placed as clusters throughout Dhaka. Mostly the settlements are heavier in
Dhaka city, Savar and partially in some parts of Keraniganj. While classifying
the study did not treat bare soil lands, sand areas or riverbanks as urban or
built up areas. For smaller cases like these they were put into the 5th and
final category which is named others. It is named such because it is of trivial
proportion compared to all the other classes and of insignificant importance
for our study.