الفهرس 
Chapter Two:Litrature ReviewOnly 14 pages are availabe for public view
 |  | from | 97 |  |  |
| | | from | 97 | | |
Abstract
With various living and non-living components, water bodies constitute
one of the most intricate ecosystems. Coastal accretion, erosion, pollution,
sediment transport, and coastal development are all examples of dynamic
natural processes that affect water bodies. As a result, monitoring water bodies
is a vital duty in environmental protection and project development worldwide.
In this regard, water body extraction and determination are critical. In the past,
detailed field measurements were used to determine the location of water
bodies on the ground. However, such a procedure takes time and effort, and it
is also inefficient in terms of cost. Aside from that, human errors are
unsystematic, making them difficult to correct. Nowadays,
modern technologies like remote sensing are necessary. The large ground
coverage, high geometric resolution, multiple spectral information, and the
ability to acquire the scene in several spectral bands are all advantages of using
remote sensing.
Based on the relationship of the band reflectance values, the current
thesis proposes a novel method to accurately extract water bodies in Egypt
from multispectral Landsat-8 satellite imageries. The proposed technique is
based on studying and analyzing previously introduced shoreline extraction
models, which are based on connections between band reflectance values, and
combining the models that have been accepted. The Near Infra-Red (NIR) and
the Short-Wave Infra-Red (SWIR) bands are chosen to properly identify water
from non-water bodies on a satellite image. A study area along the northern
part of the Suez Gulf, Egypt has been utilized in processing and analyzing the
satellite images and applying the new approach.
The perpendicular distances between the results of each model and the
true boundary (ground truth) are measured at points along the boundary that are
1 km apart in the Digital Shoreline Analysis System (DSAS) module. This
distance is calculated as the Net Shoreline Movement (NSM) in DSAS, but it is used to evaluate the model results here. Based on visual interpretation
capabilities, the true boundary is extracted using an onscreen digitizing process.
The mean and RMSE (root mean square error) are two statistical measures
computed from these NSMs to represent the distance between the two
shorelines (the first deduced from the model and the second digitized).
There are 16 models of band ratios that can be used to distinguish between
water and other classes. It has been found that ten of the 16 band ratio models
have an accuracy of less than 1/2 a pixel, which is less than 15 m. The ten
models produced different results but based on the results of the ANOVA test,
the mean and standard deviation of each of the ten models have no significant
difference with 0.05 significance level. Since the ten models considered equal,
the combination of the results of the ten models are conducted, where the
summation of the results of the models is conducted. The final combination
scenario is selected to the one where at least seven models’ results were agreed.
Moreover, another spatial region has been chosen for validating the
results. Also, the accomplished findings show that the study area’s accuracy
after the combination is 11.831 m, whereas the validation area’s accuracy is
13.074 m. Based on the achieved results, it is recommended that this procedure
should be used to extract water bodies. It is also recommended to use this
procedure in the Mediterranean Sea and the Nile River, beside the Red Sea, for
accurate extraction and investigations of water bodies in Egypt.