Potential impact of future land use/cover dynamics on the habitat quality of the Yayo Coffee Forest Biosphere Reserve, southwestern Ethiopia

Abstract

Human activities, including agricultural expansion, urbanization, and industrial advancement, have led to land use/cover change (LULCC). These changes have had negative consequences, such as the loss of species and the degradation of forest areas. The Yayo Coffee Forest Biosphere Reserve (YCFBR) is undergoing changes due to forest fires and encroachment from coffee plantations, which are predicted to increase in the future. The objective of this study was to simulate the dynamics of LULCC and its impact on habitat quality in the area over the next three decades. The study used classified land cover (LC) maps of 1992, 2022, and 2052 (predicted), as well as factor and constraint maps of the area. In addition, household surveys, key informant interviews (KII), and focus group discussions (FGD) were conducted to understand the factors influencing changes in habitat quality. Habitat quality was simulated under three different future scenarios (Business as Usual; S1), conservation scenario (S2), and development scenario (S3) using CA-Markov, Future Land Use Simulation (FLUS), and Integrated Valuation of Environmental Services Tradeoffs (InVEST) models. The result showed a 28% increase in open forest and a 35% increase in agricultural area under S1, with a minimal increase in built-up area. The medium-level habitat quality increased from 11.3% in 1992 to 27.8% in 2052 S3. The highest average haitat quality value was 0.62 in S2, while the lowest was 0.51 in S3. If S3 is the case, this value is expected to fall below 40%, indicating the lowest possible level of habitat quality in the future. The model also predicts a 7% increase in high forest areas in S2, indicating the possibility of an alternative path to save high forest loss in the area. The main causes of habitat quality deterioration include population growth, agricultural land expansion, a lack of diverse incomes, land tenure issues, and settlement expansion. Based on the findings, S2 appears to be the best future scenario for maintaining habitat quality. This study provides useful information that will help planners and decision-makers effectively prepare future conservation strategies.

Keywords:

• Biosphere Reserve
• scenario
• habitat quality
• InVEST model
• Ethiopia

1. Introduction

The increase in urbanization and industrialization since the 1950s has led to a significant rise in human activities occurring within natural ecosystems. This has resulted in major adverse effects such as the decline of species diversity, the fragmentation of habitats, and the deterioration of ecosystems. These outcomes present a danger to the welfare and stability of humans. Habitat quality is the ability of ecosystems to offer appropriate conditions for individuals or populations, which is a vital indicator of biodiversity. It is an indication of the general well-being and stability of an ecosystem (Gaglio et al. 2017) and highlights the importance of protecting biodiversity and maintaining a healthy and flourishing ecosystem (Mengist et al. 2021; Zhu et al. 2022).

Changes in LULC happen globally and are influenced by a range of factors including the environment, socioeconomic conditions, political dynamics, and regulations (Dohong et al. 2017; Kleemann et al. 2017; Walcott 2019; Fida et al. 2023). This has an impact on the original condition of natural habitats and converts them into ecosystems that are predominantly influenced and regulated by human activities, leading to the fragmentation and separation of habitats. As the land is increasingly disrupted and changed, the state of the habitat deteriorates, putting people’s well-being and overall quality of life at risk (Sharma et al. 2020; Han et al. 2021). The construction of infrastructure has reformed the physical characteristics of the environment, leading to various environmental issues encountered by countries across the globe (Zhang et al. 2020a). The increase in human settlements and population, along with the expansion of agricultural land and road networks, played a role in the increase of degraded and severely degraded habitats in Biosphere Reserves (BRs) (Coughlan et al. 2017).

Most studies on land-use change and habitat quality have concentrated on their past and present states, but there is a lack of research on studies conducted for the purpose of monitoring over extended periods, simulating future scenarios, and predicting trends in habitat quality. The majority of these studies have concentrated on examining only one land-use scenario and have not taken into account the possibility of multiple scenarios or the diverse array of future regional development plans (Shahumyan et al., 2014; Boron et al., 2016; Tang et al., 2016; Tang et al., 2021; Tyagi et al., 2022; Ji et al., 2023).

Several studies have been conducted to assess the spatiotemporal pattern of habitat quality (Wu et al. 2015; Huang et al. 2020; Yang et al. 2021). Other studies have examined the impact of land consolidation on habitat quality (Zhong and Wang 2017; Bai et al. 2019; Zhang et al. 2020; Yang et al. 2021). Additionally, there has been research on the relationship between habitat quality and landscape patterns (Wu et al. 2015; Liu et al. 2019). Lastly, studies have explored the trade-off between habitat quality and other ecosystem services (Güneralp and Seto 2013; Guillem et al. 2015; Asadolahi et al. 2018). Regarding areas of study, there are numerous studies that concentrate on urban areas, watersheds, nature reserves, etc. (Güneralp and Seto 2013; Aneseyee et al. 2020). While previous research has focused on evaluating how past and present land use (LU) changes affect the quality of habitats, there have been limited efforts to predict the future distribution of habitat quality based on different LU scenarios (Lei et al. 2017). At the same time, there was limited attention given to areas that are environmentally sensitive protected areas and BR; particularly, investigations into how the LULCCs affect the habitat quality, and the potential future scenarios have not been extensively modeled (Kalacska et al. 2017).

The methods used mainly involve the InVEST model (Aneseyee et al. 2020; Wang et al. 2021; Zhang et al. 2022), MaxEnt model (Wu et al. 2016), FLUS model (Green et al. 2021), and the combination of grid evaluation and landscape patterns (Ao et al. 2022). The InVEST model applies to different regional scales, showing better ecological process integration and good spatial display effects (Li et al. 2021), and its habitat quality module can quickly assess the impact of different threats and land use types on biodiversity (Xu et al. 2019). Studies have shown that this model effectively assesses biodiversity and habitat quality (Hack et al. 2020; Leek et al. 2020; Wang et al. 2022). Moreover, the CA-Markov model, FLUS model, and InVEST model have achieved good evaluation findings in their individual research domains, but they have rarely been employed together in a regional habitat quality scenario analysis. We attempted to combine the CA-Markov model, FLUS, and an InVEST model to study future habitat quality change under different development scenarios, and we chose a typical area to test the effectiveness of this combination in multi-scenario habitat quality prediction (Tang et al. 2021).

