Impacts of the surrounding land use land cover changes on Suba Sabeta forest, Ethiopia, and associated community perception

ABSTRACT

Impacts of the surrounding land use land cover changes on Suba Sabeta Forest over the past three decades from 1990 to 2020 and associated community perception were assessed employing an integrated approach of Landsat images analysis, household survey, key informants interview and focus group discussion. The study involved collection of both quantitative and qualitative data which were analyzed using descriptive statistics. The results revealed that forests and shrub declined from 9,220 ha and 1335 ha to 2702 ha and 783 ha while settlement and bare land increased from 866 ha and 273 ha to 5,589 ha and 3,978 ha, respectively. Cultivated land increased from 12,162 ha in 1990 to 14,329 ha in 2005 and then declined to 10,811 ha by 2020. Respondents’ responses also indicated a drastic decline in the forest cover related to fuel wood collection (81.6%), settlement (13.8%), cutting trees for construction (3.3%) and expansion of cultivated land (1.3%). According to the respondents poverty (79.08%) and population growth (20.92%) were the underlying causes of the forest cover decline. Responses further revealed disappearance of indigenous plants (biodiversity loss) (73.2%), soil erosion (18%) and decline in agricultural production (8.8%) related to the decline in forest cover. Thus, protection of the remnant forest, reforestation and developing renewable alternative energy sources might help to mitigate further decline in Suba Sebeta Forest cover and associated impacts.

KEYWORDS:

• community perception
• forest
• images
• impacts
• land use/cover

1. Introduction

Land uses and land covers change over time in response to evolving economic, social, and biophysical conditions (Lebow et al., 2012). Globally, land cover today is altered principally by direct human use: by agriculture and livestock raising, forest harvesting and management, and urban and suburban construction and development (Meyer, 1995). In developing countries like Ethiopia land use land cover changes (LULCCs) are evident pertaining to population pressure, resettlement programs, climate change, and other human and nature-induced driving forces (Shiferaw et al., 2021). Ethiopia was experiencing rapid deforestation, with 150,000–200,000 ha of forest lost in a single year (Teketay, 2001), and forests shrunk from covering 65% of the country and 90% of the highlands to 2.2% and 5.6% respectively (Hussein, 2023). According to Tsegay and Wana (2019), about 58.4% of total forest land was cleared from 1957 to 2017 in central highland of Ethiopia.

The anthropogenic LULCCs are currently accelerating with intense implications on biodiversity loss, distress in hydrological cycles, increase in soil erosion, and sediment loads at local, regional and global scales (B. Miheretu & Abegaz, 2017). Evidently, Ethiopian highland is facing a strong LULCC which is very likely intensified by climate change and the country has endangered and threatening the high diversity of its endemic plants and animals of the unique massif (Hussein, 2023). Further, the LULCCs have increased degraded lands, soil erosion, sediment and nutrient loads to water bodies while reducing livestock number and products, crop yields, and fish populations (Gisha Kuma et al., 2022). According to Kalsido and Berhanu (2020), land-use change has also been a factor that alters the hydrologic response of the watersheds leading to influencing on sediment yield changes. This implies that, the LULCC in Ethiopia is currently leading to unsustainable practices and depletion of the forest resources. Thus, this study was conducted to assess impacts of the surrounding LULCCs on Suba Sabeta Forest over the last three decades from 1990 to 2020 and associated local community perception.

2. Materials and methods

2.1. Description of the study area

Suba Sabeta Forest is situated between Wolmera and Sabata Hawas District at about 50 km west of Addis Ababa, the capital of Ethiopia (Figure 1). The forest is one of the six sub-branches that are administered under Oromia Forest and wildlife Enterprise of Finfine branch, and it is the oldest protected area in Africa (Jemal & Getu, 2018). It is one of the few remaining highland forest blocks in the Central plateau of Ethiopia, and dominated by Juniperus procera (Alemayehu et al., 2009).

Figure 1. Map of the study area.

The topography of Suba Sabeta Forest and the surrounding environs comprises an impressive range of landscape, hills and escarpments separated by deep river valleys. Altitude ranges from about 2,200 at the Southern slope of mount Wechecha at the edge of the forest to about 3385 m.a.s.l. at Dhamocha Mogele which is the highest peak at the vicinity of Addis Ababa. According to the data from the National Meteorological Agency (NMA) the mean annual temperature in the area is 16.3°C. The range of mean monthly minimum and maximum temperature is 9.8°C and 23.6°C, respectively. The hottest month is May with a maximum temperature of 26.1°C, and the coldest is December with a minimum temperature of 7.3°C. The mean annual rainfall according to the same data source is 984 mm with main rainy season ranging from June to September, and the highest rainfall occurring in July and August.

