Development of Land Use and Groundwater …
225
(4) supervised classification, (5) land use/cover change detection. These are common
steps used in processing and analyzing the land use.
(1) Image pre-processing and layer stacking
Layer stacking is the process of combining separated bands to form a single
multispectral image file for further analysis. In this study, it was applied using ENVI
5.1 software. This process was followed by defining the wavelength for all bands
according to the electromagnetic spectrum of each Landsat sensor.
(2) Image sub-setting
After the investigation of the images, it was found that the data set covers not
only the study area but also a large part of the Western Desert. Therefore, ENVI 5.1
software was used to subset the images by the vector of Rashda boundary to include
only the study area. This process decreases the amount of digital data in order to
speed up processing which is important when dealing with multiband data.
(3) Image post processing
Digital image post processing aims to obtain more information from satellite
images, which is difficult to get from the raw data.
(4) Supervised classification
Supervised classification was applied using ENVI 5.1 software. Supervised classification was used to generate land use and land cover classes using Support Vector
Machine (SVM) classifier. Comparative analysis clearly revealed that substantially
higher overall classification accuracy (95%) was observed with the object-based
SVM, compared with that of traditional pixel-based classification (89%) (Devadas
et al. 2012).
(5) Land use/cover change detection
Land use of Rashda was classified to five classes, that are (1) built-up area, (2)
agricultural area (including fallow agricultural land detected by visual interpretation),
(3) barren land, (4) sand dunes, and (5) water bodies (drainage ponds/lakes).
Due to the low level resolution, the classification of land use/cover using 1968
Corona image was difficult. For the areas that were difficult to distinguish either
they were barren land or agricultural area, they were assigned as agricultural area.
Therefore, it should be kept in mind that 1968 land use/cover detection tends to
overestimate the agricultural area.
The post-classification change detection analysis describes and quantifies differences between images of the same scene at different times. The classified images
were used to calculate the area of different land use/cover and to observe the changes
between different years which in this case are 1968, 1988, 2003 and 2018.
This analysis is very much useful to identify various changes occurring in different
classes of land use as shown by Hegazy and Kaloop (2015). Post-classification
comparison change detection was done after classifying the rectified images separately. The classified images were exported to the ArcGIS 10.4 software for
vectorization, calculation and comparison of the areas among the different dates.
225
(4) supervised classification, (5) land use/cover change detection. These are common
steps used in processing and analyzing the land use.
(1) Image pre-processing and layer stacking
Layer stacking is the process of combining separated bands to form a single
multispectral image file for further analysis. In this study, it was applied using ENVI
5.1 software. This process was followed by defining the wavelength for all bands
according to the electromagnetic spectrum of each Landsat sensor.
(2) Image sub-setting
After the investigation of the images, it was found that the data set covers not
only the study area but also a large part of the Western Desert. Therefore, ENVI 5.1
software was used to subset the images by the vector of Rashda boundary to include
only the study area. This process decreases the amount of digital data in order to
speed up processing which is important when dealing with multiband data.
(3) Image post processing
Digital image post processing aims to obtain more information from satellite
images, which is difficult to get from the raw data.
(4) Supervised classification
Supervised classification was applied using ENVI 5.1 software. Supervised classification was used to generate land use and land cover classes using Support Vector
Machine (SVM) classifier. Comparative analysis clearly revealed that substantially
higher overall classification accuracy (95%) was observed with the object-based
SVM, compared with that of traditional pixel-based classification (89%) (Devadas
et al. 2012).
(5) Land use/cover change detection
Land use of Rashda was classified to five classes, that are (1) built-up area, (2)
agricultural area (including fallow agricultural land detected by visual interpretation),
(3) barren land, (4) sand dunes, and (5) water bodies (drainage ponds/lakes).
Due to the low level resolution, the classification of land use/cover using 1968
Corona image was difficult. For the areas that were difficult to distinguish either
they were barren land or agricultural area, they were assigned as agricultural area.
Therefore, it should be kept in mind that 1968 land use/cover detection tends to
overestimate the agricultural area.
The post-classification change detection analysis describes and quantifies differences between images of the same scene at different times. The classified images
were used to calculate the area of different land use/cover and to observe the changes
between different years which in this case are 1968, 1988, 2003 and 2018.
This analysis is very much useful to identify various changes occurring in different
classes of land use as shown by Hegazy and Kaloop (2015). Post-classification
comparison change detection was done after classifying the rectified images separately. The classified images were exported to the ArcGIS 10.4 software for
vectorization, calculation and comparison of the areas among the different dates.