1 Introduction

Historical development of land use systems and climate variability is anchored in underlying theories such as Millennium Sustainable Development Goals (SDGs), Forest Transition Theory (FTT), Climate Vulnerability and Adaptation, as well as, the concept of sustainability. Obtaining accurate and up-to-date information on the spatial dimensions of land use systems and climate variability is essential in gaining a comprehensive insight into a country’s overall political, socio-economic, technological and environmental status. Few currently deny that extreme weather and climatic conditions are among the most pressing problems of our times. There is general agreement among some schools of thought that humans are intrinsically part of the problem and the solution (Jia et al., 2019; IPCC, 2000). Land use cover change (LUCC) and climate variability first emerged as earth and atmospheric concerns rather than a societal problem; it soon became clear that understanding and addressing its global and local scope would require more than a narrowed approach. Historically, the link between climate change, land use and land cover (LULC) systems, ecosystem services and livelihoods in developing countries has been well established (Wang, 2016; Boon & Ahenkan, 2011; Guo, 2002; Wang et al., 2001). Global satellite launch and data capture commenced in the late 1950s (Sodango et al., 2017). However, data capture by countries varied across the globe post-satellite launch. Ghana’s spatiotemporal data capture for some key natural resources and climate variables commenced in the 1970s. Global lands and other natural resources have undergone severe pressure, mainly due to a host of factors and competitive interests. These pressures are driven by natural and anthropogenic factors. For instance, conversion of forests into other land cover types and climate variability have impacted several ecological processes and functions. Whether separately or combined, these two phenomena pose devastating threats to all localities, especially those in least developed countries (Obeng & Aguilar, 2015; Yiran et al., 2012; Käyhkö et al., 2011; Yue et al., 2007; MEA, 2005; Mitchell & Csillag, 2001).

Climate variability and LUCC have been observed to be interlinked with evidence at various scales according to IPCC’s special report on land and climate (Jia et al., 2019). On-going global warming is associated wit2011h a mesoscale element like LUCC (Gichangi et al., 2015; Manatsa et al., 2014). They constitute those that affect individual households through the regional and continental scales (Mattah et al., 2018; Peters et al., 2008). Negative implications comprising induced single large climate events (Lambin & Meyfroidt, 2011; Fenger, 1999) increase in air pollution (Cheng et al., 2014; Lü et al., 2006), deterioration and impediment of water resources (Cai et al., 2016; Mensah et al., 2018; Schumann et al., 1991), alteration of natural habitats, physiological processes and behaviour of some ecosystems (Gorgani et al., 2013; Kalnay & Cai, 2003; Liu et al., 2011, 2014; Peters et al., 2008; Yue et al., 2007) have been noted. Among the effects experienced by poor individual households are the multiplicity of health-related issues as well as poor living conditions. Vegetation co-exists with climatic conditions (Chase et al., 1999; Yiran et al., 2012). Land use system analysis is an essential component of LULC. Its development and uses are tied to several needs. First and foremost, fundamental research into spatiotemporal development of LUCC inspires understanding on land use systems and processes. Secondly, vegetation, built-up and water indices have a relationship with LST and precipitation (Ullah et al., 2020). Hence, there is the need to establish the link between various indices and climate variability in the study domain. The quest of achieving Millennium Sustainable Development Goals (SDGs) 1, 2, 13 and 15 have increased the growing concerns for further commitments, investments, needed political will, capacity building programs, change in lifestyles and behaviours to avert the known and unknown (UN, 2016).

The past 50 years have seen unprecedented LUCC due to significant transformations in the spatial development of land use systems in Ghana. This change is closely related to the development of agriculture. The agricultural sector remains the main stay of Ghana’s economy. The sector houses 50% of labour force and contributes 20% to the country’s GDP (Doe et al., 2018; MESTI, 2015; GSS, 2014; NDPC, 2010). Livelihood dependence on the sector results in the use of several tracts of land, increasing concerns to curtail risks associated with global food security, micro/macro-economic stability, foreign revenue through exports, employment rates and so on. Policies embedded in these parameters have fuelled interests among interested parties to invest in large- and medium-scale (commercial) agricultural activities. Several researches have attributed the root cause of LUCC in the country to expansion of croplands. Economic growth, on the other hand, was mentioned as an important driver with accompanying rapid urbanization and industrialization processes (Acheampong et al., 2018; Damnyag et al., 2017; Hengazy & Kaloop, 2015; Gougha & Yankson, 2012; Kusimi, 2008). However, wildfire, pests, disease infestation and other natural disasters are not major drivers since their impacts are insignificant towards land cover change in a country for several years (FAO, 2011; Attua, 2010; Chuvieco & Congalton, 1988). Consequently, this unceasing and unprecedented LUCC in Ghana has resulted in habitat, productivity and biodiversity loss, pollution risks, alteration in prevailing microclimatic conditions, desertification, waste accumulation and other associated problems. Several studies have confirmed a great deal of efforts to mitigate the negative impacts of LUCC by the government of Ghana. The rate of forest cover decline was estimated to be 2.24% per annum between 2005 and 2010 (Forestry Commission, 2015; MESTI, 2015; NDPC, 2010). Presently, Ghana’s Forestry Commission projects a 2% (1154 km2) annual rate of deforestation (Forestry Commission, 2015; Oduro et al., 2015). Recent trends show the most obvious LUCC in Ghana is the remarkable increase in agricultural lands across the length and breadth of the country. Surprisingly, these unprecedented changes can be observed in the northeast, east-central and southwestern regions of Ghana (Forestry Commission, 2015). The increasing rate of farming activities and human settlements is unheard of in the country’s history, foraying other land cover types which comprise Ghana’s savannas, forests and woodlands. Between 1975 and 2000, agricultural lands amplified from 13 to 28% per the total land area of Ghana. This continuous dilation persisted from 2000 to 2013 where agricultural activities skyrocketed, reaching 32% of the country’s total area size. This expansion has significance beyond the simple area numbers.