2.2. Research design and methods

The research design involved analyses of landsat images of Suba Sebta Forest and its surrounding and a cross-sectional survey of local community perception. Direct satellite image analysis was used to estimate different land use land cover changes at the study area for the last 30 years (from 1990 to 2020) using GIS and remote sensing. Household survey, key informant interview and focus group discussion were also conducted to assess local community perception related to impacts of the surrounding land use land cover change on Suba Sebeta Forest.

2.2.1. Satellite image analysis

The images used in this study were downloaded from the official website of Global Land Cover Facility (GLCF) and USGS. Detailed characteristics of the sources of data are Landsat TM for 1990 and 2005 and Operational Land Imager (OLI) 2020 images the three scenes corresponded with path 169 and row 054 of the WRS-2 (Worldwide Reference System). The scenes were cloud-free, had good radiometric quality and were the best available satellite images for the study period. The imageries were processed in the Geographic Information System (GIS) environment. Image pre-processing including geometric, atmospheric and topographic corrections were carried out to ensure spatial and temporal comparability of the datasets. The three sets of images were first geo-corrected and geo-referenced because effective image processing is critical to successful forest cover mapping and change detection. Image pre-processing allows for conformity between multi-temporal imagery necessary for quantification and spatial comparisons.

The 1997 and 2017 scenes were co-registered to the base image using additional GCPs into UTM projection with geometric errors of less than one pixel, so that all the images have the same coordinate system. Image enhancement method was carried out using the histogram equalization and the differences in the pixels were compensated for to produce uniformly distributed pixels along the output axis.

The preparation and cropping of the images to cover the specified area required were achieved using subset prepared arc GIS. Ground truth information was collected and combined with the images to assess the accuracy of image classification. All data were developed into Universal Transverse Mercator (UTM) coordinate system, zone 37, with World Geo coded System (UTM WGS 84) projection parameters. The technical details of satellite images used in this study are given in Table 1.

Table 1. Characteristic of the landsat image

Image	Path/row	Acquisition year	Source
Land sat 5TM	169/054	1990	USGS
Land sat 7TM	169/054	2005	USGS
Land sat 8,OLI	169/054	2020	USGS

Apparently, five major land cover types were identified in the study area for image classification (Table 2).

Table 2. Descriptions of the land use land cover categories

LULC	Description
Forest	Areas covered with dense trees including Eucalyptus and coniferous trees, riparian trees, household and community forests
Cultivated land	Areas used for rain-fed crop production and scattered rural settlements
Settlements	Areas occupied by urban residential houses, buildings, and industrial uses
Bush/shrub land	Land with shrubs and bushes, scattered small trees mixed with grasses used for communal grazing
Bare land	Areas with no or very little vegetation cover natural or because of human activities; characterized with shallow and rocky surface along the flooded areas of the local stream valleys, over gentle and steep mountain slopes

2.2.2. Determination of normalized difference vegetation index

The Normalized Difference Vegetation Index (NDVI) was constructed from the Red band (R) and near infra-red (NIR) and calculated as:

(1) $\mathrm{NDVI}=\frac{\mathrm{NIR}-\mathrm{R}}{\mathrm{NIR}+\mathrm{R}}$(1)

The index is one of the most usable techniques for detection of vegetation cover and is important to measure the periodical change of vegetation coverage (Meeragandhi et al., 2015). High NDVI value indicates the high vegetation density while lower NDVI value shows the low density of vegetation. The NDVI is highly useful in detecting the surface features of the visible area and is extremely beneficial for policy makers in decision making.

2.2.3. Household survey and characteristics of the respondents

The survey involved households from three administrative units (Kebeles) around the forest named Barfa Tokofa, Garasu Sida and Nano Suba which were selected purposively from the 10 Kebeles of Wolmera District based on geographic proximity and degree of dependence. Sample size was determined following the formula developed by Yemane (1967) with a 90% confidence level as:

(2) $\mathbf{n}\mathit{\hspace{0.17em}}=\mathit{\hspace{0.17em}}\mathbf{N}/\mathbf{1}+\mathbf{N}\mathit{\hspace{0.17em}}\left({\mathbf{e}}^{\mathbf{2}}\right)$(2)

Where, n is the required sample size; N is the target population size and e is the desired level of precision. Accordingly, n = 2,292/1 + 2,292(0.1²), n = 96

The 96 respondents were selected randomly from the three Kebeles based on percentage allocation (Table 3), and both open and close ended questionnaires were used to collect the necessary data.

