Implications of land use and land cover change in Mampong municipality, Ghana

ABSTRACT

Understanding and managing land use and land cover (LULC) changes are crucial for addressing environmental challenges, promoting efficient utilization of natural resources, sustainability of environmental and ecological systems for the well-being of both present and future generations. This study was conducted to map the geo-physical features of Mampong municipality to assess the extent of LULC dynamism using Landsat 7 TOA (top-of-atmosphere) image of 2006 and Landsat 8 TOA (top-of-atmosphere) images of 2013 and 2020. Supervised random forest machine learning classification algorithm was applied to classify and examine the dynamics of land use changes in the study area. Markov chain simulation was adopted to examine the magnitude of transition of land cover changes. The results of the study revealed a substantial change in land cover uses in the study area over the period (2006–2020). The trend analysis in area percentage changes indicated that natural forest and deciduous forest land had reduced drastically mainly for the purpose of agriculture production and build-ups for human settlement as revealed by the negative change or reduction in typical biodiversity hotspot reservations: natural forest (−22.71%), deciduous forest (−26.74%) and water bodies (−32.11%) with a positive change in components such as built-up area (29.11%), barren land (4.96) and agriculture (37.91) in the study area. The study, however, revealed serious trend in water bodies extinction within the municipality as indicated by the dynamic change analysis results, suggesting that 7.86 km² area that was originally under water pathways have been converted for human settlement and 0.76 km² of agriculture lands are exposed to erosion (barren land). A Markov chain simulation prediction of the transition for the year 2032 revealed a positive probability increasing rate in urban and surface build-up and agriculture at a rate of 2.33% and 1.69%, respectively, with substantive reduction in natural biodiversity resources in future given the present scenario. Given the dynamics of LULC changes, we recommend that major stakeholders including environmental protection agency, forest guides, Ghana water company and municipal assembly authorities should develop feasibility holistic preventive measures that seek to ensure effective and efficient utilization of resources and promote the livelihood of individuals who depend on these resources for survival.

KEYWORDS:

• Remote sensing data
• Markov chain simulation
• land use change and land cover changes
• supervised random forest algorithm

1. Introduction

Managing land use and land cover (LULC) changes is crucial for addressing environmental challenges due to the necessity to modify earth’s surfaces for the construction of social structures and facilities to meet the aesthetic lifestyle of the increasing contemporary globally geometric human population which is predicted to rise to 9.7 billion by 2050 (United Nations, 2022) and the sustainability of ecological systems. This rapid expansion in human population accompanied with its concomitant changes in lifestyle, taste and preferences has necessitated and induced the modification of land covers to increase agricultural production to an extent of 70% to provide the necessary quantum of food for the nutritional security and economic development of the growing population (Food and Agriculture Organization, 2022). However, the extent of these alterations to land surfaces has direct or indirect effects on the geo-physical properties with their adverse impacts on the environmental resources, biodiversity and global climate changes (Quintas-Soriano et al., 2016; Winkler et al., 2021), soil erosion and salinity of soils and permanent loss of land fertility and productivity (Foody, 2002; Hegazy & Kaloop, 2015; Liu & Yang, 2015), leaving most of the land covers sustainable for future generation. These transformations have been unprecedented in different geographical parts of the world due to large scale and extravagant utilization of natural resources by anthropogenic activities (Foley et al., 2005; Lambin et al., 2001), social, economic and political factors (Biney & Boakye, 2021; Wijitkosum, 2016). Managing LULC changes has received global attention and has been guided by international conventions because of climate change implications and management of earth surface and ecosystems to promote parsimonious use of available resources. The geophysical composition of many countries has undergone dramatic changes due to rapid expansion of built-up areas in various metropolis reducing vegetation cover dramatically as a result of the associated rise in commercial activities (Agidew & Singh, 2017; Banba & Shaw, 2017; David et al., 2016).