Studies have noticed a positive correlation between the extensiveness of agriculture or economic growth, and environmental changes across the globe. According to Abbam et al. (2018), Lwasa (2014) and Silitshena (1996), such changes constitute LUCC, biophysical and biogeochemical changes as well as fluctuations in hydrological cycles. Apparently, most of these studies were carried out using GIS and remote sensing imagery. In spite of the complex nature of climate variability and LUCC, the essence of remote sensing/GIS tools, existing literature, expert knowledge and experience on LUCC, driving forces, stakeholder involvement, effectiveness of institutional and policy frameworks, among others, cannot be overlooked (Sarfo et al., 2019; Yiran et al., 2012). Importantly, using satellite imagery and existing literature have proven to be vital instruments in assessing historic changes associated with spatiotemporal development of land use systems and microclimatic conditions (Barakat et al., 2019; Bora & Goswami, 2016; Changkakati, 2019).

Southwestern Ghana hosts the country’s rich natural resources which constitutes gold, timber, bauxite, coffee, crude oil, cocoa, ecological reserves, national parks/wildlife (eco-tourist sites) among others. Southwestern Ghana (wettest part of the country with an average rainfall of 1600 mm per annum) receives the highest amount of rainfall among the other zones in Ghana. Again, the study area is noted to produce more than half of Ghana’s cocoa productivity (Tsekpo, 2018) as well as timber. The competitive interests in the use of these natural resources for various reasons among relevant stakeholders have placed pristine environment under enormous pressure. These reasons have partly driven the need to carry out this study, aimed at assessing the spatiotemporal development of land use systems and climate variability in Southwestern Ghana between 1970 and 2020. The study would propel practical capacity building at various regional and district levels through the development and strengthening of local institutions in aspects related to sustainable land and natural resource use, adaptation practices and climate-smart agriculture.

1.1 Forest transition theory (FTT) analysis: in the context of selected SDGs (1, 2, 13 and 15)

The Forest Transition Theory (FTT) as propounded by Alexander Mather during the 1990s seeks to primarily assess forest recovery dynamics under numerous circumstances across differing geographies (Rudel et al., 2005). It closely monitors changes in forest cover from a state of deforestation to that of reforestation. Thus, it offers opportunities to analyse the push and pull conditions which play key roles in shifting non-forested areas to forested areas (and vice versa), despite the theory’s focus on deforestation (Singh et al., 2017). Arguably, some common causal factors responsible for loss of forested land area include population increase, land right issues, agricultural modernization and forest reliant practices among others. Conversely, sustainability policy implementation regimes, climate change mitigation and adaptation, coupled with reforestation initiatives among others, contribute to gains in forested land area (Grimm et al., 2008; Singh et al., 2017), and promote forest transition efforts.

It is critical to mention that some popular studies have looked at the FTT concept from differing standpoints. For instance, earlier in 2007, Perz argued that the quest for a modern economy by nations defines their journey towards forest transition, thereby focusing on the modernization theory’s linkage to forest transition (Perz, 2007). Later, Barbier et al., (2010), in their research viewed FTT as a ‘natural’ trend of ‘forest loss followed by forest gain’ affected positively or negatively by policy favourability or non-favourability, respectively. This review however, is not aimed at delving into differing views on FTT.

A close observation of some underlying factors for shifting land areas from degraded to reforested status as indicated above, explicitly underscores key elements of ‘sustainability’. In this regard, the objective of this brief analytical study is to review the relationship between FTT and the Sustainable Development Goals (SDGs); specifically, SDGs 1 (No Poverty), 2 (End Hunger), 13 (Climate Action), and 15 (Life on Land). This shall encompass the assessment of socio-economic dynamics of land use, climate change impacts and forest resource management. Studies have directly and indirectly reflected on the linkage between ‘Sustainability’ and ‘FTT’ (Rudel et al., 2005; Turner et al., 2007; Mather et al., 1998). When viewed through the lens of the SDGs, making gains in FTT application may complement global efforts at achieving the SDGs 1 (No Poverty), 2 (End Hunger), 13 (Climate Action) and 15 (Life on Land) in various ways (Fig. 1).

Fig. 1
figure 1

Source Adopted from Mather (1998)

The forest transition theory (FTT).

Full size image

2 Methodology

2.1 Description of the study area

The study area (Figs. 2 and 3) lies within latitude 5.3902°N and longitude 2.1450°W. Southwestern Ghana is approximately 23,921 km2 (9236 sq. mi) representing about 10 percent of Ghana’s total land surface with a long stretching coastline of about 202 km from Ghana’s border with Cote D’Ivoire to the area’s boundary with the central region. About 75% of the nation’s high forest vegetation can be found in Southwestern Ghana. The population of the region according to 2010 Population and Housing Census stood at 2,376,021 with 1,187,774 males and 1,188,247 females.

Fig. 2
figure 2

Location of the study area: a Map of Ghana b Southwestern Ghana

Full size image
Fig. 3
figure 3

Source Adapted and modified from deGraft-Johnson et al. (2010)

Forest vegetation in the study area.

