AN ASSESSMENT OF SPATIAL VARIATION OF LAND SURFACE CHARACTERISTICS OF MINNA, NIGER STATE NIGERIA FOR SUSTAINABLE URBANIZATION USING GEOSPATIAL TECHNIQUES

Rapid urbanization rates impact significantly on the nature of Land Cover patterns of the environment, which has been evident in the depletion of vegetal reserves and in general modifying the human climatic systems (Henderson, et al., 2017; Kumar, Masago, Mishra, & Fukushi, 2018; Luo and Lau, 2017). This study explores remote sensing classification technique and other auxiliary data to determine LULCC for a period of 50 years (1967-2016). The LULCC types identified were quantitatively evaluated using the change detection approach from results of maximum likelihood classification algorithm in GIS. Accuracy assessment results were evaluated and found to be between 56 to 98 percent of the LULC classification. The change detection analysis revealed change in the LULC types in Minna from 1976 to 2016. Built-up area increases from 74.82ha in 1976 to 116.58ha in 2016. Farmlands increased from 2.23 ha to 46.45ha and bared surface increases from 120.00ha to 161.31ha between 1976 to 2016 resulting to decline in vegetation, water body, and wetlands. The Decade of rapid urbanization was found to coincide with the period of increased Public Private Partnership Agreement (PPPA). Increase in farmlands was due to the adoption of urban agriculture which has influence on food security and the environmental sustainability. The observed increase in built up areas, farmlands and bare surfaces has substantially led to reduction in vegetation and water bodies. The oscillatory nature of water bodies LULCC which was not particularly consistent with the rates of urbanization also suggests that beyond the urbanization process, other factors may influence the LULCC of water bodies in urban settlements.