In Ghana, the change has been tremendous and rapid due to the rise in population from 6.7 million in 1960 just after independence to 18.9 million in 2000, 24.2 million in 2010, and 30.8 million in 2020 with an average annual rate growth of 2.15% (Ghana Statistical Service, 2021). This is projected to reach 34.4 million by 2025 and about 50 million by 2050 (United Nations, 2022). Moreover, significant high levels of migration, urbanization, mining pressures resulting in urban sprawl (Amenyo-Xa et al., 2017; Nyatuame et al., 2023), agriculture and herding activities (Yenibehit et al., 2024), cultural and technological factors (Sarfo et al., 2022) are critical factors leading to land deprivation, and climate variation which consequently trigger changes in LULC. Analysis on LULC changes and the rate of change provides critical information to enhance decision-making process in sustainable ecological management and environmental planning.

In Ghana, myriad of studies have been conducted with the concentration on impact of land use dynamics on urbanization and development of cities (Ampim et al., 2021; Iddrisu et al., 2023), sustainability of natural environment and biodiversity (Arfasa et al., 2023; Baidoo et al., 2023), watershed management (Abungba et al., 2022), and forest dynamics (Dembele et al., 2024; Frimpong, 2015; Kullo et al., 2021; Nyatuame et al., 2023; Yeaboah et al., 2017) as a means for justifying, validating, and supervising natural environment with the application of remote sensing (RS). The usage of RS and GIS (Geographic information system) in establishing and identifying the rate and extent of land cover change is to critically provide synoptic view of high-level management information for monitoring and mapping of ecological and biophysical properties, a vital input for the prioritization and rationalization of natural resource management as well as for the development of climate change adaptation strategies (Enoh et al., 2023; Patel et al., 2024; Zhang & Li, 2022). Monitoring LULC dynamics using geospatial techniques requires sophisticated multistage data processing skills with numerous, if not dozens or hundreds of, classifications, methods and algorithms to group image pixels into meaningful classification schemes (Mehmood et al., 2022; Zhao et al., 2024). However, the choice of techniques used depends on the availability of algorithms with the image processing software and experience of the researcher with the different methods. Previous geospatial scientists have established and used various LULC mapping techniques that vary from normal techniques like minimum distance and maximum likelihood (Abungba et al., 2022; Mahmoud et al., 2023; Norovsuren et al., 2019), convolutional neural network (Fang et al., 2019; Lui et al., 2018; Xie et al., 2019), artificial neural network (Buğday & E, 2019; Eshetie et al., 2023; Sisay et al., 2023), the Bayesian maximum likelihood (Ali et al., 2018), hybrid machine learning (ML) models: support vector machine and K-nearest neighbors (Akshya & Priyadarsini, 2019; Alimjan et al., 2018; Prasad et al., 2017), light gradient boosting (Abdi, 2020; Ali et al., 2018) and random forest [RF] (Ghayour et al., 2021; Maxwell et al., 2018; Tavares et al., 2019).

In Ashanti region, Mampong municipality is one of the fastest-growing areas where adequate information on LULC change over the period is almost non-existent. This leaves a gap in understanding the dynamics of LULC change in this municipality. Therefore, the need to research into the implications of LULC change in this municipality using RS and GIS (Geographic information system) is appropriate and cost-effective and has well-established technologies in understanding the landscape dynamics. Thus, this study was conducted to investigate land use changes and dynamics of terrestrial surface changes in terms of natural and man-built types in the municipality and to understand and monitor the overall dynamics of natural environment. This paper, therefore, contributes to previous studies (Frimpong, 2015; Kotei et al., 2013) by putting emphasis on implications of land use dynamics in Mampong using RS and advanced econometric modelling techniques. The remaining sections of the paper are organized as follows. Section 2 provides a brief description of data and the econometric modeling methodology, while section 3 describes the empirical analysis and results. Section 4 contains the concluding remarks.