In Ethiopia, there is a continuous transformation occurring in the patterns of LULC dynamics. Rural areas are currently mainly focused on agricultural activities, which are expanding into natural forested areas. Around 66.2% of the forests in Munessa-Shashemene, located in Oromia, Ethiopia, have been transformed into agricultural land. Additionally, more than half (60%) of the montane forests in southwest Ethiopia have been lost as a result of clearing and deforestation. (Woldetsadik et al. 2003; Munro et al. 2008; Gashaw et al. 2017). Southwestern Ethiopia is part of the eastern Afromontane region, known for its high biodiversity. This region is anticipated to play a significant role in Ethiopia’s economic growth and development (Tilahun et al. 2017). However, there has been a decrease in ecological resources in the area in recent times. This decline can be attributed to various factors such as deforestation, migration, population growth, the intrusion of invasive species, expansion of agriculture, changes in culture, and the absence of land ownership in the area (Tadesse et al. 2014; Getahun et al. 2013). The degradation of habitat quality in these areas is being caused by the expansion of human-modified landscapes like settlements and agricultural lands (Mengist et al. 2021). Similarly, the YCFBR in southwestern Ethiopia is experiencing negative impacts from human activities. These include forest fires, rapid interventions, and the expansion of coffee plantations in the BR, which pose a threat to the rich diversity of plants and genetic resources in the area. Additionally, the establishment of projects such as the Geba Dam, Yayo coal mining, and Yayo fertilizer industries, along with the construction of extensive roads and power transmission lines in the vicinity of the BR, have posed a significant risk to the environmental condition of the area (Huluka and Wondimagegnhu 2019; Liu et al. 2014; Abera et al. 2021; Daba and You 2022).

Despite the significant efforts and positive impacts of various initiatives within the YCFBR, there is still a lack of comprehensive data on the dynamics of LULC and their potential impacts on local habitat quality. Future changes in LULC could impact habitat quality; however, there has not been a thorough examination that takes into account the simulation-based collective analysis of LULC dynamics. Therefore, it is important to simulate LULC dynamics and their impacts on habitat quality to ensure the sustainable conservation of biodiversity. The main objective of this study was to investigate the following questions: (1) How will LULC change in 2022–2052 under different future scenarios? (2) Which specific LULC classes will experience conversions to other classes between 2022 and 2052? (3) How will the changes in LULC affect future habitat quality under various potential scenarios in 2052? This study is important to provide existing evidence on historical patterns of LULC changes and their impact on habitat quality, thereby assisting ecologists and decision-makers in formulating an LU plan with a focus on sustainable land management. In addition, it could help ecologists and decision-makers concerned with biodiversity conservation develop an LU plan. The main objective of the research was to simulate future LULC dynamics in the YCFBR and assess how these changes would affect habitat quality under different scenarios for the next three decades.

2. Materials and methods

2.1. The study area

YCFBR is located in south-west Ethiopia in the Oromia National Regional State, in the Ilu Aba Bora and Buno-Bedele zones, 510 km Southwest of Addis Ababa, Ethiopia. It is one of the five BRs in Ethiopia and is part of the Eastern Afromontane Biodiversity Hotspot in the country. It is also one of the largest and most important forest areas with wild populations of coffee arabica diversity. It was registered as a UNESCOBR in June 2010 and has three management zones: core (277.33 km²), buffer (215.52 km²), and transitional (1177.36 km²). The core zone is fully protected, while the buffer zone allows anthropogenic entrance if consistent with reserve objectives. The transitional zone includes agricultural land, wetlands, grasslands, settlements, and forest fragments (Gole et al. 2009). The geographical location of the area lies within 8° 10′ 0"to 8° 40′ 0"N and 35° 30′ 0" to 36° 0′ 0"E (Figure 1).

Figure 1. Map of study area.

Based on the NASA Power Meteorological data sets for 2021–2022, YCFBR experiences annual rainfall ranging from 9210.74 mm to 686.58 mm, with high rainfall between June and August and low rainfall in January and February (Figure 2). The area experiences the warmest temperatures from February to April and the coldest from August to October, with monthly maximum and minimum temperature ranging from 30.05 °C to 12.34 °C, respectively (Figure 2).

Figure 2. Monthly temperature and rainfall of YCFBR.

Source: (NASA/POWER; http://power.larc.nasa.gov; Fida et al. 2023).

2.2. Data type and sources

In this study, we used various sources of data to predict future impacts of LULCC on habitat quality. Frist the two LULC maps (1992 and 2022) were mapped into six LULC classes using the ERDAS imagine 2015 from the Landsat imagery (Table 1(A,B)). The common image pre-processing techniques such as geometric and radiometric correction algorithms. A supervised image classification method using maximum likelihood algorithms was used to create LC maps consisting of six different LULC types. Secondly, future LULC maps (2052) was simulated in LCM in terrSet software using LULC maps (1992 and 2022) and factor and constraints maps (Table 1(B) and Figure 3) as inputs for simulation. The, thirdly, future LU demand for each of the designed future scenarios of 2052 following Mengist et al. (2021) was calculated taking the simulated LC maps of 2052 with driving data (Figure 3). Fourth, the threat Information table, threat raster maps, the half-saturation constant, and Suitability/Sensitivity information of the habitat types to each threat and with LC maps, etc., (Table 4(A,B)) were used for the InVEST habitat quality modeling (Table 3).

Figure 3. Factor and constraints maps of YCFBR.

Table 1. (A,B) Descriptions of Landsat images used for the study.