Table 3. Sample respondents taken from the selected Kebeles

Kebele	Total Households	Sample taken
Nano Suba	792	33
BarfaTokofa	907	38
GeresuSida	593	25
Total	2,292	96

Source: own survey.

Of the total households surveyed, 75.73% (73) were males and majority of the respondents (79.5%) were married while 16.7% and 3.8% were single and divorced, respectively (Table 4). About 48.5% of the respondents were within the age range of 35–44 while 38.2% were above 45 years. This shows that most of the respondents were within productive age group with the potential to analyze the previous year’s land use system and can understand impacts of the surrounding land use change on the forest. About 42% of the respondents were illiterate while 36% attended elementary school and 18% completed secondary education. Thus, most of the respondents were with the education level below secondary school and this indicates that their exposure to proper land utilization system and concepts about sustainable natural resources and realities was low. The respondents were divided into six economic classes based on their monthly income (USD) (as < 8.92 very poor, 8.92–17.85 poor, 17.85–26.78 lower middle class, 26.78–35.71 middle class and > 35.71 upper class) (Table 4) (Belete et al., 2021). About 35% of the respondents earn monthly income of less than 8.92USD while about 37% earn between 8.92–17.86USD. The survey result revealed that about 68.6% of the respondents earn below 17.85USD per month and this encourages dependence on the forest resource. Increase in households’ income leads to energy transition to modern fuels while reducing the use of traditional fuels like wood.

Table 4. Characteristics of respondents

Attribute	Frequency	Percentage	Attribute	Frequency	Percentage
Sex			Monthly income(USD)
Male	73	75.73	<8.92	33	34.7
Female	23	24.27	8.92–17.85	33	33.9
Age			17.85–26.78	7	7.1
35–44	47	48.54	26.78–35.71	17	17.6
45–54	30	30.96	>35.71	6	6.7
55–64	1	0.84	Life time in the area
>65	6	6.28	5–10	1	0.8
Marital status			11–15	11	11.3
Single	16	16.74	16–20	35	36.8
Married	76	79.50	21–25	26	27.2
Divorced	4	3.77	26–30	12	12.6
Educational level			>30	11	11.3
Illiterate	41	42.26
Elementary	35	36.4
Secondary school	17	17.99
> Sec. school	3	3.35
Total	96	100.00	Total	96	100

Source: own survey, 2022.

2.2.4. Key informants interview

Key Informant Interview (KII) was aimed to obtain detailed information about impacts of the surrounding LULCCs on Suba Sebeta Forest. The researcher collected the necessary information by using a semi-structured interview method with 25 informants (Table 5) consisting of 6 community leaders, 6 government officials, 2 Environmental Protection experts, 6 Forest Grades and 3 OFWE Experts and Environmental leaders who had rich knowledge about the LULCC of study area.

Table 5. Key informants

Key informants	Admin. unit	Number	Remark
Community leaders	Kebele	6	2 from each Kebele
Government officials	Kebele	6	2 from each Kebele
Environmental Protection experts	District	2	1 from Sabata Hawas and Welmera Districts each
Forest Grades		6	selected from the surrounding
OFWE Experts	District	3	1 from each Head, Branch and District offices
Environmental leader	District	2	1 from Sabata Hawas and Welmera Districts each
Total		25

2.2.5. Focus group discussion

Focus Group Discussion (FGD) was conducted with 10 participants comprising 3 local community representatives (1 from each Kebele), 3 leaders of traditional social institutions (Idir and Iqub) (1 from each Kebele), 2 local government officials (1 from each District) and 2 land/forest administration (1 from each District). The participants were selected based on their knowledge about impacts of the surrounding LULCCs on Suba Sebeta Forest.

2.2.6. Field observation

Field observation was another primary data collection tool employed to gather information about the existing or current situation of land use land cover change and its impacts on the forest resource by direct personal observation. The researcher transited the study area watershed and observed topography/relief, vegetation covers, development interventions, and community livelihood with special focus on forest resource dependence. This was used as a supportive technique to collect data that may complement or set in perspective data obtained by other means.

