MONITORING URBAN LAND COVER/LAND USE CHANGE IN ALGIERS CITY USING LANDSAT IMAGES (1987-2016)

Monitoring the Urban Land Cover / Land Use change detection is important as one of the main driving forces of environmental change because Urbanization is the biggest changes in form of Land, resulting in a decrease in cultivated areas. Using remote sensing ability to solve land resources problems. The purpose of this research is to map the urban areas at different times to monitor and predict possible urban changes, were studied the annual growth urban land during the last 29 years in Algiers City. Improving the productiveness of long-term training in land mapping, were have developed an approach by the following steps: 1) pre-processing for improvement of image characteristics; 2) extract training sample candidates based on the developed methods; and 3) Derive maps and analyzed of Algiers City on an annual basis from 1987 to 2016 using a Supervised Classifier Support Vector Machine (SVMs). Our result shows that the strategy of urban land followed in the region of Algiers City, developed areas mostly were extended to East, West, and South of Central Regions. The urban growth rate is linked with National Office of Statistics data. Future studies are required to understand the impact of urban rapid lands on social, economy and environmental sustainability, it will also close the gap in data of urbanism available, especially on the lack of reliable data, environmental and urban planning for each municipality in Algiers, develop experimental models to predict future land changes with statistically significant confidence. ∗ Corresponding author The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-2/W7, 2017 ISPRS Geospatial Week 2017, 18–22 September 2017, Wuhan, China This contribution has been peer-reviewed. https://doi.org/10.5194/isprs-archives-XLII-2-W7-1083-2017 | © Authors 2017. CC BY 4.0 License. 1083


INTRODUCTION
Urban change, in particular, is an important issue of Urban development and the environment (Roberts et al., 2016).The main purpose of this research is to follow the land cover land use LC/LU especially the Urban Land and its influence of decreasing the agriculture.In this perspective, the mapping of the region not only provides the value of scientific research information pertaining to a group but also provides general information about government planners and decision-makers (Triantakonstantis et al., 2015), particularly in Urban Land.Using the satellites TM/ETM+/OLI images dated for 1987-2016 years respectively.Covers most Urban Land with many spectral bands that can be used to determine the soil-related debilitating areas in urban (Bauer et al., 2004).This approach is based on the study of the urban tissue that develops faster with an annual growth rate, which oversees the evaluation of urban change detectors in each municipality and control the speed of expansion.The rapid increase in urbanization in Algiers.And the changes in agricultural caused by the extreme use of urban expansion (Siciliano., 2012), mainly due to population growth between 1987 to 2016.As urbanization progresses rapidly, it is attractive more important to know its spatial and temporal dynamics for urban environment improvement (Shahtahmassebi et al., 2014).
The massive expansion of the population (Campante et al., 2012) which must meet in the past through a variety of facilities that do not have sufficient time to meet quality standards and consistency and sustainability, which was on the roads of unfinished buildings that lead to nowhere.A relatively short period of time, urbanization has emerged as a major environmental problem facing many parts of the land (Montgomery., 2008).Algiers needs to be updated, which was developed from the same independence, the launch of the Strategic Plan and the protection of the environment (Hadjri et al., 2004).
In order to define urban dynamics, various methods have been developed, including pixel-based classification (Wang et al., 2004), subpixel classification (Mitraka et al., 2015) Oriented approach (Hussain et al., 2013).And mechanical learning algorithms (Chrysoulakis et al., 2015).Urban expansion is one of the typical models of land use and land cover (LULC) change, which was very important for city planning, environmental management, land integration (Yin et al., 2011).
Predictive modeling of Land-use change can be used as an unambiguous to achieve the path that a particular region can take in the light of factors (Wu et al., 2006), it is also a tool for managing the organization and space in order to maintain an often unstable balance, which in the end may lead, if not saved, to disrupt regional dynamics.Modern technology innovations have resulted in remote sensing and geographic information systems and space modelling for a wide range of tools to be used in research to develop reliable maps of land use and Monitor changes over time (Yang et al., 2002), taking into account the temporal dimension and finally predict future analysis (Aussenac., 2000).All of these techniques and using them collectively enable us to provide concrete and measurable responses to land-use changes and determine participation in these process factors.
The analysis of cartography and land-use projections in 2020 confirms (Tewolde., 2011) the extension of the process in fact and is particularly aggravated by the creation of new Urban Centers for a coherent development policy for the city, which can reduce the effects of this phenomenon in the medium and long term, which depends on the current dynamic assessment and forecasting future development (Marbà et al., 1996) in the region of Algiers .Two objectives are covered by this research 1) to assess the degree of change in land use based on the comparative analysis of the different satellite images available.We also identify the external factors that favor these spatial changes; 2) Evaluate the trajectory of the Algiers statistical offices in terms of land use for 29 years, which could show a further increase in the spread of urban sprawl to the detriment of agriculture.Entry into an unprecedented land crisis.