Full size image

2.2 Satellite imagery and classification of land use systems

In this study, six Landsat images archived for the given period (1970–2020) from Landsat 5 MSS, Landsat 4 TM, Landsat 5 TM, Landsat 7 ETM + and Landsat 8 OLI/TIRS were acquired from the United States Geological Survey’s (USGS) website (http://earthexplorer.usgs.gov/). Images obtained from USGS website had visible and clear weather conditions which cover the entire study area. ArcGIS 10.6, ENVI 5.0 and 5.3 were used for the image pre-processing. Other image pre-processing procedures which were performed include image calibration, layer stacking and supervised classification (Fig. 4).

Fig. 4
figure 4

Flowchart illustrating data and image processing procedure

Full size image

The flow-diagram above illustrates a system process on where and how data were acquired for this study, processed and analysed. LUCC type was determined using ENVI (Maximum Likelihood Classification Algorithm (MLCA). Topographic maps of the study area were designed using GIS tools. All data sets used in the study were geometrically calibrated using UTM Zone 30 North projection (Tables 1 and 2).

Table 1 Description of imagery data used for land cover change study in Southwestern Ghana
Full size table
Table 2 Description of land cover types identified in the study area
Full size table

2.3 Change detection analysis

Change detection analysis was run to ascertain the regularity of land use systems and its drivers in southwestern part of Ghana (1970–2020). The present study applied image differencing, NDVI, post-classification and GIS techniques in determining the spatiotemporal development of land use systems in Southwestern Ghana. These are common standard change detection analysis techniques, used in several environmental monitoring and modelling related studies. ArcGIS 10.6 was used to produce absolute pixels and the percentage of pixels in each LUCC type.

2.4 Temperature analysis

2.4.1 Image calibration (radiance)

Radiometric correction (radiance) was done to rectify atmospheric effects and enhance clarity. Gap-filling was done for images that may have had stripes. Distortions in images were removed in the calibration process. It was done in order to enhance the image quality before other processes were carried out.

2.4.2 Digital number (DN) to radiance

The ETM + DN values range between 0 and 255. A conversion of DN to spectral Radiance was done according to Coll et al. (2010) as retrieved from USGS Landsat User handbook. The mathematical expression below was used to determine the radiance for the study area.

$${\text{L}}_{{\uplambda }} = \frac{{\left( {{\text{LMAX}}_{{\uplambda }} - {\text{LMIN}}_{{\uplambda }} } \right)}}{{\left( {{\text{QCALMAX}} - {\text{QCALMIN}}} \right)}}{\text{x }}\left( {{\text{DN}} - {\text{QCALMIN}}} \right) + {\text{LMIN}}_{{\uplambda }}$$
(1)

where Lλ is cell value as radiance in \(W/\left( {M^{2} *sr*\mu m} \right)\), \(LMAX_{\lambda }\) is the sensor spectral radiance that is scaled to \(\left( {QCALMAX} \right)\) in \(W/\left( {M^{2} *sr*\mu m} \right)\), \(LMIN_{\lambda }\) is the sensor spectral radiance that is scaled to \(\left( {QCALMIN} \right)\) in [\(W/\left( {M^{2} *sr*\mu m} \right)\)]. \(\left( {QCALMAX} \right)\) is the maximum quantized calibrated pixel value to \(LMAX_{\lambda }\) [DN], \(\left( {QCALMIN} \right)\) is the minimum quantized calibrated pixel value corresponding to \(LMIN_{\lambda }\) [DN] and QCAL is the quantized calibrated pixel value [DN]. The equation above can be observed from header files ETM + and TM datasets from USGS website. The LMIN and LMAX are the spectral radiances for each band at digital numbers (DN) 1 and 255 for Landsat 7 ETM + , 1 and 65,535 for Landsat 8 OLI/TIRS. λ is the wavelength. This was done using ArcGIS 10.6.

Conversion of Spectral Radiance (Lλ) to Kelvin with emissivity value:

$${\text{T}} = { }\frac{{{\text{K}}_{2} }}{{\ln \left( {\frac{{{\text{K}}_{1} {\text{*E}}}}{{{\text{L}}_{{\uplambda }} }}{ } + 1} \right)}}$$
(2)

Therefore, \(k_{1}\) and \(k_{2}\) become a coefficient determined by effective wavelength of a satellite sensor (Table 3).

$$BT = \frac{{K_{2} }}{{\ln \left[ {\left( {{{K_{1} } \mathord{\left/ {\vphantom {{K_{1} } {L_{\lambda } }}} \right. \kern-\nulldelimiterspace} {L_{\lambda } }}} \right) + 1} \right]}}$$
(3)
Table 3 ETM + and TM thermal band calibration constants.
Full size table

2.4.3 Conversion of spectral radiance (Lλ) to kelvin with emissivity value from Landsat 8

ENVI 5.0 software was used for the correction of thermal band 10 in order to remove atmospheric distortions from the thermal infrared data (Table 4).

Table 4 Thermal constant, Band 10.
Full size table

Since temperature is required to be in Degree Celsius (°C) (Tc), results for various temperatures must be converted from Kelvin (K) (TB) to degree Celsius (°C). Conversion of Kelvin to Degree Celsius:

$${\text{T}}_{{\text{C}}} = {\text{T}}_{{\text{B}}} - 273.15$$
(4)

where TB is value at satellite brightness temperature (K) and TC is temperature in Degree Celsius.