A
Year	Satellite Image	Path/Row	Resolution (m)	Source
1992	Landsat 5 TM	170/54	30 m	USGS earth explorer
2022	Landsat 8 OLI	170/54	30m	USGS earth explorer
B
Input Layer		Source
Terrain factors	Elevation	DEM from USGS
	Slope	Estimated from DEM
	Aspect	calculated using from DEM in QGIS.
Accessibility measures	Distance from Road network	Digitized on the satellite image
	Distance from river	Digitized on the satellite image
	Distance from settlement areas	derived from the classified LULC maps
LC map (1992)		Produced by classification in ERDASS imagine 2015
LC map (2022)		Produced by classification in ERDASS imagine 2015
LC map (2052)		simulated using CA-Marcov in TerrSet

Table 2. (A,B) Scenario weighting matrix and descriptions of LU plan in each scenario.

A
Scenario Type	Scenario Weight Matrix wn Value
S1	________________
S2	(0.75, 0.95, 1, o.5, 0.6, 0.7)
S3	(1, 1, 0.9, 0.9, 1, 1,)
B
Scenario	Description	LU plan
S1	The LU pattern is only influenced by the historical transitional rules and is simulated without restrictions.	The expansion of areas into BR areas would continue until 2052 and the BR protection policy is not effective
S2	The scenario focuses on the balance between nature conservation and sustainable development for and with communities according to UNESCO policy. BR management should address the needs of stakeholders in meeting conservation, development, and logistical support needs. In this case, due to the tightened strict regulations, BR has been effectively protected and increased implementation of the protection policy,	Effective implementation of the BR protection policy. An increase in forest area compared to the baseline is due to a tree planting campaign in the country.
S3	Due to the lack of an effective LU control mechanism, population growth and technological advances, coffee plantations are expanding rapidly. Another reason is the expansion of farmland. Expansion of built-up area in urban and rural residential area. Extensive narrowing of traffic routes	The increase in coffee plantations, cultivated areas, built-up areas and road facilities increased compared to the baseline.

Table 3. Input data for the InVEST model.

Input data	Description-
LULC maps Baseline LC map, 1992, Current LC map, 2022 Future LC map, 2052	A standard GIS raster dataset with a numeric LULC code for each cell. The LULC grid should include the area of ​​interest plus a buffer the width of the greatest maximum threat distance. The raster should not contain any other data. The LULC codes must match the Lc (LC) type sensitivity codes to each threat.
Threat data	A CSV table of all threats had to be included in the model. the table provides information about the relative importance or weight of each threat and its impact on the space as a whole. each row is a source of degradation. each column contains a different attribute of each source of degradation and must be labeled threat, max-dist, weight, and deterioration (Table 4 B).
Threat raster	GIS raster files were created detailing the distribution and intensity of each individual threat, showing how each threat affects the habitat. However, the techniques used for each threat grid varied depending on the data type (threat grid data for soil erosion differs from that for pollution threat grids). Therefore, five threat grid datasets were created (Table 4(A))
Habitat types and sensitivity of each habitat to threats	A CSV table of LULC types provides information on whether a habitat has been identified (absence/presence of habitat) and their specific sensitivity to each threat (Table 4(B)). Sensitivity scores range from 0 to 1, with 0 representing no sensitivity to a threat and 1 representing the greatest sensitivity to a threat (Polasky et al. 2011). Sensitivity values ​​are determined from expert knowledge using the APH method
Half saturation constant (k)	The scaling parameter (or constant) of 0.5 was the default value for the InVEST model. The InVEST model uses a half-saturation curve to convert habitat degradation scores to habitat quality scores (Sharp et al. 2016). It is determined as the inverse relationship between the degradation rating and the habitat quality rating. It helps in visualizing the heterogeneity of quality across the landscape.
Accessibility	A polygon shape file containing data about protected areas that represent relative barriers to threats. Conservation areas and protected areas were considered sites with minimum accessibility and were given threat level 0, while polygons with maximum accessibility were given threat level 1 (Polasky et al., 2011; Tallis et al., 2013; Himlal et al., 2014).

Table 4. (A,B) YCFBR habitat suitability and its relative sensitivity to different threat sources.

(A)
		Threat Source
Land-Use Type	Habitat Suitability	Cultivated Land	Construction Land	National Roads	Provincial Roads	Forest Coffee
High Forest	1	0.5	0.1	0.1	0.3	0.9
Open Forest	0	0.4	0.1	0.1	0.2	0.8
Grazing Land	0	0.3	0.1	0.2	0.4	0.7
Agricultural Land	0	0.3	0.1	0.2	0.3	0.7
Built-up Area	0	0.1	0.1	0.1	0.1	0.3
Water bodies	0	0.2	0.2	0.4	0.6	0.8
(B) Threat factors and weight in the study area
Threat Type	Max distance	Weight	Decay
Built up area	10	1	exponential
Coffee Forest	2	0.4	exponential
Transportation land	3	0.6	linear
Mining area	2	0.8	exponential
Agricultural land	2	0.7	exponential

2.3. Methodology for simulating habita quality change

The methodology in this document consists of six steps (Figure 4). The first step was to classify LULC (1992 and 2022) in ERDAS imagine 2015. The second step was to simulate future LU change in 2052 using the Markov chain in LCM in terrSet software. The third step is to develop future LU scenarios (Table 2(A,B)). According to the Policy and Legal Framework for Biospheres in Ethiopia (César and Ekbom 2013). we have defined three scenarios namely business as Usual (S1), Conservation Scenario (S2) and Development Scenarios (S3) (Table 2B). The fourth step is to Calculating the future LU needs using the FLUS model under the three scenarios. Fifth step is to assess and compare the spatial-temporal distribution of habitat quality in the three scenarios using the InVEST model.

Figure 4. The research flow diagram.

