Landscape metrics analysis of land use patterns and changes in suburban local government areas of Ibadan, Nigeria

This study examined the spatiotemporal pattern of urban growth and magnitude of changes in selected Local Government Areas (LGAs) in the suburban area of Ibadan using remote sensing cum spatial metrics. Data for the study were obtained from administrative maps of the study area, population data and satellite imageries. All of these were complemented with ground validations using Global Positioning System (GPS) receiver. Periods of change analysis were divided into two epochs; 1986 to 2002 and 2002 to 2017. The imageries were classified into four landuse/cover classes based on Anderson’s modified version of supervised classification scheme. Accuracy assessment of the imageries was carried out using 100 randomly sampled ground truth points. Ten spatial metrics were selected for analysis using Fragstats program. Results show a significant loss in vegetation due to conversion (Land Consumption Rate, LCR = 0.025, 0.019 and 0.027; Land Absorption Coefficient, LAC = 0.012 and 0.048). Moreover, there was a persistent increase in Number of Patches (NP) indicating a scattered and fragmented but continuous development. However, dwindling values of Patch Density (PD = 90.94, 27.07 and 30.30) indicate expansion through conversion of other landuses at varying rates. Results also indicate an incidence of fragmented low density development in the fringe areas (Area Weighted Mean Patch Fractal Dimension, AWMPFD = 1.37 and 1.39). The study highlights the chaotic land development and unrestrained urban expansion in the study area.


Introduction
Settlements provide a spatial emphasis for most of human activities (Ahmed and Ahmed 2012).
Investigating the spatial forms of settlements can contribute to greater understanding of land use changes, ecological processes, cultures and lifestyles.Examination of the growth pattern of settlements at different time-scales can be used for effective planning of landuse for maximum use of land resources (Abolade 2007;Olayiwola et al. 2014).Hence, the spatial arrangement and the underlying forces of settlement growth and development are important topics of analysis in urban studies (Aguilera et al. 2011).Moreover, there is a revived attention of scholars on the expansion of cities into the neighbouring settlements (Ferreira et al. 2018;Karg et al. 2019).
A suburban area is a transition zone between the urban cores and the neighbouring rural areas (Lawanson et al. 2012;Adedire 2018).In many of the developing countries, this zone is usually characterised by sporadic built-up areas and lower population density.Therefore, they are generally described as newly urbanised zones (Adedire 2018).Suburban development is caused by population and spatial expansion of the metropolitan city, decline in urban environmental quality and easy access to land for building and commercial purposes (Dutta 2012;Appiah et al. 2014).The pull factors at the suburb include improved infrastructure and socio-economic activities and quest for affordable housing (Lawanson et al. 2012;Acheampong and Anokye 2013;Appiah et al. 2014).On the whole, there are changes in both the physical processes and human activities in the suburban areas (Appiah et al. 2014;Karg et al. 2019).
The advent of land-change technology, including progresses in GIS and remote sensing, has provided a platform to scrutinise landscape alterations through space and time (Guindon and Zhang 2009;Estoque and Murayama 2015).For instance, there is an increased awareness in the use of spatial metrics for characterizing the landscape structure and spatio-temporal pattern of landscape transformations (Pham et al. 2011;Fan and Myint 2014;Estoque and Murayama 2015).Shyamantha et al. (2016) reiterated that spatial metrics are developed based on information theory measures and fractal geometry.They noted that the ability to quantitatively describe the landscape structure is a prerequisite for studying landscape functions and transformations.
Spatial metrics is a quantitative way of representing spatial structures and patterns (Chen et al. 2016).In landscape analysis, spatial metrics are usually employed to detect landscape pattern and fragmentation (Dezhkam et al. 2017).Therefore, they are used to analyse the structure or patches of an area with homogeneous thematic entities (Ferreira et al. 2018).Also, they can be employed to quantify the spatial heterogeneity of patches of a common class (Kumar et al. 2018).In settlement studies, spatial metrics are adopted to assess the spatial characteristics and structures of settlements (Ayila et al. 2014;Wondrade et al. 2014;Rastandeh and Zari 2018).The metrics can be calculated as patch-based indices (such as size, shape, edge length, patch density, fractal dimension) or as pixel-based indices (such as contagion and lacunarity) computed for all pixels in a patch (Singh et al. 2017;Inkoom et al. 2018).
Previous studies on Ibadan and its environs have revealed series of settlement expansion patterns (Fabiyi 2006;Toyobo et al. 2011;Agbor et al. 2012;Adegboyega and Aguda 2016).These studies relied on just simple change detection techniques which yielded a single estimate of spatial growth between two dates.Also, the previous studies combined remote sensing with descriptive statistics to analyse growth pattern of the study area.Furthermore, the studies concentrated on either a Local Government Area (LGA) or one settlement in Ibadan suburb; none of them incorporated all or, at least, combined a few of these LGAs or settlements.In effect of these defects, the present study examined the spatiotemporal pattern of urban growth and magnitude of changes in the suburban areas in the North-eastern part of Ibadan using remote sensing cum spatial metrics.