2.5 Accuracy assessment

Accuracy assessment of land use classes was carried out for each period (1970–2020) using different ground truthing sampled points from random locations using ENVI and ArcGIS. These samples were overlaid on google earth pro for verification. Hundred (100) samples were generated from each class (Table 2) in the classified images for the accuracy assessment, making a total of five hundred (500) samples. The user and producer’s accuracy assessment were employed along with a confusion matrix that culminates randomized and overall sampled points. The expression below was used to calculate the accuracy assessment:

$${\mathbf{Accuracy}} {\mathbf{Assessment}}\left( {{\mathbf{A}}.{\mathbf{A}}} \right) = { }\left[ {\left( {{{{\text{ASP}}} \mathord{\left/ {\vphantom {{{\text{ASP}}} {{\text{TSP}}}}} \right. \kern-\nulldelimiterspace} {{\text{TSP}}}}} \right) \times 100} \right]$$
(5)

where ASP = Number of sample points that accurately falls on each required feature (ASP = 460). TSP = Number of total sample points generated (TSP = 500). A.A = Accuracy Assessment [(460/500) X 100 = 92%].

Therefore, the present study had 92% accuracy for the given years (1970–2020). A higher value gives vivid accuracy to the raster map and features depicted on the map.

2.6 Data analysis

The present study used GIS tools coupled with existing literature on LUCC and climate variability in the study area for its analysis. Using satellite imagery and existing literature have proven to be vital instruments in assessing historic changes associated with spatiotemporal development of land use systems and microclimatic conditions. Using satellite imagery and existing literature to detect and establish historical changes does not only validate information on environmental changes observed in the study area but provide some strategic directions to inform policies and choices for communities.

2.6.1 Establishing a relationship between various input variables

The average maps of the input variables were stacked, and a random pixel sampling (10%) (n = 2,722,431) was performed to obtain information for each variable. The total number of pixels for the study area are 28, 725, 130. The resulting dataframe (df) was cleaned, normalized and used to calculate the Pearson's correlation between the variables using the Pandas (a python data analysis library). Correlation coefficients are statistics that measure the relation between variables or features of datasets.

2.6.2 Pearson correlation: pandas implementation

Pearson’s correlation coefficient factors a dataset with two features: x and y. Given the premise, the first value x1 from x corresponds to the first value y1 from y, the second value x2 from x to the second value y2 from y, and so on. Again, there are n pairs of corresponding values: (x1, y1), (x2, y2), and so on. Each of these x–y pairs represents a single observation. In addition, it measures the linear relationship between two features and connotes the ratio of the covariance of x and y to the product of their standard deviations. The coefficient is denoted using the letter “r”. It can be expressed using the mathematical function:

$${\text{r}} = \Sigma_{i} \left( {\left( {{\text{x}}_{i} - {\text{mean}}\left( {\text{x}} \right)} \right)\left( {{\text{y}}_{i} - {\text{mean}}\left( {\text{y}} \right)} \right)} \right)\left( {\sqrt {\Sigma_{i} \left( {{\text{x}}_{i} - {\text{mean}}\left( {\text{x}} \right)} \right)^{2} \sqrt {\Sigma_{i} \left( {{\text{y}}_{i} - {\text{mean}}\left( {\text{y}} \right)} \right)^{2} } } } \right)^{ - 1}$$
(6)

From the expression above, i takes on the values 1, 2,…, n. The mean values of x and y are expressed with mean(x) and mean(y). Interpretation of “r” values:

  • “r” can assume any real value in the range −1 ≤ r ≤ 1.

  • “r = 1” connotes a strong positive linear relationship between x and y.

  • “r > 0” indicates positive correlation between x and y.

  • “r = 0” means x and y are independent.

  • “r < 0” indicates negative correlation between x and y.

  • “r =  − 1” shows a strong negative linear relationship between x and y.

In summary, a larger absolute value of r indicates stronger correlation, closer to a linear function. A smaller absolute value of r indicates weaker correlation.

3 Results

3.1 Change detection analysis between 1970 and 2020

Land cover classification from the five classes identified over the past five decades indicate an overall conversion of each land cover type (Tables 5 and 6) based on class total and image differencing results. Results generated for this study present evidence in alteration and declension in LUCC over the past 50 years. The arch land use that increased progressively over the study period was built-up and farmlands/shrubs. In all, between 1970 and 2020 (see Annex I), bare land obtained an area of 342.20 Km2 (39.65%) and lost 75.43 Km2 (−60.35%) to other classes. Built-up obtained an area of 7431.62 Km2 (217.48%) following a 6896.36 Km2 (117.48%) increment gained from other classes. Water bodies obtained an area of 635 Km2 (51.84%) following a 239.48 Km2 (−48.16%) loss to other classes. Farmlands and shrubs obtained an area of 8382.52 Km2 (137.55%) following a 6598.30 Km2 (37.55%) increment whilst forests covered an area of 3572.50 Km2 (40.48%) and lost 16,739.92 Km2 (−59.52%) to other classes (Table 6; Fig. 5).

Table 5 Area coverage for LUCC in Southwestern Ghana (1970–2020)
Full size table
Table 6 Land cover changes between periods (%)
Full size table
Fig. 5
figure 5

LUCC changes over the past five decades (1970–2020) in Southwestern Ghana

Full size image

Considering the overall conversions of LUCC systems (land cover changes between periods) as depicted in Tables 5 and 7, bare land have decreased at a rate of 18.06% (with 0.36% decline per year), built-up increased at a rate of 1288.36% (with 25.77% increase per year), waterbodies declined by 27.39% (with 0.55% decline per year), farmlands and shrubs increased by 369.81% (at a rate of 7.4% increase each year) whilst forests over the study period declined by 82.41% (at a rate of 1.65% each year).