Introduction
The rapid rate of population growth has triggered the need for settlements that provide for shelter, human socioeconomic activities and food supplies that meet human need. The resultant activities on the earth's surface grow over time as well as create a parasitic relationship with other land uses. The increase in population growth and human settlements is one of the major constructs that reshape the nature of land use cover system of the earth surface with one growing at the expense of another. This relationship according to Coale and Hoover (2015) do not only affect planning and societal development but also influence energy consumption (S. Wang, Ma, & Zhao, 2014), pollution (Farhani and Ozturk, 2015;Liddle, 2014;Q. Wang, Zeng, & Wu, 2016) and global warming and climate change scenarios (Elmhagen et al., 2015;Garschagen and Romero-Lankao, 2015;Henderson, Storeygard, & Deichmann, 2017).
Population growth on the other hand has resulted in a rapid rise in urbanization in both developed and developing parts of the world. The menace of this rapid urbanization rate impact not only on human comfort, but also exert significant influences on the nature of Land Cover Change (LCC) patterns of the environment, in addition to pressures exerted on other components of the human environment including the depletion of available (Feng, Chen, Hayat, Alsaedi, & Ahmad, 2017;Guo and Shen, 2015;Jiang, Wu, Liu, & Deng, 2014;Koop and van Leeuwen, 2015;Prosdocimi, Kjeldsen, & Miller, 2015), depletion of vegetal resources Price and Bradstock, 2014;Zhou, Zhao, Zhang, & Liu, 2016) and modification of human climatic systems (Henderson, et al., 2017;Kumar, Masago, Mishra, & Fukushi, 2018;Luo and Lau, 2017;Mathew, Chaudhary, Gupta, Khandelwal, & Kaul, 2015).
The urbanization process results in Land Use Land Cover Change (LULCC) of a given location. The observed LULCC is due to; the influence of human activities on land cover through various activities such as farming, deforestation, construction and sand filling of former colonies of water bodies . The modification of the land cover types arises from the quest to meet man's requirements for survival in urban centers. While these demands required in urban centers increase proportionally to the rate of urbanization especially construction, the vital cover of the earth surface that is being modified frequently exerts a potency of seeding a feedback action to the human environment. This feedback either negative or otherwise, underscore the needs to critically study the effect of urbanization LULCC.
In light of the foregoing, Many researchers explore the potential of aerial photography (Fensham and Fairfax, 2002;López, Bocco, Mendoza, & Duhau, 2001;Mas et al., 2004) and remote sensing (Green, Kempka, & Lackey, 1994;Joshi et al., 2016;Lillesand, Kiefer, & Chipman, 2014) data in studying the effect of urbanization on vegetal cover change with surface vulnerability (Mukherjee, Krishna, & Patel, 2018;Tayyebi, Shafizadeh-Moghadam, & Tayyebi, 2018;Wellmann et al., 2018) over a long period, but are often deficient in terms of data consistency. The application of Remote Sensing (RS) data as well as RS techniques of data mining provide a frontier through which events of the past can be compared to what is obtainable today. Changes over a given period of time are capable of revealing great a deal of information regarding nature, and magnitude of change that may occur over time. The strength of remote sensing as a tool for data collection is based on it's capabilities to store information that has occurred. Thus, providing a data bank that enables continues tracking of changes that might exist when comparisons are made with present day data.
Geographical Information System (GIS) to provides a tool that allows for the integration of different but spatially referenced data about a phenomenon under study with a high degree of accuracy and appealing results. GIS is a cost effective-tools through which geospatial information related to changes on the earth surface can effectively be managed and integrated towards determining cause and effects. In addition, GIS provides an effective technique for the integration of multi layer information that might have direct or indirect impact regarding scenario occurrences (Ferreira et al., 2015;Hughes et al., 2016;Kaliraj, Chandrasekar, & Magesh, 2015 Shrivastava and Nag (2017); Shuaibu and Sulaiman (2012). These studies successfully implement RS techniques and other auxiliary data in GIS environment by treating each of the data sets as a separate geospatial data either as a point, polygon, line or linear feature. The processing of these data sets often employ different tools available in the commercial geospatial processing applications.
The evolution computer technology resulting in the availability of effective data storage, retrieval and manipulation along with large storage capacity of data has aided the effective integration of RS technology with GIS. Exploration in the utilization of this advancement has resulted in effective integration of RS data with GIS for effective decision making. The utilization of remote sensing data offers a useful insight about changes that reshape the earth cover over time, this is often better utilized if, and these changes can be tracked at regular intervals to provide knowledge regarding how rapid these changes take place. In an attempt towards the effective capturing of slight changes on land cover, the RS techniques becomes an inadequate tools since multiple layer information will be required.
Rather the use of numerical approach through the utilization of RS data to generate series of algorithm and mathematical expression may be explored. Studies that have combined the use of RS data and development of mathematical algorithm includes; Metternicht (2001); Nemmour and Chibani (2006); Oguz and Zengin (2011);Willhauck, Schneider, De Kok, & Ammer (2000) among several others. These researches had proven effective in the aspect of data integration and predictive capabilities, however; spatial reference becomes inadequate especially in mathematical modeling. In fussy logic, arguments and string conditions become too many to reflect reality thus, the need to develop a more geospatial application that can effectively manage multi layer information in determining land cover changes.
The integration of these techniques according to Corner, Dewan, & Chakma (2014);Dewan and Yamaguchi (2009), along with the pre-and post-classification techniques has extensively been explored with interest on pre-change vector as well as multi date classification. Other studies in an attempt to achieve the same goal of analysis, adopted Normalized Difference Vegetation Index (NDVI) (Rawat and Kumar, 2015;Yengoh, Dent, Olsson, Tengberg, & Tucker III, 2015;Zhu et al., 2016) and principle component analysis (Dronova, Gong, Wang, & Zhong, 2015;Rokni, Ahmad, Solaimani, & Hazini, 2015). The premise of these techniques is that change in the pixels as a result of change in spectral reflectance value for the period under study can be infrared but however, they are deficient in the effective identification of change nature (Dewan and Yamaguchi, 2009). To effectively manage the pre and post classification comparison techniques along with methods, (though may often create difficulties) have emerged as the most effective method for identification of LULCC studies.