2.3.1. Simulating future LULC change

The CA Markov model was employed in order to simulate future LULC changes (Fida et al. 2023). It is hybrid model, which combines both the Markov chain and cellular automata (CA) techniques. This model is employed to forecast the upcoming LULCC in the year 2052 using LCM in TerrSet software. This model provides several benefits including its capacity for dynamic simulation, effectiveness in situations with limited data, and ease of calibration. Additionally, it enables simulations of various LC and complex patterns (Memarian et al. 2012; Hyandye and Martz 2017; Subedi et al. 2013; Singh et al. 2017; Risma et al. 2019; Atullley et al. 2022).

2.3.2. Land use scenario design

Scenario analysis aims to describe and analyze the various development possibilities and inform policy formulation by comparing the status of different development scenarios (Gao et al. 2021). We summarize the policy orientation of national documents (MoARD, and World Bank 2007; Gebeyehu et al. 2017), and combined with existing studies, set three weighting values for different land use change (LUC) scenarios (Table 2A): (1) The Business as usual (S1) is a continuation of the law of LUC from 1992 to 20222 and does not change the conversion rules in the land use category. (2) The Conservation scenario (S2) is based on the biodiversity conservation policy, which focuses on balancing both nature conservation and sustainable development. (3) The development scenario (S3) is simulated by increasing coffee plantations, cultivated areas, built-up areas and road facilities (Table 2B).

2.3.3. Calculating future LU demand

The future LU needs under the three possible scenarios of 2052, were calculated using the FLUS model. The FLUS model consists of the artificial neural network (ANN) and the adaptive inertial competition mechanism. The ANN is effective in identifying the relationship between natural, social, and economic elements and LU change. The adaptive inertial competition mechanism is useful to overcome the uncertainty and complexity of mutual transformation, which can solve the complexity of local transformation and parameter determination in traditional cellular automata (Sfa et al. 2020). Using different factors such as terrain, accessibility, and simulated LC cover maps of 2052, (Table 1(B) and Figure 3) the LULC demand for each scenario in 2052 is calculated (Table 5).

Table 5. Area changes according LULC type.

Area in Km^2 (percentage change)
	2022	2052
LC type	Baseline	S1	S2	S3
Open Forest	219.03	279.58 (+28%)	226.24 (+3%)	290.73 (+33%)
Built-up Area	119.33	135 (+14%)	165.55 (+39%)	187 (+57%)
Agriculture	344.66	463.58 (+35%)	391.11 (+13%)	481.87 (+40%)
Water bodies	7.07	6.46 (–9%)	7.35 (–4%)	4.87 (–31%)
Grazing Land	132.73	96.74 (–27%)	117.09 (–12%)	112.80 (–15%)
High Forest	836.88	692.39 (–17%)	752.34 (–10%)	596.25 (–29%)

2.3.4. Habitat quality assessment

In this study, the habitat quality module of the InVEST model was used to assess the habitat quality. The assessment by this module was based on the influences of human activities on the ecological environment. It was assumed that the more frequent the human activities, the worse the habitat quality in this area. In contrast, the better the natural environmental conditions and the lower the influence of threat factors, the higher the habitat quality was assumed to be (Trisurat et al., 2019; Yang et al., 2018). The model combines the sensitivity of different LU types to threat factors and the intensity of external threats, acquires the habitat degradation degree by computation, and finally calculates habitat quality (Fu et al., 2018; Vermeij, 1995; Xu et al., 2019b). The formula used to calculate the habitat quality index is as follows: (1) $$H_J\int 1-\left(\frac{D_X^2}{D_{X+K^2}^2}⟩\right.⌉$$(1) where Qxj is the habitat quality of grid x in LU type j; Dxj is the habitat degradation degree, which represents the habitat degradation degree in grid x for LU type j; Hj is the habitat adaptability of grid x for LU type j; and k is a half-saturation constant. The habitat quality value ranges between 0 and 1; the higher the value, the higher the habitat quality. The formula used to calculate the habitat degradation degree was as follows: (2) $$D_{\mathit{\text{XJ}}}={\sum }_{r=1}^r{\sum }_{r=1}^y\left.\left(\frac{{\omega }_{\gamma }}{{\sum }_{r=1}^r{\omega }_{\gamma }}\right.\right)$$(2) (3) $$i_{\mathit{\text{rxy}}}=\left.\left(\frac{d_{\mathit{\text{xy}}}}{d_{r \mathrm{\text{max}}}}\right.\right)\mathrm{i}\mathrm{f} \mathrm{\text{linear}},\mathit{\text{or }}{i}_{\mathit{\text{xy}}}=\mathrm{\text{exp}}\hspace{0.17em}[-\left(\frac{2.99}{d_{r \mathrm{\text{max}}}}\right)d_{\mathit{\text{xy}}}]\mathrm{i}\mathrm{f} \mathrm{\text{exponential}}$$(3)

where ωr is the weight of different threat factors; ry is the intensity of the threat factor; ßx is the anti-interference level of habitat; Sjr is the relative degree of sensitivity of different habitats to different threat factors; r is the habitat threat factor; y is the grid in the threat factor r; dxy is the distance between grid x and grid y; and drmax is the scope of influence of the threat factor r. The habitat degradation degree varies between 0 and 1; the higher the value, the higher the habitat degradation.

The spatio-temporal distribution of future habitat quality within the three future LU scenarios was evaluated by utilizing the InVEST Habitat Quality Model. An open-source Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model of habitat quality, version 3.6.0 (https://www.naturalcapitalproject.org/invest/), was developed at Stanford University (Sharp et al. 2018). The inputs for InVEST model is mentioned in the following Table 3.

2.4. Assessment of driving factors for habitat quality degradation

To assess the driving factors of habitat quality degradation in the area over time, we conducted a questionnaire survey (household heads, n = 84) with 6 FGDs and 48 KIIs in June 2023. The questionnaires were developed based on literature and expert discussions. By using the methods, we were able to cross-reference and confirm the answers, which ultimately increased the validity and consistency of the results. Respondents were asked about the current status of BR habitat quality in the area. They were asked for their views on the main factors of habitat quality degradation listed in Table 9 in pre-listed questions by experts and knowledgeable local elders based on the actual situation of YCFBR.