Table 7 Rate and magnitude of change (sq.km) of LUCC in Southwestern Ghana (1970–2020)
Full size table

3.2 Drivers of land use cover change

From a range of socio-cultural, economic, land reforms and institutional/state policies, demographic, technological and biophysical factors, over eight (8) key drivers (proximate and underlying causes) have been highlighted in existing literature as major influences of LUCC in the study area (Fig. 8) (Table 8). Major historical events (i.e. underlying causes of change) and the linkages between the main divers and the most important LUCC transitions are the implications of the changes drawn from findings of previous studies and spatial results generated for this study over the past five decades (Table 8).

Table 8 Major events, key drivers, consequences and transitions of LUCC in Southwestern Ghana (1970–2020)
Full size table

3.3 Temperature analysis

Figure 6 shows the temperature range on average was between 27.78 °C and 20.23 °C in the 1970s. However, the average temperature range for 1980s was between 30.44 °C and 27.78 °C, which could be attributed to biophysical factors (i.e. bushfires/wildfires and prolonged dryness that occurred in the 1980s), which caused significant rise in surface temperatures in the study area. The average temperature range for the 1990s was between 28.88 °C and 25.45 °C. Average temperature range for 2000, 2010 and 2020 were between 30.12 °C and 23.67 °C, 31.66 °C and 24.44 °C, as well as 33.76 °C and 24.54 °C, respectively. Dark red and yellowish areas indicate areas with high or moderately high temperatures (these areas are heavily characterised by built-up) whilst dark blue areas represent areas with low temperature (mostly characterised by forest reserves or deciduous forests) with transient light blue zones.

Fig. 6
figure 6

Temperature variations over the study period (1970–2020) in the study area

Full size image

3.4 Precipitation analysis

Rainfall data (in mm) archived for the study period were acquired from Trading Economics’ website (https://tradingeconomics.com/ghana/precipitation). The precipitation analysis was performed using the Inverse Distance Weighted (IDW) interpolation to depict areas within the study area with high and low average rainfall (Fig. 7).

Fig. 7
figure 7

Changes in rainfall amounts (in mm) received in specific zones over the study period (1970–2020) in the study area

Full size image

The precipitation range was between 50.27 mm and 100.54 mm for the 1970s; that of the 1980s ranged between 38.25 mm and 50.11 mm. However, the precipitation range for the 1990s ranged between 25.85 mm and 70.77 mm. A range of 39.17 mm and 45.21 mm can be observed for the 2000s; 10.70 mm to 39.85 mm is recorded for 2010, and 0.70 mm to 20.88 mm observed in early 2020.

The distribution above presents correlation analysis between various indices and climatic variables. Considering the results generated in relation to “r” shows the impact or degree of evidence identified by the present study as negligible; very weak or weak; least moderate or moderate, and robust or strong correlation. Here,

0.5 \(\ge\) r ≤ 1 connotes robust or strong correlation (evidence of association) between both variables

0.5 ≥ r ≤ 0.6 means least moderate or moderate correlation (evidence of association) exist.

0.2 ≥ r ≤ 0.49 means very weak or weak correlation (evidence of association) exist.

Below 0.2 (r < 0.2) means limited or no evidence of association (negligible).

4 Discussion

4.1 Major influences of LUCC and its implications

The comparison of each class over the study period showed there has been a remarkable change in LUCC of Southwestern Ghana during the last 50 years. Much of the changes is directly correlated to human-induced factors especially the conversion of forest areas to built-up areas and farmlands/shrubs. This therefore affirms the claim (Acheampong et al., 2018; Damnyag et al., 2017; Kusimi, 2008; Mensah et al., 2019) that economic forces are direct stimuli to the dramatic change in land cover or pristine environments. Table 6 illustrates specifically how built-up lands and farmlands/shrubs have increased at the expense of natural vegetation, waterbodies and bare land.

Overall, the natural vegetation (forests) and waterbodies in Southwestern Ghana over the given period reduced by 82% whereas areas covered by waterbodies reduced by 27%. The changes as illustrated in (Fig. 5) can be spatially observed across the length and breadth of the study area. We observed that the changes experienced in land use systems during the last 50 years, as depicted in Table 8 is primarily driven by economic factors which constitute rapid increase in population growth and distribution, agricultural activities, mining and illegal logging, state policies driven towards social interventions and industrialization, boosting exports through foreign exchange, bridging housing deficit gaps and poverty alleviation. The significant change, observed between 1980 and beyond, was mainly as a result of biophysical factors (i.e. prolonged dryness and wildfires) that were experienced in 1983. Various reforms and economic policies-initiated post-famine period till date are driven towards micro- and macro-economic stability as well as enhancing social infrastructure. Technological improvement and the development of new transportation networks which comprise railway networks to link major towns and cities, as well as, new roads for cocoa growing areas, mining and crude oil exploration towns have triggered changes contributing to the dramatic increase in built-up environment during the last 50 years. Presently, most farmers among other indigenes are trained in the use of several technologies, used to enhance productivity of cash crops and other perennial crops grown in the region. Improving soil fertility and adapting to climate-smart agricultural techniques among a host of technological equipment for farming and mining by multinational and small-scale operators have degraded forests and land resources in Southwestern Ghana (Fig. 8).