The pre and post classification technique is most effective in urban environments where similarities in spectral response patterns in different Land Cover (LC) features may appear highly similar. In addition, the technique can be employed using RS data from sensors with different spatio-temporal spectral resolution (Dewan and Yamaguchi, 2009). RS has offered an effective tool for relating the interaction between population, environmental changes and human environment (Dewan and Yamaguchi, 2009;Miller and Small, 2003;Tuholske, Tane, López-Carr, Roberts, & Cassels, 2017). Space-born satellites provide useful information that enables the evaluation of a such scenarios over time.
Although, RS data might sometimes be vulnerable to interference by factors such as cloud and other aerosols affecting visual evaluation (Weitkamp, 2006;Winker et al., 2009), information generated from spaced based instruments is cost effective, durable, relatively accurate and reliable compared to the conventional methods of survey (M. Chen, Mao, & Liu, 2014;Congalton and Green, 2008;Hyyppä et al., 2000). Moreover, the challenges of vulnerability to interference by cloud cover and other atmospheric contents can effectively be managed using different image correction technique and additional ground reference data.
RS-based data sources and techniques provide a method through which data related to structural variation in LULCC pattern can be understood to mitigate irreversible changes that might occur as a result of LC alteration. For effective change pattern estimation, rate, types of LC change and prediction of feature magnitude changes, RS technique becomes a viable tool.
In addition for every sustainable development, data related to structural changes in LC is critical for policy formulation, implementation and effective conservations. indicates the lack of land use pattern categories in the state and might translate to government planning to be based on population data without reference to land use patterns within the metropolis. Numerous factors identified by Dewan and Yamaguchi (2009) to be responsible for the lack of this vital information to include; lack of geospatial expertise in government agencies saddled with responsibility of developing such a vital information, bureaucracy, financial constrain and absence of coordination among the relevant agencies. This research therefore, attempts to study the LULC change pattern of Minna, Niger State using geospatial techniques, remotely sensed data and socioeconomic data to provide empirical studies that will guide policy makers regarding patterns of land used changes in the study area.
Different techniques of remote sensing data analysis exist; ranging from image identification and classification based on spectral characteristics to object based classification schemes. The later classification systems have been applied in several remote sensing image analysis researches with wide variability of success. The spectral based classification techniques employed the use of image pixel where features of the same spectral pattern are grouped into one class representing a particular land cover type (Aplin and Smith, 2008;Blaschke et al., 2014;Hussain, Chen, Cheng, Wei, & Stanley, 2013). Although this approach has tested most effective but it still suffers limitations when pixels of an image containing two or more spectral properties are to be identified and classified, especially when the pixel's spatial extend varies from land cover extent of interest (Myint, Gober, Brazel, Grossman-Clarke, & Weng, 2011;Nogueira, Penatti, & dos Santos, 2017;Teodoro, Gutierres, Gomes, & Rocha, 2018;L. Wang, Sousa, & Gong, 2004;Whiteside, Boggs, & Maier, 2011;Yu et al., 2006). The implication of spectral variation within one pixel often results to misclassification and conflict of information. In the light of these shortfalls of the pixel based image classification technique, an object based classification scheme was developed to enhance accuracy and as an improvement over the conventional approach. Various approaches to object based oriented classification techniques exist (J. Chen et al., 2015;Cheng et al., 2015;Cheng, Han, Zhou, & Guo, 2014;Gokturk et al., 2015;Li, Zang, Zhang, Li, & Wu, 2014).
These techniques have proven effective in much literature but the effectiveness and application depends on the spatial extent and homogeneity of the surface characteristics. For example, in an urban environment object based classification can be done at ease since urban surfaces contain features of similar characteristics which can be identified and classified into land cover type. However, when the spatial extent increases, land cover characteristics becomes complex thus, affecting accuracy and reliability of classification. In addition, object based image classification will requires input data of high resolution which often requires large storage space; and due to their size, processing becomes extremely slow affecting the application of object based interpretation. Advancement from object based image classification is the use of object based oriented classification techniques that utilizes different machine learning languages to identify and group similar colors into a single land cover type. Successful application of this technique has been shown in the work of Lang  (2015) is to operate at the spatial scale of the object of interest rather than at the extent of the image pixels. This makes the technique an efficient, versatile, reliable and cost effective method of image analysis application. The issue of remote sensing sensors which influence image quality prior to classification according to Bukata, Jerome, Kondratyev, & Pozdnyakov (2018); Gu, Lv, & Hao (2017); Toth and Jóźków (2016) has direct correlation with the accuracy of classification especially the pixel based approach. The advancement in science and technology has favored the fabrication and development of multisensory spaced platforms that can measure different environmental parameters thus, improving reliability and accuracy (Desheng and Xia, 2010;Myint, et al., 2011).
This research therefore; identifies the different LULC types and explores the characteristics of LULC patterns using remotely sensed data. The specific objectives of this research however, is (i) to identify the land cover types in Minna, (ii) to evaluate the LULC changes of Minna between 1976Minna between to 2016 to determine the spatio temporal change characteristics of land for the period under review.  All satellite imageries were studied to enable the identification of features of similar spectral bands. Image classification in this research was employed to assigned unique spectral band to a feature by utilization of a process referred to as signature editor. In this process, each feature on the map were given a unique signature as inferred from the composite image and the ground based control points used in developing training samples. The processed satellite images where subject to qualitative evaluation using their spectral characteristics and properties to ascertain t-he uniqueness of different LULC types in the study area prior to classification.