2. Materials and methods

2.1. Study area

The present study focused on Mampong municipality located in the Ashanti Region of Ghana. The Mampong municipality was purposively selected from the rest of the districts in the region for the study as result of the rapid changes in land use due to upspring of universities and other tertiary institutions as well as its elevation into a municipality in 2007 and to police regional command, resulting in concomitant upsurge in population. The municipality as depicted in Figure 1 is bounded to the east by Sekyere Central, south by Sekyere South, and Ejura-Sekyedumasi to the West. The municipality is located within longitudes 0° 05″ W and 1° 30″ W and latitudes 6° 55″ N and 7° 30″ N, and covers a total land area of 782 km² with 79 settlements of which about 61% being rural. The total population of the municipality is projected as 116,632 with an annual growth rate of 1.3% (Ghana Statistical Service, 2021). The municipality is generally low lying and rise gradually through rolling hills stretching southwards towards the Mampong scarp with the highest point of 2,400 m above sea level, while the lowest is about 135 m. The municipality has an average annual temperature of 27°C with variations in mean monthly temperature ranging between 22°C and 30°C (Ghana Meteorological Service, 2021). The mean annual rainfall, which is bimodal and fairly distributed, ranges between 800 and 1500 mm.

Figure 1. Map of the study area.

2.2. Data and image pre-processing

The study employed secondary datasets, which were derived from Landsat satellite imagery. Landsat TOA (top-of-atmosphere) satellite image datasets of 2006, 2013 and 2020 were used for the study in the municipality.

The images were selected over equal time intervals as dynamic changes in geomorphological land features are usually geological and gradual. Images were captured at approximately the same period to ensure that the selected image had the same reflectance values, which were derived from Google Earth Engine (GEE). The selected imagery was radiometrically corrected to eliminate possible noise of haze, and also geometrically corrected and projected to the Universal Transverse Mercator zone 30 North. According to Yu et al. (2019) and Lu and Weng (2007), band 4, 5 and 6 are the best three-band combination and band 1, 2, 5 and 7 are the best four-band combination which produces almost identical performance for LULC classification. Therefore, bands 1, 2, 5 and 7 of the images were stacked with the application of QGIS (Geographic information system), and R software algorithms were used for the classification of images in the present study, as shown in Figure 2.

Figure 2. Classification scheme.

2.3. Vegetation cover analysis

The normalized difference vegetation index (NDVI) was adopted to study vegetation productivity or mapping of vegetation encroachment and land degradation in the study area. NDVI was estimated by following the mathematical expression:

$\mathit{NDVI}=\frac{\mathit{NIR}\mathit{\hspace{0.17em}}-RED}{\mathit{NIR}+\mathit{RED}}$

where NIR and RED are spectral information in the near infra-red and red bands, respectively, of the satellite imagery. By design, the NDVI values varies between −1.0 and +1, with higher and positive values indicating higher photosynthetic activity or vegetation greenness. Values near zero (0) are usually non-vegetated areas and negative values indicate water bodies (Wegmann et al., 2016).

2.4. Classification algorithms and classes of land cover type mapping

In this study, supervised random forest (RF) machine learning (ML) classifier algorithm was adopted for the classification of the selected imagery. RF is a robust algorithm for classification and regression of data with high dimensionality and potentiality of discrimination and grouping of digital levels of image classification which produces good results that are suitable for the classification of multisource RS and geographic data. This is due to its ability to minimize the correlation between the classifiers in the ensemble by bootstrapping and aggregating o ensemble to form classification and regression tree (Kulkarni & Barrett, 2016; Wegmann et al., 2016). The RF classifier is not sensitive to noise or overtraining as the resampling is not based on weighting, and it is computationally much lighter based on boosting as well as Out of the Bag error estimate and evaluation of variable performance (Ghayour et al., 2021; Tavares et al., 2019). Moreover, the RF was adopted due to its ability to handle Hughes phenomenon as a result of the two-randomness dimensionality and, the capability to produce good classification results with a relatively reduced number of supervised training samples (Zhao et al., 2024). The land classes as described in Table 1 were developed and served as the supervised training sites (Region of interest) fed to the RF algorithm for the classification of the geomorphological land use features in the present study area. The trained sites were developed using the Google Earth map along with expert knowledge on spectral signature of landmark features based on the geological composition of the area.