The Study area
The study was conducted in the Northeast suburban areas of Ibadan, Nigeria.Ibadan, the capital city of Oyo State, consists of eleven LGAs; while five constitute the municipal, the remaining six LGAs form the suburb.The area selected for the study consists of three LGAs, these are: Akinyele, Lagelu and Egbeda LGAs.The study area is located between Latitudes 7º 21'N and 7º 41'N; and Longitudes 3º 47'E and 4º 09'E (Figure 1).The National Population Commission (NPC 2006) revealed that the study area contained 643,587 people spread over a total land area of about 969.08km² (Table 1).The major and aboriginal indigenes of study area are the Yorubas, though there are other tribes from different parts of the country (Fabiyi 2006).The study area falls within the tropical climate characterised by a long wet season which lasts for about eight months between March and October and a dry season of about four months between November and February.The mean annual temperature is 26.6ºC with seasonal variation.While the mean annual rainfall is about 1,260mm, the relative humidity could be as high as 76% (Ogolo and Adeyemi 2009;Egbinola and Amobichukwu 2013).The natural vegetation is derived forest consisting of such trees as Iroko, Mahogany, Obeche, and other similar varieties of hard wood (Toyobo et al. 2011).Whereas the northern part of the study area consists of the pre-Cambrian igneous rocks composed mainly of the medium grained gneiss, the southern section is dominated by sedimentary rocks.The major soil type in the study area varies from savannah soils in the northern part and forest soils with deep well drained sandy loamy and sandy clay loamy soil in the eastern and western parts (Oladele and Bakare 2011).
As a result of these geographical accounts, agriculture is the major human occupation in the area.
Among the cash crops cultivated in the area are oil palm, cocoa and kolanut (Adelekan et al. 2014).
Other economic activities include trading, public service employment and transport.Also, there are many local craft industries in the area that are using local materials to produce highly sophisticated farming tools and other items.However, there are few modern manufacturing industries in the area (Fabiyi 2006;Oladele and Bakare 2011;Adelekan et al. 2014).

Data Preparation
Hardcopies of the administrative maps of the three selected Local Government Areas in the suburban area of Ibadan were scanned, georeferenced and digitized for further analysis within GIS environment.Also, population data were used in calculating consumption rate and urban landuse coefficients of the study area.All the imageries used have been geometrically corrected and georeferenced by the providers, therefore no geometric correction was carried out again; they were just projected.The digitized boundary was used to clip each of the imageries so as to extract only those portions that is within the study area.Imagery enhancement, contrast stretching and false colour composites were created to improve the visual interpretability of the imagery by increasing the apparent distinctions between the features.Furthermore, the imageries were classified into four landuse/cover classes using Anderson's modified version of supervised classification scheme to produce land-cover maps of the study area (Anderson 1971).In addition, accuracy assessment of the imageries was carried out using 100 randomly sampled ground truth points (Foody 2002).