Fig. 8
figure 8

Drivers and consequences of LUCC in Southwestern Ghana

Full size image

Again, evidence of human-induced factors driving this dramatic change is exacerbated by some natural or biophysical factors emanating from climate change as depicted in Table 8 and (Figs. 6 and 7). In effect, these drivers demonstrate a “causal chain” where one cause (underlying cause) could lead to the other (direct cause). Perhaps, “an effect” could drive “a cause”. For instance, reduction in forest areas through various socio-economic activities could influence prevailing microclimatic conditions (Figs. 6 and 7). Changes in prevailing microclimatic conditions could eventually cause droughts and flooding which could affect farmlands, shrubs, waterbodies and so on. Findings of this study are congruent with the stand points of Acheampong et al. (2018) and Damnyag et al. (2017), who posited those economic and socio-political factors like urban sprawl, mining, state/institutional policies, population growth and distribution as well as expansion in agricultural activities, are major factors driving LUCC in the study area. It is worthy to mention that efforts by the Government of Ghana through the Forestry Commission, International organizations and some Community Based Organizations (CBOs)/NGOs have over the past two to three decades, initiated various schemes to avert the acceleration of deforestation and land degradation. Among such interventions are re-afforestation projects outlined in sustainable forest and land resource management policies like: GYEEDA, planting for food and jobs, REDD + Hotspot initiatives or policies, as well as sensitizing farmers and indigenes on climate-smart agriculture and alternative livelihood sources that do not degrade land and forest. Some green projects or alternative livelihood source initiatives in the study area constitute the Community Resource Management Area (CREMA) and the Additional Livelihood Support Scheme of the Enhancing Natural Forests and Agro-forest Landscapes (ENFAL) projects. These initiatives provide a practical avenue for local communities to economically benefit from managing their resources sustainably. Creating green alternative livelihood sources for local folks in various rural communities across the entire study area could reduce the rate of urbanization and cater for future requisition of processed timber among other factors that cause forest and land degradation. In the quest of achieving SDGs 1, 2, 13 and 15, several livelihood options like tree nurseries, mushroom farming, bee keeping, grass cutter rearing, soap making, rabbit and snail rearing among other sustainable and cost-effective green projects could manage competing interests among relevant stakeholders in the region. LUCC in the study area is tied to several needs of various stakeholders. Hence, it is needful to adhere to policies or initiatives that will meet the needs of interested stakeholders while simultaneously protecting natural resources.

4.2 Various indices (NDVI, NDBI and NDWI), climatic variables and change detection from 1970–2020

4.2.1 NDVI change detection of SW Ghana from 1970–2020

The estimated NDVI range for the 1970s was between −0.96 and 1. The range for 1980s was between −0.97 and 0.79. The 1990s had a range of -0.93 and 0.81; 2000s had a range of −0.85 and 0.75; 2010 ranged between −0.87 and 0.70, and 2020 depicted an NDVI range of −0.90 and 0.64. Figure 9 illustrates a steady decline in vegetative index over the study period. Larger values of NDVI represent forest areas due to higher green biomass of trees and other vegetation. These areas as observed over the study period (1970–2020) constitute mainly forest and wildlife reserves/parks, closed (dense) and open canopies. Decrease in NDVI based on study findings could be attributed to the main drivers highlighted in Table 8, which are primarily, economic, socio-political and biophysical factors. This aligns with the assertions of Acheampong et al., (2018), Damnyag et al. (2017), Kleemann et al. (2017) and Kusimi (2008) that the major factors causing deforestation in the study area are economic and socio-political which constitute population and economic growth, weak governance.

Fig. 9
figure 9

Changes in NDVI over the study period (1970–2020) in Southwestern Ghana

Full size image

Quantification of differences in vegetation in Southwestern Ghana was visualized in image differencing using NDVI for the study periods (1970s, 1980s, 1990s, 2000, 2010 and 2020). Areas marked with violet (Fig. 9) represent a highly negative change, thus, major reduction in vegetation cover as observed in the 1970s and 1980s. Such areas as depicted in Fig. 9 are subdued by the sea or built-up environment. Yellowish and greenish areas indicate areas with moderate and dense vegetation cover, respectively, with an increasing rate of agricultural areas (between 2000 and 2020).

4.2.2 NDBI change detection of SW Ghana from 1970–2020

Figure 10 illustrates changes in NDBI over the study period in Southwestern Ghana. It is observed that NDBI ranged between −0.80 and 0.29 for the 1970s. The 1980s had an NDBI range between −0.77 and 0.37, and −0.75 to 0.49 for the 1990s. Again, the NDBI range for the 2000s was between −0.70 and 0.62. A significant increment was observed in 2010 when NDBI ranged between −0.85 and 0.77; NDBI range for 2020 was between −0.83 and 0.79. There is clear evidence of continuous expansion of settlements over the study period in the study area. Dark red and yellowish areas indicate high presence of built-up environment. Light green and green areas represent areas covered by farmlands and shrubs as well as less dense vegetation. Dark blue areas represent areas covered by forest and wildlife reserves (deciduous and semi-deciduous zones) or water bodies as shown in Fig. 10.