Table 1. Description of land class for classification

Land cover type	General description
Deciduous forest	This class describes all areas that are characterized by sparsely located trees and shrubs.
Forest	Areas dominated by closely knit trees, dense vegetative cover and all vegetated areas that are not exposed to bare soil.
Agriculture/farmland	Lands under annual tillage and, lands that have been cleared for preparation of crop cultivation, and lands that are left untilled, such as fallow lands, are identified as agriculture lands.
Built-ups/urban	Areas demarcated for community service (parks, playing grounds, lorry parks), residential areas, commercial and industrial areas were classified as built-up areas.
Barren lands	Areas exposed to erosion, degradation and open patches land were characterized as barren lands.

2.5. Assessment of accuracy

Assessing the accuracy of classification algorithms enables the evaluation of the error incurred during preparation of final thematic LULC maps. Therefore, the prepared LULC maps were subjected to an overall accuracy and Kappa coefficient statistical test. Kappa coefficient, a robust nonparametric statistical method that measures and describes the interrelationship between categorical variables in confusion (error) matrix (Olofsson et al., 2013; Rossiter, 2004), was determined by using the following equation:

$\hat{K}=\frac{n{\sum }_{i=1}^r{x}_{\mathit{ii}}-{\sum }_{i=1}^r{x}_i+\mathit{x}+\mathit{i}}{n^2-{\sum }_{i=1}^r{x}_i+\mathit{x}+\mathit{i}}$

The overall accuracy of image classification was determined by the following equation:

$\mathit{Overall}\mathit{Accuracy}=\frac{{\sum }_{i=1}^r{x}_{\mathit{ii}}}{\mathit{x}}$

where n stands for total number of observations (pixels), while r is number of rows in matrix. $x_{\mathit{ii}}$ represents number of observations in row i and column i. On the other hand, marginal sum of row i and column i is denoted by $x_i\mathrm{and}\mathit{x}+\mathit{i}$, respectively (Blaschke, 2010; Islam et al., 2018).

2.6. Dynamics of land use and land changes

Markov chain analysis was used to examine the dynamics of LULC changes in the study area. Markov chain approach was based on the probability distribution that a given class of land change under current study depends on distribution in the previous period that could be represented by a stationary first order Markov chain. Given the classes of land use dynamics, stochastic matrix with the probability A_ij of movement of area under a given land class i in period t to another land use class group j in period t +1 is called transition probability matrix (TPM) which is expressed as follows:

(1) $\mathrm{P}=\left[\begin{array}{ccc}{\mathrm{A}}_{\mathrm{ij}}& \cdots & {\mathrm{A}}_{\mathrm{is}}\\ ⋮& \ddots & ⋮\\ {\mathrm{A}}_{\mathrm{ij}}& \cdots & {\mathrm{A}}_{\mathrm{ss}}\end{array}\right]=⟨\frac{\mathrm{Q}}{0}|\frac{\mathrm{R}}{\mathrm{I}}⟩$(1)

where $A_{\mathit{ij}}\ge 0$ and ${\sum }_j{A}_{\mathit{ij}}=1$

Here, P is a square matrix (n × n), with n being the total number of land classes under study. Each element (A_ij) represents the probability of moving from one land use class i to other j. Following Kemeny and Snell (1976) and Tariq et al. (2022), the final probability matrix (F) for each state through finite Markov chain theory was defined as follows:

(2) $\mathit{F}=\left[\begin{array}{c}\prod _{\mathit{i}=1}^{\mathit{s}-1}{A}_{\mathit{i},\mathit{i}+1}\\ \prod _{\mathit{i}=2}^{\mathit{s}-1}{A}_{\mathit{i},\mathit{i}+1}\\ \prod _{\mathit{i}=3}^{\mathit{s}-1}\underset{\underset{{\mathit{A}}_{\mathit{ss}}}{⋮}}{{\mathit{A}}_{\mathit{i},\mathit{i}+1}}\end{array}\right]$(2)

Given the F matrix, a Markov chain simulation of the transition probability was carried out to project the area change for the selected land use class for next decade based on the assumption that the dynamics follow a constant nonlinear growth rate shift as defined in equation 4(4) $\mathit{InY}=\mathit{In}\mathit{a}+\mathit{t}\mathit{In}\mathit{b}+\mathit{u}$(4) . The calculated land area of each class for the period under study (2006–2020) was converted into time series data for growth rate analysis to understand the dynamics of land use changes over years by fitting exponential growth model expressed as follows:

(3) $Y_t=\mathit{a}{b}^t{e}^u$(3)

where $Y_t$ is the estimated land area under each class, $\mathit{a}$ is intercept, $\mathit{b}$ is the regression coefficient, $\mathit{t}$ is the time variable and $\mathit{u}$ is the error term. The growth rate was obtained from the logarithmic form of the equation (1)(1) $\mathrm{P}=\left[\begin{array}{ccc}{\mathrm{A}}_{\mathrm{ij}}& \cdots & {\mathrm{A}}_{\mathrm{is}}\\ ⋮& \ddots & ⋮\\ {\mathrm{A}}_{\mathrm{ij}}& \cdots & {\mathrm{A}}_{\mathrm{ss}}\end{array}\right]=⟨\frac{\mathrm{Q}}{0}|\frac{\mathrm{R}}{\mathrm{I}}⟩$(1) as

(4) $\mathit{InY}=\mathit{In}\mathit{a}+\mathit{t}\mathit{In}\mathit{b}+\mathit{u}$(4)

The percent growth rate (g) was derived using the following expression:

(5) $\mathit{g}=\left(\mathit{Anti}\mathit{In}\mathit{b}-1\right)\ast 100$(5)

where $\mathit{g}$ is the growth rate in percentage.

3. Results

The study of the geo-anthropogenic effect on the LULC in the present study area began with examination of the biophysical properties and landmark features in the geographical study area to enable proper identification of physical objects of the earth. However, bands 5, 4 and 3 were combined to form false color composition (FCC) to examine the spectral signature and landmark features in the study area as depicted in Figure 3. The FCC was used to identify the various land use class based on their respective spectral signatures.

Figure 3. False color composition from image.

3.1. Vegetation cover analysis

The study of the dynamics in vegetation was analyzed using normalized differenced vegetative index (NDVI) as presented in Table 2 and Figure 4. The results revealed that there has been significant change in vegetation due to the reduction in NDVI values over the study period. The reduction in the average NDVI value of 0.43 in 2006 to 0.26 in 2020 suggests that there has been a complete degradation of vegetation cover. However, changes cannot be identified with the changes in NDVI, and thus the exact changes are examined in the next sub-section.

Figure 4. Plot of NDVI map.

Table 2. Summary of NDVI

NDVI	2006	2013	2020
Minimum	0.12	0.06	0.05
Maximum	0.58	0.49	0.56
Mean	0.43	0.31	0.26
SD	0.05	0.06	0.09

3.2. LULC classification map

The classified images of the original multispectral image and its decomposition into land use thematic map of the various land classes under present study using the RF algorithms are shown in Figure 5. The classification accuracy estimated for the thematic maps produced from the original image was determined by the overall train accuracy and Kappa coefficients. The validation of the thematic classified image produced an overall accuracy value ranging from 91% for the 2006 classification map to 96% for 2020 thematic map with the kappa coefficient of 0.88 and 0.94, respectively. The kappa coefficient of 0.94 in 2020 implies that the classified classes were 94% totally identical to the ground truth features, whilst 88% and 89% identicality were observed for 2006 and 2013 images, respectively.