Data Analysis
All analyses in this study are based on the contiguous suburban areas in the North-eastern part of Ibadan; all the three LGAs are treated at the whole landscape level.Therefore, the Landsat imageries used in this study were classified into four feature classes, namely: built-up area, vegetation, exposed surfaces and water bodies using maximum likelihood classification algorithm as a supervised classification technique (Table 3).Source: Adapted after Anderson (1971).
In the attempt to quantify the spatio-temporal extent and rate of growth in the study area, results of landuse/cover classification were combined with class area spatial metrics (Cushman et al. 2008;Alphan et al. 2009;Deng et al. 2009;Yeh and Huang 2009;Buyantuyev et al. 2010;Kumar et al. 2018).In this study, ten spatial metrics were selected for analysis based on the potential of each to best describe the landscape pattern in the study area (Table 4).All the metrics were calculated using Fragstats program (McGarigal et al. 2002).

Imagery Classification and Accuracy Assessment
Table 5 shows that the overall accuracies of the classification were 99.98% (1986), 99.62% (2002) and 99.21% ( 2017).These values indicate that there were high significant agreements between reference points and the extracted classes (Foody 2002;Herold et al. 2005).Figure 2 shows the results of the classification and the landuse/landcover maps of the study area.

Physical Changes in the Suburban Areas of Ibadan
Table 6 shows that vegetation covered the greatest proportion of the land at 78.89%, 74.84% and 49.63% in 1986, 2002and 2017, respectively. Exposed surfaces occupied 12.16% in 1986, 12.28% in 2002and 24.87% in 2017. The built-up area occupied 8.82% in 1986, 12.71% in 2002and 25.30% in 2017.Water body occupied the least part of the landcover at 0.13% in 1986, 0.17 Furthermore, Table 6 shows that there was a decrease in LCR from 0.025 in 1986 to 0.019 in 2002 which later increased to 0.027 in 2017.This shows that as the population increases over the years, the level of compactness increased between 1986 and 2017.This is an indication that there is a significant level of fragmentation and an outward spread of the built up class.Also, there was an increase in LAC from 0.012 to 0.048 between the two epochs indicating that as the population increases, rate of conversion of other classes of landuse to built-up also increases (Table 6).The situation can be attributed to the emergence of built-up patches which occurred as the built-up expands outward.Increase in the NP between 1986 and 2002 (from 94 to 809) shows increased fragmentation of the built-up area during the study period (Table 7).This suggests that the period between 2002 and 2017 experienced tremendous expansion rate through the conversion of other landuse classes.Therefore, it can be summed that some patches were merging to form a homogeneous landscape.This is further manifested in the increasing NP and a much faster increase in the CA.The cumulative effect is revealed by the gradual increase in LPI for the built-up land; this is a sign of agreement between the metrics Furthermore, AWMPFD was 1.37 in 1986 and 1.39 in 2017.These results imply that for a unit increase of patch area, the perimeter increased by 37% and by 39% in 2017.The higher value obtained in 2017 shows that the geometry of urban patches is getting more complex over time.This could be an indication of the prevalence of fragmented low density development in the fringe areas.
Results reveal that development pattern was not uniform in all directions and varies with the availability of space for expansion.Several studies have analysed the land use and land cover in the study area using GIS and remote sensing techniques, though at individual settlement or Local Government level (Fabiyi 2006;Toyobo et al. 2011;Agbor et al. 2012;Adegboyega and Aguda 2016).This study has combined spatial metrics analysis with the previous techniques, which helps in understanding the direction of growth and the nature of the landscape in the study area.For instance, the continuous rise of NP indicates that heterogeneous and fragmented growth process was taking place in the study area.However, a comparison of CA and NP metrics reveals that there was rapid urban growth resulting from increase in and expansion of the built-up area.Also, the increasing intensity of buildings in the area increased the LPI of the built-up area leading to a tremendous urbanization pressure over the other land uses.The increase of PD in built-up areas is a result of the urbanization process that occurred rapidly between 2002 and 2017.Increase in population necessarily aggravated the demand for more houses, subsequently creating new built-up areas at the fringe areas.Furthermore, a visual comparison of the classified imageries indicates an infill form of development between 1986 and 2002 and more of edge expansion during the second epoch.