Using a systematic random sampling method from 1024 household lists from the three kebeles (the smallest administrative unit in Ethiopia), 84 heads of households were selected for an interview. These were specifically chosen for their proximity to the forest area, where biodiversity protection is most urgent. Furthermore, the purposive expert sampling technique was used to select 48 key informants and 6 FGDs with 6–9 members from whom detailed information was subsequently obtained. The people selected for KII and FGD were specifically selected based on their position and knowledge within the BR. The sample size (n) for the questionnaire survey was determined using the following formula from Yamane (1967) at a 95% confidence level. (4) $$\mathrm{\text{The}} \mathrm{\text{sample}} \mathrm{\text{size}}\left(n\right)=\frac{N}{1+N\left(e^2\right)}$$(4)

Where n is the sample size, N is the population size, and e is the desired level of precision.

n = 523/1+523 (0.14) 2 = ≈ 29 from selected keele 1

n = 229/1+229(0.14)2 = ≈ 27; from Keele 2

N = 272/1+272 (0.14) 2 = ≈ 28; from keele 3, and then a total of 84 household heads were interviewed.

Both qualitative and quantitative analysis methods were used to analyze the data in this study. All the data collected was checked, refined, and specifically examined. The data were analyzed using Relative Important Indices (RII) analysis in the Microsoft Excel program.

2.5. Model accuracy test

The model validation process is the accuracy assessment of pre-diction and contrast made between the predicted and observed LC maps. In order to validate the CA-Markov model prediction, the simulated land use for 2020 was compared with the classified land use map for the same year (Nadoushan et al. 2015; Chen and Nuo 2013), using the map comparison kit (MCK) software to find Kappa values. Magnitude and location are the two important issues in land change modeling and properly validating the model for predicting future changes will require effective validation of location and magnitude (Noszczyk 2018). As recommended by Pontius (2000), three indicators were used to validate model performance; Kappa for location (K location), Kapa for quantity (Kno), and Overall Kappa. Simulations are excellent when the value of indicators is equal to 1 and unsatisfactory when the indicators are equal to 0 (Singh et al. 2015). According to Eastman, 0.80 is an acceptable accuracy rate to make reasonable future predictions (Eastman 2006). Based on the trend of land-use change from 1992 to 2022, after predicting the quantity and scale of land use in 2022 using the CA Markov in TerrSet software, this study used the FLUS model to simulate the land-use situation in the year 2052 (Figure 5). The simulation results were compared with real land use in 2022 and then the overall accuracy and kappa coefficient were used to verify the accuracy.

Figure 5. Comparison between the classified and simulated LULC map of 2022.

3. Result

3.1. Model validation and simulation performance assessments

Based on the trend of land-use change from 1992 to 2022, after predicting the quantity and scale of land use in 2022 using the TerrSet software, the simulation results were compared with real land use in 2022 (Figure 5) and then the overall accuracy and kappa coefficient were used to verify the accuracy. The closer the overall accuracy and kappa coefficient are to 1, the more accurately a model can simulate the spatial change of regional land use. When the kappa coefficient is greater than 0.8, the model achieves better simulation accuracy with statistical significance (Wang, Tang, et al. 2020). In this study, the overall accuracy of the model was 0.867, and the kappa coefficient was 0.893, indicating that the CA-Markov model was suitable for simulating land-use change in the YCFBR.

3.2. Future LULC dynamics under different scenarios

In scenario S1, the increase in built-up area was minimal compared to the S2 and S3, but more significant than in the baseline situation. The level of conservation of high forest in S2 is quite worthy, as shown by the fact that the area of ​​open forest devoted to coffee plantation would decrease by 25% compared to S1 (Table 5). According to Table 7, there is an increase in agricultural area of ​​around 40% in the S3 basis. The FLUS scenario simulation results on spatial changes in LULC in 2052 under the three scenarios (S1, S2 and S3) and in the reference year 2022 are shown in Figures 6 and 7.

Figure 6. LULC map of 2052 under the three scenarios (S1, S2 & S3).

Figure 7. LU types trends in the three scenarios of 2052 in YCFBR.

Table 6. Habitat quality level and change rates in KM² and percentage share from 1992 to 2052 (S1, S2 & S2).

Habitat Quality grade	1992		2022		2052
					s1		S2		S3
	Km^2	%	Km^2	%	Km^2	%	Km^2	%	Km^2	%
Poor	97.0	5.8	157.0	9.5	268.0	16.2	162.0	9.8	279.0	16.8
Medium	188.0	11.3	337.0	20.3	323.0	19.5	346.0	20.9	461.0	27.8
Good	357.0	21.5	328.0	19.8	376.0	22.7	399.0	24.1	323.0	19.5
Excellent	1017.0	61.3	837.0	50.5	692.0	41.7	752.0	45.3	596.0	35.9

Table 7. YCFBR habitat quality level during 1992 to 2052 (S1, S2 & S2).

Habitat quality grade	Year
	1992	2022	2052
	Baseline	Current	S1	S2	S3
Poor	0.28	0.35	0.40	0.43	0.45
Moderate	0.79	0.60	0.43	0.48	0.52
Good	0.97	0.84	0.51	0.67	0.46
Excellent	0.98	0.90	0.80	0.88	0.61
Average	0.76	0.67	0.54	0.62	0.51

In addition, the LULC map of S1 shows two major changes in LU classes, namely the increase in open forest area by a maximum of 28% and the expansion of agricultural area by up to 35% (Table 5). These increments occur primarily through the deforestation of high forests and grazing lands. This extension will reach the edge of the BR (Figures 6 and 8). In the S2, there is an estimated increase in built-up area of ​​39% within the area in which settlement expansion is permitted, i.e. the areas in BR transition zone. However, expansion into the buffer and core zones is illegal. Much of this increase comes primarily from grazing lands and agricultural land. In summary, due to the strict implementation of conservation policies, as shown in Table 5, the S2 model predicts a 7% increase in areas of high forest vegetation, while open forest areas will decrease compared to the situation in S1.