Figure 5. Land use classification map.

The results from the visual inspection of the thematic maps as depicted in Figure 5 indicate a substantive change, suggesting land covers have undergone dramatic shift over the years due to anthropogenic effect in the study area, as depicted by colour changes in the map. However, further analysis was conducted to estimate the area covered by each class and the extent of change among the various classes of land cover features under present study as presented in Table 3. The results of percentage trend analysis revealed that forest land cover has reduced drastically by 22.71% at a growth rate of −0.97% per annum. This magnanimous change in forest has serious implication on loss of natural resources and biodiversity as indicated by recent remarkable changes in climate variables and rainfall pattern in the study area.

Table 3. Area of land cover

Land cover	Area (km^2) 2006	Area (km^2) 2013	Area (km^2) 2020	Percentage change 2006–2020	CGR (%)
Barrenland	74.48	73.63	78.04	4.56	0.24
Urban	116.31	141.79	164.06	29.11	1.88
Farmland	80.19	120.75	129.17	37.92	2.69
Water	42.75	38.14	32.36	−32.11	−1.30
Forest	159.56	143.24	130.03	−22.71	−0.97
Deciduous	91.90	82.35	72.51	−26.74	−1.12
Grassland	117.89	83.18	76.91	−53.28	−1.92
Total	683.08	683.08	683.08

CGR = Compound Growth Rate.

Moreover, during the study time interval (2006–2020), there has been a considerable reduction in deciduous forest area at an exponential rate of −1.12% per annum translating into −26.74% change in total land area of deciduous forest. The drastic change and degradation in deciduous and typical natural forest are very likely to be associated with the rapid expansion of urbanization/built-ups which increased at an exponential rate of 2.23% per annum during the study period. The expansion of road network for easy accessibility to newly upcoming towns and villages within the municipality and intensification of agriculture which also witnessed an increase at an exponential rate of 2.69% during the period (Table 3) to produce food to meet the demand of the population which is increasing geometrically at an annual growth rate of 1.3% (Ghana Statistical Service, 2021) within the municipality, and the increased exploitation of resources for gold and timber and firewood harvesting as a local source of energy for cooking. These results are in line with a study conducted by Boakye et al. (2008), Frimpong (2015)and Kullo et al. (2021) who reported similar trend of forest resources degradation in an adjacent district to the present study area. The significant growth in urban settlements over the study period is inevitable and expected in a developing municipality where urbanization and migration exacerbate much pressure on existing land and land resources for urban structures, such as housing and commercial activities, as the expansion and development of these facilities depend directly on adjacent natural environment. This result is similar to a study conducted by Amenyo-Xa et al. (2017) and Tuffor-Mills et al. (2020) on LULC in Ghana who argue that built-up areas were one of the main principal factors responsible for reducing vegetation cover in Ghana due to upsurge in the demand for housing, and increasing commercial activities. However, barren land during the period increased substantially to 4.56% at an annual growth rate of 0.24%. The increase in barren land could be due to the activities of charcoal burning in the area as the activity dwells on forest product to meet the increasing demand of local energy for cooking resulting in deforestation as revealed by the drastic reduction in forest resources. Regarding croplands, agriculture land in the municipality is largely dispersed due to the nature of land ownership and type of farming practices, increased significantly by 37.92% at an annual exponential rate of 2.69%. This signifies the importance of crop production and availability of food in the municipality which is essential for food security, accounting for relatively low prices of locally produced food commodities within the municipality relative to other districts in Ghana. Moreover, water and river bodies in the district also reduced drastically at a rate of −1.30 per annum. This result supports a study conducted by Kotei et al. (2013) on the effect of land use pattern on hydrological systems in the study area that reported a substantive 12.25% reduction in the annual mean streamflow in the study area. The drastic changes on water bodies could be due to the filling up of water bodies pathway, especially mushy areas within and around upcoming commercial area for building new structures for commercial activities such as car washing bays and sand wining. In this regard, further analysis was carried out to examine the dynamics of changes as discussed in the next sub-section.

