Multi-Temporal Study of Land Use Land Cover Changes

Anthropogenic activities have profound impact on Land use and land cover (LULC) the world over affecting all abiotic and biotic components in all environments. Studying the effects of past, present and future LULC on forest cover and other dynamic land uses provide valuable information for environmental, land planning and climate change mitigation measures. This study uses a combined methodology of Remote Sensing and Geographical Information System to map historical, current and future LULC through Satellite imagery (Landsat TM, Landsat ETM+, ALOS, Disaster Monitoring Constellation-DMC and SENTINEL) covering Greater Kumasi in a 40-year period. LULC classes such as Agriculture, Built-up, Close Forest, Open Forest, and water were considered based on the predominant geographical sceneries, remote sensing data and field study. Markov Cellular Automata modelling was employed for the estimation of likely LULC changes for the year 2040. The study discovered a perturbing development of forest loss (forest degradation and deforestation), decreasing agricultural activities whiles the Built-up share ballooned. Increasing urbanization at the expense of forest cover and transformation of agricultural lands into human settlements were observed. The forecasted LULC map for 2040 indicated an upward growth in Built up areas at the detriment of the other LULC categories. The outcomes validate the urgent necessity for critical review of regulations in LULC policy strategy, design, and development for the protection of forests and other critical ecosystem services to be preserved. This trend encompassing historical, current and future LULC necessitates that prudent resolutions have to be made to guarantee forest cover, make available land for agriculture and to mitigate the effects of the climate change


Introduction
The Climate Change phenomenon is reckoned to be linked to Land use land cover (LULC) changes and the activities of humans [1][2][3]. Consequently, Land use land cover change (LULCC) information at varying spatial and time-scales is vital in appraising abiotic and biotic components in all environments trends [4,5].
For optimum planning, management, regulation and utilization of the of the earth's resources it is vital for resource managers and users to understand the nature of change and the rate of LULCC [6,7]. Kagombe et al. [8] reasoned that, information on LULCC is also essential in resource economics as it impacts the dynamics of proximal livelihood opportunities as well as distal interrelated economic activities and resources.
Gatsi & Appiah [9] project Ghana's urban population to hit 65% by 2030 from the current 52% of the total population to about 65% by 2030. This surging rate undoubtedly has serious ramifications for natural environments in urban hubs as economic forces escalate [10]. For a long time, Ghana's second largest city (Kumasi) was fondly referred to as West Africa Garden city, unfortunately this is not the case anymore. This prestige has been lost due to urban development [11]. An economic analysis study by Quartey [12] in Kumasi via a Hedonic Price Model contrasted the overall economic usage value of forests with the user value of the forest. The assessment revealed that Kumasi (Ghana) loses a net minimum of US$ 35 million each year in carbon credits because of deforestation.
Remote sensing and Geographic information system (GIS) provide a cost-effective technology which have unique environmental capabilities to monitor vast region of the earth alternative of mapping landscape resources and analyzing changes over the traditional ground-based surveying methods. While the latter methods will continue to be important in ground-truthing exercises for validation and calibration of remotely sensed data, it is generally agreed that application of remote sensing technologies for mapping of resources over large areas and with need for temporal replication is far much economical in comparison to traditional methods. They provide effective tool for analyzing the land use dynamics of a region, as well as for monitoring, mapping and management of natural resources, [15,16]. Indeed, remote sensing approach employing the moderate resolution satellite imageries like Landsat Thematic Mapper, Landsat Enhanced Thematic Mapper Plus, SPOT Vegetation among others has widely been accepted [17,18]. It has gained prominence in wide range of application such as landscape resource assessment, resource monitoring, land cover change analysis, drought monitoring, and biomass estimation among others.
This study intends to produce past, current and future LULC map of Greater Kumasi. The study seeks to assess the LULCC for the period 1990 -2020 in a ten-year interval. This is envisaged to offer an insight into the extent, and pattern of LULCC. The study also purposes to offer a synopsis of the key pressures causing the unbridled urban growth and to forecast the LULCC extent of the region by 2040.

Study area
Greater Kumasi is centrally located in the middle belt of Ghana and its lies within longitude 1058'W and 1011'W and latitude  [19]. The population of the region is anticipated at 5,792,200 as at 2019 from 1,109,133 in 1960 [19]. The climate conditions in study area features both wet and dry conditions with stable temperature throughout the course of the year, an average of 1400mm of rain per year. The topography of the study area is undulating with a number of rivers running through the study area and has an average elevation of 250m above Mean Sea Level (MSL). The Owabi and Berekese head works located in the study area are the main source of portable drinking water. Lake Bosumtwi is natural lake (formed from antique meteorite impact crater) is located within the study area. This lake offers means of livelihood such as a fishing industry to the neighboring towns and villages. It is also a hot hub tourism attracting foreign and local and tourist [20].

Data acquisition
Inventory of Satellite data used for LULC classification and Reference Data are itemized in Table 1

Image preprocessing
Parsa et al. [21] posit that image pre-processing is essential in The images were subsequently enhanced using Histogram Equalization. The 1990 and 2000 Landsat images were found to be hazy and were corrected.

Land use classes
The following broad LULC classes were chosen based on satellite image availability and study of literature (Table 2).