Figure 8. Spatial distribution of habitat quality in baseline, 2022 and under each scenario in 2052.

3.3. Future habitat quality under the three LU scenarios

In ArcGIS software, we used natural breaks to classify habitat quality into four classes: poor (0–0.4), moderate (0.4–0.6), good (0.6–0.8) and excellent (0.8–1). We then determined the proportion of habitat area for each quality level from 1992 to 2022 and projected it until 2052 under the three scenarios (Table 7). Figure 8 shows the base maps and the current and future (in three scenarios) habitat quality maps. The results suggested that habitat quality would continually decline throughout the duration of the study except the situation under S2.

The proportion of the study area with medium-level habitat quality increased significantly from 11.3% in 1992 to 27.8% in 2052 (S3) (Table 6, Figure 9). The proportion of poor habitats increased significantly from 5.8% in 1992 to 16.8 in 2052 (S3) (Table 6). The other habitat quality level has experienced only minimal area changes. The results show that the majority of the study area exceeded the average habitat quality level (Table 6). The result of the simulated future average habitat quality level in 2052 decreased significantly in S1 and S3 compared to S2 (Table 6, Figure 9). This indicated that the S2 scenario is the best scenario for better habitat quality in the study area under the future situation in 2022–2052.

Figure 9. The ratio of habitat quality grade in 1992, 2022 and under each scenario in 2052.

The statistical analysis indicated that the average habitat quality for the years (1992, 2022, and 2052) was 0.76, 0.67, and 0.54, respectively under business-as-usual scenario (S1) (Figure 9). This reveals that even though there was a slight decrease over the time, slightly more than half of the area possessed excellent habitat level. However, by the year 2052, it is projected that this value will decrease to below 40% if scenario S3 would be the case, indicating that the most unfavorable situation in terms of habitat quality is projected to occur in 2052 S3. During this time, the percentage of habitat quality with poor quality level is anticipated to rise from 5.8% in 1992 under S1 to 16.8% 2052 under S3 (Table 6). These results point out that the S3 characteristics which is expressed by the socio-economic progress, expansion of forest coffee plantations, agricultural investments and related activities for the purpose of socio-economic development in the region instead of putting much effort towards conservation of the natural resource base, including habitat quality.

In addition, the majority of the habitat quality area was rated excellent over the specified time period, closely followed by good and moderate scores. Approximately 61.3%, 50.5% and 41.7% of the area have excellent habitat quality (Tale 6). This means that for S3 in 2052 the proportion of higher quality habitats increased by 25.4% compared to the base year. Of the three scenarios, the proportion of poor habitats in S2 was lowest at 9.8%, while S1 and S3 had proportions of 16.2% and 16.8%, respectively. This finding once again showed that the proportion of areas with poor habitat quality would be less with the successful implementation of the biodiversity protection policy in S2. More specifically, in all the three scenarios, there is an increase in the moderate level habitat quality range from 19.5% to 20.9% and 27.8% in S1, S2 and S3, respectively (Table 6). Based on the three scenarios, the average habitat quality of S2 is close to the highest average habitat quality of the baseline, with only a slight difference of 0.14 between them (Table 6).

Moreover, across all three scenarios in 2022–2052, the worst habitat quality is observed at the poor level, while the highest habitat quality is found at the excellent level, with an average value of 0.54, 0.62, and 0.51 for scenarios for the year 2052 S1, S2 and S3 (Figure 9). Among all options, habitat quality had the highest average value under S2 at 0.62 of the 2052, while it had the lowest average value under S3 at 0.51 (Table 6 and Figure 9). In all the three 2022–2052 scenarios, it is evident that the overall pattern of habitat quality in the area has significantly changed in S₃ (Figure 9). Moreover, S1 indicated that if the current situation remains unchanged, the overall declining trends in habitat quality is continuing (Table 7 and Figure 9).

In addition, we observed the relative changes in the distribution of habitat quality within the BR area. In the S1, habitat quality will deteriorate by up to 9%. In the S2, the habitat quality in the BR area would be improved by up to 0.85% (Table 8).

Table 8. Percent change in habitat quality in the YCFBR.

Snow.	Scenario Percentage Change in Habitat Quality in the BR
	S1	−0.09
	S2	+0.85
	S3	−0.04

3.4. Habitat degradation under different scenarios

To show how severely habitats have been impacted in different areas of the BR, we used the reclassification technique in ArcGIS. Accordingly, habitat degradation level in the area was classified into five levels: no degradation (0.3–0.7), mild degradation (0.2–0.3), moderate degradation (0.1–0.2), moderate severe degradation (0.04–0.1) and severe degradation (0.02–0.04) (Figure 10). The results revealed a gradual decrease in habitat quality in the study area over the period. The proportion of habitat area that remained undamaged decreased from 61.3% in 1992 to 35.9% in 2052 (S1) (Table 6). In contrast, the proportion of areas with significant habitat degradation, as measured by severity, increased. Specifically, the proportion of areas with severe habitat degradation increased from 0.2% to 4.3%, while the proportion of areas with moderate habitat degradation increased from 1.4% to 19.5% between 2022 and 2052 (S1).

Figure 10. Habitat degradation degree of the YCFBR in percentage from 1992 to 2052.

In terms of spatial distribution, the degree of habitat degradation was high and moderately high in agricultural areas and near large urban areas and in road networks (Figure 11). In areas with high forest cover, there was often little or no damage. The reddish and pink colored part of the protected area indicates a deterioration in habitat quality, the green colored part indicates high habitat quality. As shown in Figure 11, the number of areas colored red increased from 1992 to 2052, especially in S3, confirming the presence of habitat quality degradation.

Figure 11. Spatial distribution of habitat degradation degree in the YCFBR from 1992 to 2052.