3.3. Change detection analysis and dynamics of land use changes

The change detection in the LULC was analyzed by using the differences in the NDVI. The result is presented in the histogram depicted in Figure 6. The distribution of the normalized differenced vegetation index indicates that there has been a reduction in vegetation cover as revealed by the negative NDVI.

Figure 6. Histogram of change in NDVI.

To further examine the exact changes that has happened in areas which cannot be revealed by the changes in the NDVI, post classification analysis was conducted to examine the extent of land changes and tendency of change. The results are presented in Table 4.

Table 4. Area changes in land use changes (km²)

Name	Forest	Urban	Barrenland	Deciduous	Grassland	Water	Farmland
Forest	0.00	1.41	0.47	4.25	0.96	0.85	9.04
Barrenland	0	0.29	00.00	0.53	1.38	0.002	1.88
Deciduous	1.01	8.43	1.07	0.00	4.83	0.66	5.72
Grassland	0.36	10.21	0.86	5.09	0.00	0.035	6.61
Water	0.27	7.86	2.58	0.44	0.11	0.00	0.87
Farmland	0	8.21	0.70	2.03	5.23	0.00	0.00

The results of the dynamic changes indicate that 8.0% of the total area change in forest, representing 9.04 km² over the study period has been converted to agricultural land with transition probability of 23% (Table 5). Moreover, 1.41 km² of forest area was converted into urban land use with 9% tendency of transition. The dynamic change results indicated that 8.43 km² of deciduous forest has been transformed into urban usage, which has serious implication on soil fertility and destruction of natural ecosystem. Moreover, we found a clear tendency of 8.21 km² of agricultural lands been converted in to urban with transition likelihood of 10% (Table 5). This could be due to the sale of relatively closer arable land to estate developer agents for urban expansion. The results, however, indicated that 7.86 km² of the total area of water bodies were converted into urban settlement for commercial activities which might have serious implication on urban flooding in future. The transition matrix results (Table 5) revealed that forest and deciduous degradation are purposely for urban settlement. However, the probability of agricultural land abandoned to barren land was about 2%. Therefore, utilization of urban space can be enhanced with proper urban planning and the application of modern urban planning architectural technology and design. The study results revealed a 33 % tendency of barren land areas to be rejuvenated into grassland with emergence of shrub. Other classes of land use are most likely to have changed by 2020, with a 13% probability of grassland being converted to farmland. The low likelihoods of farmland, forest and grassland areas being converted into barren land cover type (about 2%, 3% and 5%, respectively) are remarkable to ensure rejuvenation and protection of biodiversity in area. This can be attributed by recent efforts of policy makers in creating awareness on the need of conversation of natural reserves through the promotion of afforestation programmes.

Table 5. Transition matrix of land use dynamics

	Barrenland	Urban	Farmland	Water	Forest	Deciduous	Grassland
Barrenland	0.28	0.39	0.00	0.00	0.00	0.00	0.33
Urban	0.00	1.00	0.00	0.00	0.00	0.00	0.00
Farmland	0.02	0.10	0.83	0.00	0.00	0.05	0.00
Water	0.00	0.20	0.00	0.80	0.00	0.00	0.00
Forest	0.03	0.09	0.23	0.00	0.58	0.00	0.07
Deciduous	0.00	0.12	0.08	0.00	0.02	0.78	0.00
Grassland	0.05	0.10	0.13	0.00	0.00	0.04	0.68

Simulation of the transition probability as depicted in Table 6 was carried out to examine the changes that would be likely to happen by the next decade years (2032) among the various land use class using the percentage nonlinear rate of change (CGR) as per equation 6 as the baseline scenario due to rapid changes in population in municipality with higher demand for social amenities and food requirement to meet the developmental agenda in the municipality. The results of the simulation process are presented in Table 6.