STUDY AREA AND DATA PREPROCESSING
Algiers, the Capital of the Algeria, situated at Mediterranean Seaside in the north of Africa, with geographic coordinates of 36°42'N and 3°09'E as shown in Fig. 1, The winters are mild and wet while the summers are hot and dry.The average annual temperature is 21•C.The rainfall is erratic from one year to another, with sometimes pronounced droughts.populated at three million inhabitants and occupied 1190 km 2 of the surface (Bouhennache et al., 2014).In the last decade, the urban spaces have rapidly been growing especially of east, west, and south of Algiers City decreasing the cultivated spaces and Bare Field land.This increase indicates that urban areas are on the base of Algiers Municipality, which is rapidly expanding in a certain City.The state of the commune is ideal for urban areas with the rapid development of the study.The Landsat image requires: (196/35 path/row) for our study to cover the Landsat area.Due to cloud cover in a few years, we had to choose only 15 Landsat images, and we had 06 Landsat (TM), 05 Landsat (ETM +) and 04 Landsat (OLI) shown in (Figure 2), covering the period from 1987 to 2016, Impervious surfaces from images Acquired in summer achieved the best results based on Spectral mixture analysis (Henits et al., 2017).All Landsat images are downloaded from US Geological Survey (USGS) Earth Explorer (https://earthexplorer.usgs.gov/)and The Landsat program has been extensively used for ecosystem monitoring (Wulder et al., 2012).And the spectral signatures of remotely sensed data, for medium spatial resolution images, are still the most important features in land use/cover classification.At the optical analysis base, all images had the correct geometric coordinate and provided level-one terrain-corrected Landsat data in WGS84 geodetic datum, Universal Transverse Mercator Map projection (UTM, Zone 31N), due to the nature of the data the geometric and radiometric distortion were previously corrected before distribution (Rawat et al., 2013).And the individual scenes were mosaicked according to the resampling method of the nearest neighbor.The mosaic images were an additional subset by the administrative boundaries of Algiers city.

Methods
Figure 3 illustrates the framework of our study, we developed an approach through the following steps: 1) Image preprocessing of the extraction of important properties.2) extract training sample candidates based on the developed methods; and 3) Derive maps and analyzed of Algiers City on an annual basis from 1987 to 2016 using a Supervised Classifier Support Vector Machine (SVMs) (Lin et al., 2011).
For the reason that our principal interests are to map an urban land trajectory, we identified four types of land covers, including Urban Land, Vegetation, Bare field, and Waterbody.the class of Urban land is defined as built environment with impervious surfaces (Li et al., 2015)

Classification
Support Vector Machine (SVM) classifier is a non-parametric supervised classification derived from statistical learning theory.SVM was developed in the late 1970s, but its popularity in Remote Sensing only began to increase about a decade ago (Zheng et al., 2015) Previous studies showed that SVM has the ability to generalize to unseen data with a small training dataset (Zheng et al., 2015) compared SVM to two other non-parametric classifiers.SVM has been familiar as an effective classifier in land-cover mapping (Clinton et al., 2015) We used the ENVI (the environment for visualizing images software), SVM classifier in this study and the parameter settings were as follows: the radial basis function was used as the kernel type, the gamma in the kernel function was set as 0.143, and the penalty parameter was set as 100.

Land Cover / Land Use Classification Result
On the basis of a time series of the Landsat images, LULCs of the study region was classified into four categories such as Urban land, Vegetation Bare field, Water bodies, the supervised classification method (SVMs) provides excellent results in all Landsat images as shown in Figure 4: Figure 5. Assessment of Land Cover Accuracy the SVM Classification.Two indices were used to evaluate the performance of the classifications: The Overall Accuracy (OA) and the Kappa index (K), the evaluation results demonstrated that the SVM algorithm with an overall accuracy and a kappa coefficient test was performed to measure the extent of classification accuracy as it not only accounts for diagonal elements but for all the elements in the confusion matrix has a higher accuracy in land use mapping (Srivastava et al., 2012).
On average, the overall accuracy is approximately 96.02% and a kappa coefficient 93.66%, demonstrating that the land-cover classification results provide a solid basis for quantifying urbanization activities in the fast-developing urban areas.