3.5. Causes of habitat quality degradation in YCFBR

Using the Relative Importance Index (RII), the analysis discovered that the main factors that lead to the decline in habitat quality were population growth, which scored 0.88, and the expansion of agriculture, specifically the development of coffee plantations in forested areas and overgrazing, which scored 0.84. Moreover, the Lack of alternative livelihood options for life sustenance of the local smallholder farmers ranked the 3rd with RII 0.83. Subsequently, the extraction of forest products took place, with a particular emphasis on the widespread problem of illegal logging, collecting fuelwood, and producing charcoal (scored 0.79). Land ownership and the deterioration of land quality (rated at 0.76) are contributing factors to the decline in habitat quality in the area, along with forest fires (rated at 0.74). Additionally, these findings indicate a lack effective law enforcement and an increase in the expansion of settlements in the area have RII scores of 0.70 and 0.62 respectively, which are significant (Table 9). The decline in the quality of habitat within a BR is affected by factors associated with closeness local communities’ interaction, which directly lead to an increase in the deterioration of the local habitat quality.

Table 9. Driving factors of habitat quality degradation in YCFBR.

Proximate drivers				Underlying drivers
No		RII	Rank	No.		RII	Rank
1	Extensive Agriculture	0.84	2	1	Population growth	0.88	1
2	Settlement Expansion	0.62	8	2	Lack of livelihood option	0.83	3
3	Forest product extraction	0.79	4	3	Lack law enforcement	0.70	6
4	Forest fires	0.74	7	4	Land tenure and Land degradation	0.76	5

3.6. Discussion

The present study examined the simulated results of the three future scenarios to demonstrate the impacts of future LULCCs on habitat quality for better biodiversity conservation and environmental management, with the aim of supporting decision-makers and conservation actors. It highlights the potential of various future LU trends and their impact on future changes in habitat quality and emphasizes the need to become familiar with these changes expected in the near future, from 2022 to 2052. The FLUS scenario simulation of the three future scenarios shows the large change in the amount and distribution of different LU classes in each scenario, along with their characteristics of future habitat quality changes. Essentially, the YCFBR maintained a relatively high level of habitat quality between 1992 and 2022, but this level gradually declined over the years. A previous study by Beyene (2014) in this area reported the conversion of forest areas to agricultural land for subsistence crops and the improper management of semi-coffee forests.

Since 1992, the number of non-degraded areas has decreased, which can be seen in the degree of deterioration in habitat quality. This decline can be attributed to human factors such as the growth of coffee forests in recent years, built-up areas, and agricultural land. Most importantly, the main cause of the decline in high forest cover is the increase in open forest cover due to the expansion of forest coffee plantations in the area. This could be particularly useful in responding to the decline in habitat quality in the areas’ high forest stands. Bartczak and Metelska-Szaniawska (2015) and Joa et al. (2018) reported that the landscape patterns of forests, farmland, and construction land should be optimized in LU planning and environmental protection. Likewise, according to Nyssen et al. (2004), LULC and habitat loss are widely considered to be the main reasons for biodiversity decline in Ethiopia. Studies by Lung and Schaab (2010) in East Africa on three protected forest areas in the region also showed that deforestation and degradation of natural forests resulting from agricultural activities have negative impacts on the biodiversity of natural forest areas.

Of the three scenarios examined in this study (i.e. S1, S2 and S3), S2 simulated a significant increase in high forest cover (protection) due to declining trends in open forest areas caused by the implementation of conservation measures under S2. This revealed that in the case of careful enforcement of the policies by concerned stakeholders, there may be an opportunity in the future to change the trend of severe degradation of the forest LC and achieve better habitat quality management in the area dealing with the management aspect of protecting biological diversity. Likewise, the results of this study are consistent with those (Wang et al., 2020; Limin et al., 2019) that provided evidence for the direct effects of LULC changes on habitat quality. This clearly implies that policymakers need to take certain measures to effectively protect biodiversity in the area. particularly to manage the expansion of coffee plantations to achieve a successful conservation outcome with good habitat quality while maintaining adequate economic growth. Newbold et al. (2015) and Lewis et al. (2015) showed that an effective forest conservation approach is needed to reduce deforestation and forest degradation. Given the different types of forest management approaches, it is useful to provide policymakers with evidence about which approaches are most effective for forest conservation (Sutherland et al., 2004). In contrast, the possibility of preserving high forest LC in the area would be lower in S1 and S3 than in S2, indicating a lack of effective conservation measures and adequate law enforcement in S1 and S3. This is worrying for the future conservation of biodiversity.

Due to the different LU policies implemented in each scenario, habitat quality varied significantly across the three scenarios. It is a clear description of how built-up areas have grown and expanded over time, regardless of the three possible scenarios. This is probably due to population growth, which is the main reason for the expansion of built-up areas and other important social service facilities. The area with poor habitat quality is found primarily in built-up and agricultural areas. Bai et al. (2019) and Briggs and Mainwaring (2017) note that rapid expansion of built-up areas and urbanization have led to a gradual imbalance in ecosystem function, an accelerated decline in biodiversity, fragmentation of habitat patterns, and degradation of habitat quality. Similarly, Zhang et al. (2020) reported that rapid urbanization is a major cause of habitat quality decline and often increases environmental risks. Furthermore, according to findings from Mengistu et al. (2021), the lowest habitat quality was concentrated in urban areas and along with road networks. A study by Gedefaw et al. (2020) also supports this result in the northern part of Ethiopia, showing that the increase in built-up area and arable land is due to the current strong demand for public land, public facilities, schools, and clinics. The studies by Teka et al. (2013) and Wubie et al. (2016) also showed that the inevitable and rapid expansion of built-up areas in various regions of Ethiopia is a consequence of population growth, an increase in rural settlements, and the expansion of urban areas. Accordingly, rapid urbanization has had significant impacts on habitat quality, leading to habitat fragmentation, degraded water quality, and freshwater shortages.