Table 6. Projected land use changes by 2032

Land cover	Area (km^2) 2020	Projected change at 2032	CGR (%)
Barrenland	78.04	60	−0.23
Urban	174.06	204.16	2.33
Farmland	129.17	140.45	1.69
Water	32.36	25.56	−0.128
Forest	130.03	125.57	−0.487
Deciduous	72.51	69	−0.289
Grassland	76.91	58.34	−0.348

The results revealed that, given the above scenario, all thing being equal, natural forest, deciduous forest and grassland are likely to reduce at rates of −0.487%, −0.289% and −0.348% per annum, respectively, from their current state to other land transformation. Water bodies are likely to extinct at a rate of −0.128% which will pose serious threat to agriculture, urban livelihoodsand loss of important biodiversity hotspots. The former is essential so as to meet the food demand of the population as production of important commercial vegetables demands water during production especially in the dry seasons whilst the later could lead to flooding of settlement due to blockages of water pathway as recently observed in major urban settlements. Moreover, other areas are likely to be converted to urban area with 1% of agricultural land being barren land by 2032. The chances that there will be a change in other land cover for urbanization is 39.3% with 2% chance that other land type will be converted for agriculture production, resulting from changes in typical forest. These results provide the impetus for the requirements of immediate, stringent preventive measures on the utilization of land cover in the municipality.

4. Conclusion

Land use and vegetation cover changes, accompanied by recent climate changes, are the contemporary challenges facing humanity as they pose threat to ecosystem and biodiversity. Various methods have been developed to study the dynamics of changes, but RS has proven to be a viable method to study earth surface changes. With the application of remote sensing, the changes and transformation of land cover in Asante Mampong municipality have been examined. The results of the study showed that there has been a substantial change in the geo-physical features in the study area during the study period (2006–2020). The analysis revealed a negative change or reduction in natural forest (−22.71%) and deciduous forest (−26.74%), and positive change in components such as built-up area (29.11), barren land (4.56%) and agriculture (37.91%) under study. The dynamic change analysis results revealed that natural forest and deciduous forest land transformation were mainly for the purpose of agriculture production and build-ups for human settlement. The study, however, revealed serious trend in water bodies extinction within the municipality as the water bodies are left to be barren land due to urbanization and illegal activities of lumbers and sand wining with serious implication on food security and biodiversity. The use of spatiotemporal data on LULC change has revealed general picture of the main land use alterations in one of the most important districts of Ashanti Region. Our Markov chain simulation of the transition probability revealed positive increasing rate in urban and surface build-up and agriculture at a rate of 2.33% and 1.69%, respectively, with substantive reduction in natural biodiversity resources such as natural water bodies that are likely to be converted for other land form utilization by 2032. However, these adverse changes in land cover, which affect biodiversity and communities, if continued could pose serious environmental threat in rainy seasons for communities and food insecurity during the dry season. This study result will serve as fundamental information to local stakeholders on changes in LULC in the municipality during development planning for managing the use of land resources. Therefore, given the dynamics of LULC changes, we recommend that major stakeholders including environmental protection agency, forest guides, Ghana water company and municipal assembly authorities should develop feasibility holistic management preventive measures that seek to ensure effective and efficient utilization of resources for future generation and promote the livelihood of individuals who depend on these resources for survival.

Author contributions

All the authors contributed by conceiving, designing the research, analyzing the data, and critically reviewed and approved the final version of the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data used for this work is a public data that was downloaded from GEE. Landsat 5 TM⁺ image period of 2006 and Landsat 8 TOA (top-of-atmosphere) satellite image for the period 2013–2020. Code for data extraction was obtained through the link https://code.earthengine.google.com/. R-codes for data collection, processing and analysis are available on request.