Close Forest
All land occupied by woody vegetation consistent with thresholds used to define Forest Land in the national greenhouse gas inventory and a vegetation structure that currently fall below, but in situ could potentially reach the national threshold values used by to define the Forest Land category in Ghana

Open Forest
All land with woody vegetation consistent with thresholds used to define Forest Land in the national greenhouse gas inventory that are degraded Close forest.

Agriculture
All cropped land, including rice fields, and plantation where the vegetation structure falls below the thresholds used for the Forest Land category and land where over 50 of any defined area is used for agriculture, this may be currently cropped or in fallow and may include areas for grazing of livestock.

Built Up
All developed land, including social utilities such as transportation infrastructure (roads and highways), built up areas, bare grounds and human settlements of any size.

Water
These include lands that are covered or saturated by water for all or part of the year (for example peatlands). It also includes reservoirs and natural rivers and lakes.

Image classification
The modified Anderson Level I classification Scheme [23] which is the worldwide accepted and widely used LULC classification scheme was used in this study to achieve desired

Change detection analysis
The study employed a Post-Classification Change Detection in assessing the LULCC that had occured over the thirty year' period

Modelling and predicting LULC change
CA is a very efficient tool for forecasting spatial dynamic Yang et al. [27] and Feng et al. [26] consequently add that, the development potential value of the cell at location ij at time t is expressed in equation (2) ( ) Veldkamp & Lambin [28], Turner et al. [29] and Feng et al. [26] continue that LR is a multi-variant discovery method that is often combined with CA. LR has the ability to reduce spatial dependency among variables and remove spatial autocorrelation.
Employing LR, the development potential of cell at location ij, (Pg) ij , based on spatial factors is given in equation (3) ( ) The 2010-2020 land-cover maps were first used as inputs in Markov module to generate a transition matrix and a set of conditional probability images between the two dates of the thematic maps. These resulting outputs were later loaded in the CA-Markov module to generate the 2020 predicted map.
Afterwards, the predicted 2020 land-cover map was compared with actual land-cover map of 2020 for validation. Following the validation, the 1990-2020 land-cover maps were used to predict the 2040 land-cover map.

Flow chart for the study
All the GIS and remote sensing processes are summarized in

Image classification and accuracy assessment
Foody [33] and Behera et al. [34] intimate that accuracy assessment is essential, and particularly so, when using post-  The Table 3   The LULC map for 2020 ( Figure 3D) shows Close and Open forests plummeting to 50% Agriculture remaining stable @ 20%;

LULC maps assessment
Built-Ups surge endures from 15% to 27%. Water share of the LULC remains same.

LULC change detection
The Table 4 show

Remote sensing and GIS in LULC appraisal
The optimal utilization of the land and its resources requires an in-depth information of the historical, current and possible future scenarios. RS and GIS provide the tools expedient for monitoring the dynamics of LULC ensuing out of both the changing demands of increasing populace and elements of nature acting to influence the landscape [39,40]. Processes emanating from natural and man-made activities cause the transformation of earth's atmosphere and land [41,42]. The assessment of the spatial-temporal patterns of LULC in forests, rural, urban and other land use forms are necessary to the understanding of the evolution of forest loss, urban systems and other critical ecosystem services. Consequently, information about LULC extent, change and forecasting are essential for apprising land cover maps and the management of natural resources [43,44].
Satellite imagery deliver a proficient means of obtaining information (data) on spatial distribution and temporal trends of LULC required for quantifying, appraising, forecasting and projecting land changes [45,46]. Additionally, in inaccessible (such as mountainous, marshy, glacial and many others) terrain, remote sensing technique is feasibly the only method of procuring the relevant data at a cost and time effective basis [47,48].

LULC trajectory
In the selection of the training sites for the supervised classification of the Greater Kumasi stratified random sampling ensured quality results. This involved employing expert knowledge, as the area of interest was disseminated into strata that maximized the dissimilarities between units and minimize the difference within each unit. Strata designated (One or more) were appraised to be main drivers of the system under scrutiny. A random sample was then selected from each unit or stratum. As recognized variances occur between the strata, stratified random sampling with well-adjusted allocation gives enhanced estimation devoid of bias. The chief advantage of stratified random sampling is that results are generally unbiased and precise. It frequently produces data that is more illustrative of the entire population due to the special attention it gives to the smaller subcategories within the population. Additionally, it provides the best avenue to get results that mirror the diversity of the population under study [49,50]. Stratified random sampling is more effective for large and in the urban centers [19]. The factors enumerated in Table 1 all apply in the study area. This unfortunate situation necessitates a quick and a well thought out plan to salvage what is left.

Conclusion
The application of GIS, remote sensing, and statistical simulations techniques has shown in this study to be a cost effective and potent means for monitoring spatial and temporal LULCC. The study revealed that, built up area grew astronomically [59] and Toure et al. [60] in Ghana.
LULC maps derived from satellite data in this study are not without errors. As Robinove [61] stated, computer produced maps via digital operation of multispectral data is never 100% precise.
The procedure of categorizing a wide spectrum of the World's landscapes into precise and frequently simplified categories brings error by delineating boundaries round geographically situated categories which are 'homogeneous' or suitably heterogeneous.
Nevertheless, these limitations are rectified by comprehensive statistical scrutiny to give reasonably correct LULC maps [62].