Human populations increase over time, which leads people to live in groups and communities. According to the development in those communities, some came to be recognized as rural settlements, while those with more progress were called the cities. The development process in the urban areas led more people to settle in those areas. This process is known as urbanization. This urbanization resulted in rapid population growth in the cities [5]. However, during this urbanization, a large number of people still lived and worked in rural settlements. It was the eighteen-century industrial revolution that attracted a large scale population towards urban areas. Approximately the world’s half population is moving towards cities. Since these areas provide comparatively more employment, better education, and increase recreation facilities. During industrialization, many countries experience this urbanization; nonetheless, those previously industrialized nations have already been through this process of urbanization. Thus, urbanization is a global phenomenon and as with other phenomena, urbanization also entails some severe disadvantages [6]. The creation of megacities is one of the essential aspects of urbanization. Megacities are cities with more than ten million populations. This, however, has now changed to cities with even double than ten million; for instance, the Japanese city Tokyo has a population of forty million. This has resulted in urban sprawl, which increases the city’s geographical area because of its population growth. In this process, the cities take up the areas which were formerly used for agricultural purposes. These suburbs of the cities now serve residential and industrial functions [7]. This sprawl calls for infrastructural development, especially for transportation, since people are likely to live far from the workplace and need daily commuting to and from one end of the city to another. LU/LC change related to urbanization produces consequences such as a rise in temperature and low air quality [8]. The population will continually grow in urban areas and their population. This will lead to further complex challenges of urbanization in the future. Planning to deal with the likely effects of this urbanization will have consequences for our world in the future.

Many factors contribute to causing global warming and climate change, but temperature is one of the main factors. A continuous rise may be found in the annual temperature worldwide in the past 150 years. The LU/LC change of a city also changes the land surface temperature (LST) profile of a city. Different land surface with different emissivity has different temperature [9,10]. LU/LC change in the present work has a remarkable contribution to changing the urban and local temperature. It has a substantial impact on the natural rounds of the atmosphere. By the year 2050, it is anticipated that more than half of the world’s population will move towards urban areas [11]. An obvious picture of the land surface and the climate’s physical properties can be declared by the LST, which is involved in the environment’s several processes. Measuring the air temperature by using temperature sensors is expensive and time-consuming as it requires the building of temperature sensors at the preferred places in the city.

Remote sensing is acquiring details about an object without any physical interaction with that object. Satellite sensors obtain this information about the object. Farmers can use remote sensing to observe their fields without physically visiting the area continuously. The farmers can monitor the overall image of their plantations, extracted through different classification tools. These satellite images give the farmers essential information to plan better for their fields [12,13]. The function of remote sensing is to provide data that helps analyze the variabilities of temporal and spatial parameters of the environment on a large scale. This data is utilized for crop yield appraisal, growth regulations, environmental damage evaluation, land-use monitoring, calculations, surrounding explorations, town planning, radiation monitoring, and geographical maps for civil and military purposes. Remote sensing’s most essential function is to classify and map the land cover aspects such as vegetation, soil, water bodies, and forests [14]. The Remote Sensing study includes photography, computer systems, spectroscopy, electronics satellite propelling, optical communication and media studies, etc. Remote sensing integrates these innovations in one complete system. The remote sensing process is achieved through several stages essential to its successful operation, as shown in Fig. 1.

Two main types of remote sensing are Passive and Active remote sensing. Passive sensors are used in passive remote sensing. These sensors utilize light to capture the image. These sensors are installed on satellites, aircraft, or drones; these sensors detect the natural radiation reflected from the objects or the earth’s area. Radiometers, infrared, and photography are some other examples of passive remote sensors. Fig. 2 shows the passive remote sensing process.

Active remote sensing uses active sensors, which respond to internal stimuli of objects. It records an object’s vitality. The energy is sent to objects or territories, sensors distinguish and the radiation is measured that reflects from objects. Examples of active remote sensing are LIDAR and RADAR [15]. Fig. 3 shows the active remote sensing process.

Image classification is the most significant field in processing research, particularly in pattern recognition. Image classification is a significant objective for ecological applications. Image pixels in classification can be characterized either by division/segmentation or by factual belongings of multivariable dependent on both spatial and measurable associations with adjoining pixels [16].

About urbanization, especially for climate systems, temperature is the most key parameter. The main objectives of the current study were;

The LU/LC classification of Peshawar region using supervised classifiers, i.e., artificial neural network (ANN) and support vector machine (SVM).

The proposed work comprises of 7 sections: In Section 2, the literature review has been discussed, Section 3 presents the materials and methods, the methodology has been discussed in Section 4 and Section 5 consists of the experiments and results which demonstrate the simulation results using ENVI & GIS tool, in Section 6 we compare our work with the previous studies and Section 7 presents the conclusion.

District Peshawar, 340 m above sea level and 1257 km², is located at 34°0′28.8″ N 71°34′42.6″ E in Khyber Pakhtunkhwa province. On three sides, the valley of Peshawar was surrounded by mountain ranges, with the fourth opening to Punjab province shown in Fig. 4. The oldest history dates back to 539 BCE, making Peshawar one of the oldest cities in Pakistan and throughout the world [20]. The sixth-largest city of Pakistan is Peshawar, according to the 2019 census, with a population of 2,133,000 (Tab. 1) (Pakistan Bureau of Statistics, 2019). According to the Koppen classification, Peshawar has a hot semi-arid climate, with mild winters and sizzling summers.