Classification and evaluation
The Classification of different class types with learning samples extracted using SVM (Fig 4).Based on an optical control, the results of the classification, and the reason for the spatial division of ground cover per year.
And it has increased in recent years of the study period of the project, and the urbanization of Algiers City clear that urban extension has increased substantially continuously from year to year, and urban areas throughout the area in 2015 was the almost year after the year 1987.
We tested the ability of Support Vector Machines (SVMs) to discriminate classes using a limited number of training samples and it was applied to time-series (Zheng et al., 2015).The empirical results of this study showed that the Urban Area in Algiers in 2014 was 44,812 Hectares, an average annual increase of 4% since 2000.The Urban Area in 2000 was 25,439 Hectares, the average annual rate of 4.7% since 2000 when Urban Areas were 13,903 Hectares to be able to assess and measure the decline of arable land depends on Urbanization and some pixels of urban land were misidentified as bare Field because of the similar spectral signatures of bright urban land and dry bare Field. .The relationship between Urban Land and Bare Field for Different Years.For this purpose, since 1987 to 2016, spatial images Landsat Satellites were used to evaluate and modify urbanization and transgressions during this period.The results showed that urban density was concentrated along a high length of Algiers and the City Centre.However, the density of vast urban areas beyond the border along the coast and southeast of the city, it shows us the city of Algiers had a growth rate.These expansions are largely related to new properties and these trends help us to better understand the patterns and processes of urbanization and suggest some problems with urbanization so far.For example, this increase in urbanization in areas negatively affects the ecosystems and services we provide, the urbanization of these agricultural lands and natural habitats in urban areas, thereby reducing the cultural identities of cities. Urban studies in the future should consider both generalities and characteristics and be directly related to urban sustainability measures.1987 to 2016 as derived from the Landsat data.The main change in land cover in urban areas is due to other types of land cover in land constructed as a result of construction and repairs, but other changes, including increasing public green space and Roads, are also very common.Remote sensing is arguably the most effective tool for monitoring land cover and Landsat TM / ETM + and OLI data are the most popular data sources on the urban scale.

Urban Sprawl on Provincial Level
In order to analyze changes in urban coverage over the past 29 years, post-classification comparisons and statistics are used.Based on the land cover properties in Algiers, we used the following four land cover types: Urban Land, Vegetation and Bare Field, Water bodies.Calculating the Urban Land of Algiers City, we applied the Supervised Classification of land cover in the Landsat Images containing the Cities in the sample typically extended beyond the set of districts containing these Cities.To calculate the Urban Land for every district in this set, the urban grid for every City in the sample needed to be clipped using the City districts Shapefile.Clipping the urban grid was done by using the Arc Map's Raster Calculator with the municipalities Shapefile set as an Analysis Mask.The total area for each Urban Land Cover / Land Use category in the study area were calculated separately from the images classified between (1987 -2016).This is shown in Table 2.   predefined land cover classes.In this way, changes in the areas extent can be calculated over time.

Urban Sprawl on Municipal Level
finally, series of acquired images with spatial and temporal resolution are available.these series make it possible to develop a monitoring of a great diversity of spatiotemporal structures and are particularly interesting for urban applications.

Figure 1 .
Figure 1.The Study Area of Algiers City in Northern Algeria.

Figure 2 .
Figure 2. The Temporal Distribution of the used Landsat TM / ETM+ / OLI Images.

Figure 3 .
Figure 3.The Overall Framework of the Study Processing and Analysis.

Figure 4 .
Figure 4. Land Cover Maps of Algiers from 1987 to 2016 Derived from Landsat Time Series Images.
Urban from 1987 to 2016 as Derived from the Landsat Data.In this study, Remote Sensing techniques have been used for data analysis for 29 years, and the grading method supervises the Support Vectors Machine (SVMs) and changes detection.1) Assessment of urban land changes and population growth; 2) to measure the hectares and percentages of Urban growth; and 3) temporary analysis of urban Land.This study could highlight Land-Use change for all Urban Areas of Algiers.He stressed the necessary statistics to quantify this phenomenon.The discovery of change was a critical contribution to the study.Mediumresolution images, such as Landsat, have enabled the extraction of urban areas, compared with the National Office of Statistics data.
Figure 7.The relationship between Urban Land and Bare Fieldfor Different Years.For this purpose, since 1987 to 2016, spatial images Landsat Satellites were used to evaluate and modify urbanization and transgressions during this period.The results showed that urban density was concentrated along a high length of Algiers and the City Centre.However, the density of vast urban areas beyond the border along the coast and southeast of the city, it shows us the city of Algiers had a growth rate.These expansions are largely related to new properties and these trends help us to better understand the patterns and processes of urbanization and suggest some problems with urbanization so far.For example, this increase in urbanization in areas negatively affects the ecosystems and services we provide, the urbanization of these agricultural lands and natural habitats in urban areas, thereby reducing the cultural identities of cities. Urban studies in the future should consider both generalities and characteristics and be directly related to urban sustainability measures.

Figure 8 .
Figure 8.The map urban expansion in Algiers City from1987 to 2016 as derived from the Landsat data.The main change in land cover in urban areas is due to other types of land cover in land constructed as a result of construction and repairs, but other changes, including increasing public green space and Roads, are also very common.Remote sensing is arguably the most effective tool for monitoring land cover and Landsat TM / ETM + and OLI data are the most popular data sources on the urban scale.

Figure 9 .
Figure 9. Urban Sprawl at the Municipal Level in three Phases.

FigureFigure 11 .
Figure 10.Urban Sprawl at the Municipal Level

Table 1 .
Spectral band specifications for Landsat sensors TM / ETM+ and OLI.

Table 2 .
Annual Change Urban Land with Population Density.
Our results show that dramatic urbanization has occurred in the region of the Algiers city, where Urban Land areas have mostly expanded to East, West, and South of the Central Regions.The average growth rate of urban areas in Algiers from 1987 to 2016 was at 707 Hectare per year, which was generally consistent with the National Office of Statistics data.