Fig. 10
figure 10

Changes in NDBI over the study period (1970–2020) in Southwestern Ghana

Full size image

4.2.3 NDWI change detection of SW Ghana from 1970–2020

From the illustrations below, the NDWI range for the 1970s was between −0.85 and 1 whilst that of the 1980s was between −0.90 and 0.95. However, the range for the 1990s was between −0.87 and 0.97. A significant change of −0.94 and 0.99 is observed for the 2000s whilst 2010 had a range between −0.96 and 0.99. Finally, 2020 NDWI ranged between −0.98 and 0.99. Figure 11 illustrates changes in water index over the study period. Dark blue areas represent areas covered by the sea while light blue areas are covered by rivers and other waterbodies. Greenish areas are areas covered by natural vegetation, forest reserves/parks whilst light green and yellowish areas are covered by farmlands/shrubs and built-up environment (settlements) as observed during the past 50 years (Fig. 11). No drastic change was observed for NDWI. Decline in waterbodies over the study period as presented in Table 5 could be attributed to the dramatic increase in built-up areas and farmlands. In addition, biophysical factors like increase in surface temperatures (Fig. 6) through evapotranspiration and other processes in Southwestern Ghana, coupled with fluctuations in rainfall patterns (Fig. 7) could partly contribute to the decline in water areas in the study region.

Fig. 11
figure 11

Changes in NDWI over the study period (1970–2020) in Southwestern Ghana

Full size image

4.3 Relationship between various variable inputs

Monitoring ecosystems that are sensitive to climate change ensures a comprehensive understanding of the link between some climatic variables and ecosystems. This insight is critical for future land use planning and management. One of the key objectives of this study was to examine the relationships between NDVI, NDBI, NDWI and some climatic variables (temperature and precipitation). Surprisingly, results indicated NDVI and temperature were positively correlated (r = 0.214) (Table 9). This means both NDVI and LST are increasing steadily which partly agrees with the findings of most studies considering the period (December-harmattan season) the data was captured among other reasons (Guo, 2002; Sun & Kafatos, 2007; Wang, 2016; Wang et al., 2001). Several studies have depicted a negative correlation (Gorgani et al., 2013; Liu et al., 2011; Yue et al., 2007) between NDVI and LST. The reasons for this result in the case of Southwestern Ghana could be attributed to the increase in farmlands/shrubs and grasslands, which does not significantly influence surface temperature. Thus, the impact of grasslands and farmlands/shrubs is marginal unlike forests or natural vegetation. Hence, as farmlands/shrubs (Table 6) (Fig. 5) increased during the study period, LST also increased steadily. More so, the presence of waterbodies in randomly sampled zones, used for the present study could somewhat cause NDVI and LST to be positively correlated as revealed by (Guo, 2002; Wang et al., 2001). However, the study area does not have major or more waterbodies that could affect the results generated since the random points generated were evenly distributed in the study area. Again, further analysis was conducted to remove sampling points that fell within zones of waterbodies to validate findings. Results generated indicated a very weak positive correlation (r = 0.214) was detected by the present study which somewhat validates the presence and increase in farmlands/shrubs and grasslands. Contextually, a very weak positive correlation generated for NDVI and LST eventually indicates very weak evidence of association exists for the two given variables in the study area.

Table 9 Pearson’s correlation coefficient (r) between various indices and climatic variables
Full size table

On the other hand, NDVI and precipitation had a negative correlation (r = −0.11). Here, the decreasing trends in vegetation coverage (forest areas) and increasing trends in farmlands/shrubs and grasslands (Fig. LULCC), had resulted in the changes in rainfall patterns (Table 9) (Figs. 7 and 9). This result corroborates with the initial findings of Changkakati (2019) and the results of Bora and Goswami (2016). They attributed the negative correlation to growing monsoons among a host of other factors. Mitchell and Csillag (2001) in their study concluded that precipitation was a critical variable for vegetation in the mixed prairie ecosystem. Only the general correlation and regression between vegetation index and some climate elements were compared in their study. A thorough study should be conducted to reveal how precipitation is critical, which period of the season is worth noting for precipitation, coupled with which vegetation type is more delicate to precipitation. A negative correlation of (r = −0.11) is however negligible; since there is insufficient  evidence of association to support the relationship between NDVI and precipitation per the results generated for this study.

We observed temperature and NDWI were positively correlated (r = 0.342) (Table 9) (Fig. 12a). The findings in the study agreed with the standpoints in previous studies (Abbam et al., 2018; Aduah & Baffoe, 2013; Logah et al. 2013). They revealed increasing temperatures in the study area could influence NDWI through high evaporation as well as expansion of waterbodies due to high temperatures. This somewhat explains why LST (Fig. 6) is positively correlated to NDWI during the past 50 years. NDWI and precipitation had a negative correlation (r = −0.305). It could also be observed that NDVI and NDWI have a positive correlation (r = 0.540). Both indices per results in (Figs. 9 and 11) show a decreasing trend. Results (r = −0.305) proved a weak correlation exist between temperature and NDWI variables in the study domain. Hence, weak evidence of association exists for the given variables.

Fig. 12
figure 12

Pearson’s correlation coefficient (r) between various indices and climatic variables. a Correlation between NDWI, NDVI, NDBI and temperature. b Correlation between NDWI, NDBI, NDVI and precipitation

Full size image

Results revealed NDBI is positively correlated to LST (r = 0.265; very weak correlation/evidence of association), NDWI (r = 0.818; the study detected a strong correlation or evidence of association between NDBI and NDWI), NDVI (r = 0.165) and negatively correlated to precipitation (r = −0.189). Positive correlation between NDBI and LST shows that, drastic increment in built-up (Fig. 10) has resulted to an increase in surface temperatures (Fig. 6). Also, the positive correlation between NDBI and NDVI shows an increase in built-up (settlements) (Fig. 10) and farmlands/shrubs and grasslands (Fig. 5) (Table 5). The negative correlation (Table 9) between NDBI (Fig. 10) and precipitation shows a significant increase in built-up environment has resulted to a decline in rainfall amounts received (changes in rainfall patterns) (Fig. 7). Our findings corroborate with the assertions of (Abbam et al., 2018; Aduah & Baffoe, 2013; Aduah et al., 2012; Mensah et al., 2019) on built-up’s influence on LST and precipitation. Precipitation and LST per the distribution above (Table 9) (Fig. 12b) show an inverse correlation (r = −0.155). In effect, an increase in LST (Fig. 6) has resulted in the fluctuations in rainfall patterns (Fig. 7) in the study area.