The free and open data sets (multispectral and multi-temporal Landsat satellite imagery) were collected from USGS (United States Geological Survey) and SUPARCO (Space and Upper Atmosphere Research Commission). The analysis was based on the two data sets, i.e., 1996 and 2019. According to the NLCD, the land cover classification has four main types, i.e., water bodies, agricultural land, settlements, and barren land. This land-use data were imitated from the 30-m resolution Landsat ETM+/OLI images. Tab. 2 shows the imagery data that were downloaded based on the satellite availability for the Peshawar region. The Peshawar region’s paths 151 and row are 36, which becomes 151/36, as shown in Tab. 2. By putting the raster form of the imagery data in GIS, the ROI was separated from the rest of the image using the subset image tool.

Pre-processing is the first step of classification and is carried out to remove noise from the image. For the proposed set-up, a reduction in atmospheric correction, fog, and band ratio discovery is made in the pre-processing phase. Fig. 6 shows the workflow of image pre-processing.

To remove noise and fog from images, another filter by the name of “convolution and morphology” is applied to the image. The objective of this filter is the evaluation of geometrical structures. For this filter, the relative ordering of pixels is essential rather than the pixels’ numerical value. Morphological filtering effectively works on digital images. Convolution and Morphology filter is available in ENVI 4.7 tool under the heading of filter tab. Click on the Filter tab in the menu and then on Convolution and Morphology. Figs. 7 and 8 show the convolution morphology tool and processing.

The median filter is applied to make the image smooth while conserving edges more significant than kernel dimensions. The central pixel value is replaced with the median value. Removing speckling noise from the image, the median filter produces better results. Fig. 9 shows the filtered image after the pre-processing and applying median filter. To sharpen an image, the median filter is provided by the Add Back value. To preserve an image’s spatial context, the Add Back value adds the original image’s back. Add back value is the percentage of the original image included in the final image output. If the Add Back value is 40%, at that point, 40% of the actual image is added to 60% of the final production. Add Back values, added to the median filter, are shown in Fig. 7. Click on Quick Apply or Apply button to add this data to the convolution and morphology tool. The input data is given to the C&M tool; the median filter is then used to acquire the desired filtered patch. A band list will appear once the convolution process is completed and we have to select the grayscale and RBG bands.

Selecting training samples is the second step of image classification. This sample is used for the automatic classification of a particular data set. This training sample is needed for each class and the number of required classes depends on research interest.

Region of Interest (ROI) are subsets of sample data. Region of Interest is selected as a polygon. ROI defines the border of an object under observation. Once the filtered image is opened, click on tools and then the region of interest. A dialogue box will pop up, showing ROI tools. Select the desired area by drawing a polygon. Several pixels selected will be highlighted as a sample shown in Fig. 10. The number of pixels and colors for the LU/LC classes are mentioned below in Tab. 3.

The decision step of image classification is based on the training sample where the system is tested for a decision. For comparison of image pattern with the desired image, we select the appropriate procedure. Once the sample is taken, image classification is carried out using a supervised classification algorithm.

There are different bands with combined features in unclassified images that cannot be distinguished. To turn these images into their features, supervised classifiers can be used [25], i.e., agriculture, barren land, water bodies, and urban area. Different band combinations were used to view the image to classify the various features. LU/LC classification can be achieved by classifying the satellite images. For each class, simple has been taken in polygon form with pixels mentioned in Tab. 3. Two supervised classifiers Artificial-Neural- Network (ANN) and Support-Vector-Machine (SVM), were used after testing. The classification algorithm is selected once the training sample is added. ANN and SVM are selected by clicking the classification menu and then supervised classification.

Visible and near-infrared electromagnetic spectrum bands are used to calculate the Normalized Difference Vegetation Index NDVI is the best tool to observe the presence of vegetation areas and their strength. The general Eq. (1) for deriving NDVI values is shown in Tab. 3:

(1) NDVI=(NIR−RED)/(NIR+RED)

The value of the NDVI index is between 1 and 1.

Fig. 12 shows the LU/LC classification of 1996 and 2019 imagery data using SVM. The study area’s imagery data were also classified into four classes, i.e., settlements, agriculture, barren land, and water bodies. 1257 km² was the total classified area. The kappa coefficient and overall accuracy of SVM are mentioned in Tab. 5.

Assessing the classification accuracy is the last step in supervised classification and is carried out after completing the classification phase. LU/LC maps will be produced with accuracy by using a confusion matrix. The LU/LC classification output of the ANN and SVM are shown in Tab. 6. The table shows the overall accuracy obtained from the post-classification in the confusion matrix: the user and producer accuracy and the kappa coefficient calculated using Eq. (2).

(2) K=N∑i=1nmi,i−∑i=1n(GiCi)N2−∑i=1n(GiCi)

where :

i is the number of classes.

Agriculture, Settlements, Water Bodies, and Barren land have an overall high accuracy in SVM. The table shows that ANN has an overall accuracy of 99.3542% and the Kappa coefficient of 0.9907, while the overall accuracy obtained from SVM is 99.8829%, and the Kappa coefficient of 0.9983. So the overall accuracy indicates that SVM is the highest in declaring higher accuracy for LU/LC classification. A comparative analysis of the Artificial Neural Network and Support Vector Machine state has been summarized in Tab. 6.