In addition, the results showed that the continuous expansion of agricultural areas over the last thirty years has historically led to the degradation of forest areas and, more importantly, negatively impacted areas of potential habitat quality, a trend that is also expected to continue in the future. The simulation result of Mengistu et al. (2021) supports this conclusion by finding that the overall habitat quality status is deteriorating across most of the Kaffa BR based on the predicted habitat quality values of the area in southwestern Ethiopia. Similarly, Leung and Liang (2019) reported that rapid LULC change can easily lead to the rapid decline of important natural habitat areas, such as high forest areas with more important habitat quality, resulting in severe degradation. (Berta et al., 2020) found that agricultural expansion impacts habitat quality and biodiversity and leads to habitat fragmentation. Mattos et al. (2021) discovered that landscape conversion from natural environments to agriculture and rangeland is leading to significant biodiversity declines in the tropics. Likewise, Dai et al. (2019) reported that increasing agricultural areas contributed to the degradation of habitat quality. Aneseyee et al. (2020) reported that LUC was the main factor driving regional habitat quality degradation, which was consistent with our results.

Ultimately, the main factors contributing to long-term habitat quality degradation were identified as population growth, agricultural expansion, and the lack of alternative livelihood options for local smallholder farmers’ livelihoods. Additionally, extraction of forest products, land ownership and degradation of land quality, and a lack of effective law enforcement of illegal activities in the forest area for protection were among the major factors listed in the area based on local perception. Wang et al. (2022): Human activities have been an important driving force for changing LU patterns and degrading habitat quality. Dai et al. (2019) found that increasing agricultural areas contributed the most to habitat quality degradation.

3.7. Policy implications

This study can provide policy implications for future LU according to the impacts of different LU patterns on habitat quality. Different LU patterns influence habitat quality. First, there are areas of land that have been altered or manipulated through human efforts for various agricultural practices. The declining trend over time in high forest areas would lead to a deterioration in habitat quality. First, land managers and planners should balance the relationship between the boundaries of the BR area and other LU extensions in the area by focusing on improving LU efficiency and avoiding illegal encroachments on areas with high habitat quality. Secondly, there are about half of the forest areas with high habitat quality in the region, and the best mechanisms for a conducive environment need to be developed to improve the efficiency of management of these areas with high habitat quality. In addition, attention should be paid to the livelihood aspects of the surrounding communities to find other alternative livelihood options to prevent the current expansion of agricultural land, especially the expansion of coffee forests, which could claim such areas of high habitat quality. A likely approach to ensure sustainable LU would therefore be to focus on alternative methods rather than constantly encroaching on ecological land.

3.8. Limitation of the study

In this study, changes in LULC and its impacts on habitat quality in the YCFBR were simulated by coupling the CA-Markov, FLUS and InVEST models. The simulation results were highly accurate and effectively reflected the relationship between regional human activities and natural habitats. However, there were still limitations. First, the habitat quality evaluation method still needs to be improved. We evaluated habitat quality based on threat sources, habitat suitability, and sensitivity (Song et al. 2020; Fang et al. 2021), and although we referred to some literature when selecting parameters and weights, they are still subjective. Secondly, although the InVEST model can spatially visualize habitat quality, it only considers the stress of each single threat factor on the habitat patch and ignores the interactive effect of each threat factor on the habitat in the estimation process. Secondly, Finally, for the influence scope and weight of threat sources, expert evaluation was only carried out by referring to relevant research, which is subjective to a certain extent.

3.9. Conclusion

With the ongoing progress of urbanization and industrialization, land-use change has resulted in multiple issues including the fragmentation of habitats, the loss of biodiversity, and the decline of ecosystem services. Therefore, we simulated impacts of future LULC dynamics on habitat quality use in the study area under three different scenarios. The integration of the CA-Markov model with the FLUS model allowed for the simulation and prediction of future LUCs in the study area, considering both spatial and temporal patterns. Moreover, the InVEST model was used to evaluate change of habitat quality in the three different scenarios of 2052. Simulating the habitat quality of conservation area based on the surrounding land-use change is an important method for understanding and evaluating the complex relationship between human activities and natural habitats. The CA-Markov–FLUS model selected in this study showed excellent simulation results and was suitable for simulating land-use changes in the YCFBR. The model test showed that the overall accuracy was 0. 867, and the kappa coefficient was 0.893, indicating that the model had strong applicability for predicting future land-use change in the YCFBR and effectively reflected the impact of surrounding human activities on the natural habitat conservation in the area. The LULC map of S1 shows two primary changes in LU and LC, particularly the increase in open forests and agricultural land. In S2, high forest area is expected to increase, while open forest area is expected to decrease in the same scenario, which is due to the strict implementation of conservation policies in S2. As a result, the average habitat quality in 2052 (S1 and S3) has decreased significantly compared to the previous scenarios (S2). This indicated that under the future situation in 2052, the best scenario for better habitat quality of the study area is the conservation scenario (S2). Furthermore, this study confirmed that although the YCFBR habitat quality current status seems good, its simulated trend indicated overtime deterioration which needs some sorts of actions to save its possible future deterioration. Therefore, it is necessary for the BR (biodiversity conservationists) to implement appropriate and timely fit management approaches to natural habitat conservation and effective LU planning.

Acknowledgments

The authors gratefully acknowledge the valuable databases from United States Geological Survey (USGS). We would also like to thank West African Center for Water, Irrigation and Sustainable Agriculture (WACWISA) and the University for Development Studies; lecturers in Department of Environment and Sustainability Sciences, librarians, and others staff members for their help and support during the research.

Disclosure statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability statement

The data that has been used is confidential.

Additional information

Funding

This work was supported by the West African Centre for Water, Irrigation and Sustainable Agriculture (WACWISA), University for Development Studies, Ghana and the Government of Ghana and World Bank through the African Centers of Excellence for Development Impact (ACE Impact) initiative.