LUCC associated with biophysical variables such as vegetation index and climatic data provides the type of information required for categorization and etching of habitats. Ghana has experienced severe loss and fragmentation of natural vegetation over the last 50 years (Forestry Commission, 2015). Findings based on major influences have accounted for the dramatic loss in vegetation, and a significant increase in built-up and farmlands/shrubs. Among the two factors stated above, human-induced factors mainly driven by economic and socio-political factors as presented in Table 8 (Fig. 8) have influenced these unprecedented changes during the last 50 years in Southwestern Ghana. The potential human drivers in the study area have been explored in several local studies in different scopes. Some studies (Abbam et al., 2018; Acheampong et al., 2018; Aduah & Baffoe, 2013; Aduah et al., 2012; Damnyag et al., 2017; Kusimi, 2008; Mensah et al., 2019) confirmed that decreasing trends in vegetative coverage and land degradation in the study area were attributed to urban sprawl, industrialization, expansion in farmlands, mining and illegal logging of timber, population growth and distribution, as well as weak governance. Findings presented in this study indicated built-up (i.e. increase in human settlements, population growth and distribution, and infrastructure among other factors) and farmlands/shrubs (i.e. expansion in agricultural activities and so on) are the two major LUCC classes, driving major changes in the region. Importantly, the need to regulate various needs, tied to the use of natural resources in the region must be prioritized in our quest to achieve the Millennium Sustainable Development Goals. The study area is vital to the growth and development of Ghana’s economy, hence, competing interests among relevant stakeholders in the use of natural resources must be major area of concern. Advancement in technology, observed in transportation networks and other infrastructure, as well as agricultural activities, mainly to boost exports and productivity, has amplified the conversion of natural vegetation zones into farmlands and built-up environments.

The present study therefore recommends the need for remote sensing in contemporary forest and sustainable land management. The study showed more land cover will be targeted for conversion as farmers expand their farmlands and adapt to new technological approaches to improve productivity, considering the increasing rate of producer price index for cocoa and other cash crops on the international market. LUCC known as a mesoscale element and a contributing factor to climate change, makes it a topical issue for effective management, planning and inventory mapping. Land conversion scenarios highlighted in the study are very useful for stakeholders to better understand the environment. Using remote sensing and GIS techniques to monitor environmental change gives a comprehensive view into the general ambience of Southwestern Ghana. The maps and statistics generated can be applied to assess the impacts of the land use changes on the local hydrology and provide a better basis for future land use planning.

5 Conclusion

The study utilised remote sensing data to assess the spatiotemporal development of land use systems, drivers and climate variability in Southwestern Ghana over the last 50 years. An attempt was made to determine the historical changes in NDVI, NDWI, NDBI, LST and precipitation. Through this study, we found that Southwestern Ghana had experienced substantial LUCC, with a decline in forested areas, waterbodies and bare land. This research further revealed a significant increase in built-up areas and farmlands/shrubs. Key variables that influenced prevailing microclimatic conditions can be associated with LUCC, considering the nature of proximate and underlying drivers, observed in the study area during the last five decades.

Here, the study established a relationship between the variable inputs mentioned above using Pearson’s correlation coefficient. Monitoring ecosystems that are sensitive to climate change ensures a comprehensive understanding of the link between some climatic variables and ecosystems. Findings revealed surface temperatures had increased steadily over the study period coupled with fluctuations in rainfall. The study depicted a positive correlation between NDVI and LST. NDBI, on the other hand, correlated positively with NDVI, NDWI and LST. In contrast, NDVI, NDBI, NDWI and LST correlated negatively with precipitation. Findings based on image differencing and spatial analysis over the past 50 years were validated using previous studies related to LUCC studies in Ghana and other parts of the world. Therefore, it can be concluded that the substantial change in LUCC in the study domain are immensely driven by human-induced factors which constitute economic and socio-political factors.

We demonstrated that the synergies of satellite imagery and GIS technologies serve as powerful tools for mapping and detecting changes in LULC. This research conducted in Southwestern Ghana using these modern technologies in conjunction with existing literature showed a remarkable change in land cover. We advocate the need for strict execution of appropriate land use policies to sustain food production, coupled with other competing interests in the utilization of natural resources in the study area within this era of changing climate, population increase and economic growth. The study serves a seminal guide to land use developers and institutors for effective and sustainable use of natural resources. We therefore propose an integrative rural–urban growth management strategy that culminates spatial planning and environmental resource governance to avert the negative consequences on the natural environment of unfettered socio-economic development. The spatiotemporal analysis and characterization of LULC transformation, presented in this study have implications for both theory and policy. The present study’s approach and underlying theories in determining the spatiotemporal development of land use systems, influences and climate variability could be extended to other areas to bridge the knowledge gap in LUCC related studies. Again, the main drivers of LUCC as presented in this study could be further analysed using multi-criteria decision-making (MCDA) tools and determining the contribution rates of various indices towards the changes in surface temperature and precipitation to substantiate the assertions or hypothesis given in this and other related studies conducted in the study area.