Conclusion
This study investigated the growth patterns of selected LGAs in the suburban areas of Ibadan using remote sensing and spatial metrics techniques.Four major landuse/landcover classes were identified for the study: vegetation, built-up, exposed surfaces and water bodies.The study revealed a significant reduction in vegetation within the study period (1986 to 2017), and a corresponding increase in exposed surfaces and built-up landuse/landcover classes.While the increase in built-up class can be attributed to increase in population, the increase in exposed surfaces could be as a result of shifting cultivation farming systems and opening-up of more lands for road and other construction works.Moreover, this study found that the dominant growth process of settlements in the study area could be attributed to a creation process whereby patches representing settlements areas are increasing.The spatiotemporal analyses showed that a large proportion of vegetation was transformed to built-up areas.A proportional transition of other land uses to the built-up areas was also found to be consistently increasing.

Implications of the Study
The urbanization process in the area has caused fragmentation of the landscape and heterogeneous land use development.The result of landscape analysis showed that the trend of growth is to some extent influenced by haphazard land development practices and uncontrolled urban expansion.However, the heterogeneous landscape development will most likely continue into the next few decades as the built-up areas in the southern part have already started to agglomerate.
Furthermore, the decreasing trend of nearest neighbourhood distance is an indication that homogeneity will increase by refill type of development after sometimes.
In view of the foregoing, the study recommends the need to incorporate the capabilities of Remote Sensing and GIS in landuse policy and decision making processes.This will greatly help in mitigating haphazard growth of the study area and also check the implications of dynamism and complexity associated with urban landuse problems like urban food crisis, poor accessibility, sprawl development at the fringe and other environmental problems that may be imminent.

Figure 1 .
Figure 1.The Study Area Source: Oyo State Ministry of Lands

Figure 2 .
Figure 2. Land use land cover of the Study Area: (a) 1986 (b) 2002 (c) 2017 Sources: Landsat Imageries (Path 191 Row 55): TM 1986; ETM+ 2002; OLI/TRI 2017 in 2002 and 0.20% in 2017.Increase in the proportion of exposed surfaces is an indication of intense human activities that might have encouraged up of the land thereby resulting in significant conversion of vegetated areas.However, a visit to the study area during the field survey revealed that most of the cleared portions (exposed surfaces) towards the eastern part of the area in 2002 have been covered by cultivated crops in 2017.Also, there was an increase in the area occupied by water body between 2002 and 2017 which might be as a result of the dredging and expansion of water channels embarked upon by the Oyo State Government of Nigeria in 2016 (Oyo State Government 2016).

Table 3 .
Classification of Landuse/cover in the suburban Area

Table 6 .
Summary of Landuse/Landcover Statistics

Growth Pattern / Overall Landscape Pattern of the Study Area Table
7 shows that the total built-up area (CA) increased from8,548.38ha in 1986 to   12,317.31ha in 2002and later increased to 24,509.97ha in 2017.The highest rate of urban growth was observed during the second period of urbanization (2002 to 2017) in which the built-up area increased more than twice the previous size within 15 years (Table7).Moreover, NP of the built-up class increased slightly in year 2002 which is an indication of more fragmentation and less of infilling development.The same scenario occurred in 2017 as proportionate increase in NP was recorded in 2017 indicating the continuing development of scattered and fragmented urban patches.

Table 7 .
Spatial Metrics of Built-up Class at the Entire Landscape Level