Landsat
The training samples were developed from a Global Position System (GPS) generated points during reconnaissance to enhanced accuracy of locating training samples. Six training samples presented in Table 2 were developed for each class and then refined, rename and merged after examination of statistical attributes as referred to by Dewan and Yamaguchi (2009). The generated training samples where run on 120-150 selected sample sites on the imageries ranging between 300-800 pixels using maximum likelihood classification. A supervised maximum likelihood classification is a classification algorithm used to obtain highly accurate results from a remotely sensed data with an assumption that each class has a   Continues growth in this pattern will result to obstruction of water ways, modification of Land-cover types Description Developed Area All residential and industrial areas, settlement and transportation infrastructural network.

Water body
River, permanent open water, lakes, ponds, canals, and reservoir.

Farmlands
All agricultural lands both on small and large scale, it involves grass-land and ridges.

Rock outcrop
This referred to exposed rock either due to denudation or human activities.

Bare surface
This referred to sand land, , and areas which has no vegetal covered.
hydrogeological structures and increase the potential of the occurrence of environmental hazards including flooding and urban heat island.  (1976, 1986, 1996, 2006, and 2016).

Results and Discussion
The LULC types of Minna are presented in Figure 2. Based on the remote sensing and ground control points, six (6) to Seven (7)  The result of LULCC in Minna from 1976Minna from , 1986Minna from , 1996Minna from , 2006  The spatio-temporal change analysis presented in Table 3 revealed that changes has taking place in the different land cover types over times. Vegetation in 1976 covered 299.85 ha considerably declined by 43.99ha as at 1986. Between 1986-1996    values. This rapid urbanization period coincide with period of many PPPA by the government resulting to increase in infrastructural development, housing, education, commercial, industrial, small and medium scale business, demographic changes, immigration from the surrounding rural areas in addition to availability of social and basic amenities in the suburb also influences this development rate. The spatial characteristics of developed areas also result to the construction and development of more transportation network especially the road transportation. Many road networks within Minna metropolis were either expanded or upgraded to a dual carriage ways while some residential streets of untilled types were tilled to ease movement within the Metropolis.
The geospatial analysis from the Landsat images revealed that expansion from 1976 to 2006 did not occur proportionally to each other. Vegetation for the period under review shows a declined from 54.03% to 18.71% in 2016. Water body on the other hand declined from 13.48% to less than 1percent in 2016. While vegetation and water body shows a tremendous declined in area coverage, farm land and developed areas indicate a significant growth from 0.41% to 10.87% and 2.25% to 27.20% between 1976 to 2016 respectively. A slight increase was also observed in rock outcrop from 4.22% in 1976 to 4.75% in 2016 while bared surface increases from 21.62% to 37.76% for the same period. The high value of change in bared surface was largely due to change in farm land to development of road network and open space development.
The study revealed that the rapid rate of urbanization of Minna has been relatively more rapid during the last decades compared to the 1976-2006 era. The rapid urbanization rate has resulted to the significant changes in LULC pattern with adverse effects on the environment. The expansion rate for the period under study shows that in the last 10years Minna has developed by more than 110 ha.

Conclusion
This research has identified, classify the LULC type and examining the Spatiotemporal changes in LULCC of Minna during the last fifty years using RS, geospatial technology and other auxiliary variables. Analysis revealled that built-up area increase from 74.82ha in 1976 to 116.58ha in 2016. Farmlands increased from 2.23 ha to 46.45ha and bared surface increases from 120.00ha to 161.31ha from 1976 to 2016. The study revealed that this growth rate resulted to substantial reduction in vegetation and water body including wetlands.
The transformation of vegetation and water body to built up areas and farmlands has resulted to a severe environmental degradation with adverse vulnerability to flood occurrences, growth of slums and ghettos. The study demonstrate the capability of effectively integrating remote sensing data and other auxiliary information in identification of LULC types and helps in understanding the dynamics of LULCC of Minna for sustainable. Areas for further study may include determinants of LULCC of water bodies in urban settlements; and an analysis of the effect of PPPA on the urbanization process.