2.2.7. Data analysis

The results of satellite image analysis were presented using maps indicating the land use land cover changes over the study period while the data collected from respondents were analyzed descriptively using frequency and percentage.

3. Results and discussions

3.1. Land use land cover of the study area

The satellite images of land use land covers of the study area are indicated in Figure 2 and the corresponding areas for the land use land cover classes are given in Table 7. Computation of the spatial extent of land use land cover after analysis and supervised classification of the 1990 image revealed that cultivation, forest, shrub, settlement and bare land covered 12,168 ha (51%), 9,220 ha (38.64%), 1,335 ha (5.6%), 866 ha (3.62%) and 273 ha (1.14%), respectively (Table 6). The spatial extent of land use land cover map of the year 2005 showed that cultivation covered the highest area 14,329 ha (60.05%) followed by forest land 5,580 ha (23.38%). Bare land, settlement and shrub land covered 1790 ha (7.50%), 1395 ha (5.85%) and 768 ha (3.22%), respectively. Results for the year 2020 revealed that crop land occupied the highest area 10,811 ha (45.31%) followed by settlement 5,589 ha (23.32%) while bare land, forest land, and shrub land occupied 3,978 ha (16.67%), 2,701 ha (11.32%), and 783 ha (3.28%), respectively.

Figure 2. Land use land cover maps of the study area.

Table 6. Accuracy assessment of classified images, 1990–2020

A) 1990	LULC class	User Accuracy
		Cultivation	Shrub	Bare Land	Settlement	Forest	Total	OmissionError(%)
Producer Accuracy	Cultivation	32	0	1	0	0	33	2.94
	Shrubs	1	23	0	0	0	24	4.17
	Bare land	0	0	20	0	0	20	0.00
	Settlement	0	1	0	13	0	14	0
	Forest	1	0	0	0	24	25	4
	Total	34	24	21	13	24	116
	Omission Error (%)	5.88	4.17	4.76	0	0
fOverall Accuracy = $\frac{32+23+20+13+24}{33+24+20+14+25}\times 100\mathrm{\%}$ = 96.55% and overall kappa coefficient = 95.16%
		User Accuracy
B) 2005	LULC class	Cultivation	shrubs	Bare land	settlement	forest	Total	Omission Error (%)
Producer accuracy	Cultivation	11	0	0	1	1	13	15.38
	shrubs	0	39	1	1	0	40	2.50
	Bare land	1	0	9	0	0	10	10
	settlement	0	1	1	25	0	27	3.70
	forest	0	1	0	0	15	16	6.67
	Total	12	41	10	27	16	106
	Omission Error (%)	9.09	4.89	10	7.40	6.25
Overall Accuracy $=\frac{11+39+9+25+15}{13+40+10+27+16}\times 100\mathrm{\%}=93.39\mathrm{\%}\mathrm{and}\mathrm{overall}\mathrm{kappa}\mathrm{coefficient}=91.14\mathrm{\%}$
		User Accuracy
C) 2020	LULC class	Forest	Cultivation	Settlement	Shrubs	bare land	Total	Omission Error (%)
Producer Accuracy	Forest	11	0	0	0	0	11	0
	cultivation	1	45	0	1	0	47	4.25
	settlement	0	0	16	0	1	17	5.88
	shrubs	0	1	0	10	0	11	9.09
	Bare land	0	0	1	1	25	27	7.4
	Total	12	46	17	12	26	113
	Omission Error (%)	8.33	2.17	5.88	16.67	3.84
Overall Accuracy = $\frac{11+45+16+10+25}{11+47+17+11+27}\times 100\mathrm{\%}=94.69\mathrm{\%and}\mathrm{overall}\mathrm{kappa}\mathrm{coefficient}=92.75\mathrm{\%}$

3.2. Accuracy assessment of classified images

Accuracy assessment was performed for classified images of the particular years using Google Earth imagery and referenced data from the field. The result of classification was then validated by creating a confusion matrix from which different accuracy measures were derived. Finally, user accuracy, producer accuracy, over all accuracy and Kappa coefficient were calculated using the images of the specified years following Othow et al. (2017) as:

$\mathrm{Overall}\mathrm{Accuracy}=\frac{\mathit{total}\mathit{number}\mathit{of}\mathit{correctly}\mathit{classified}\mathit{number}\mathit{of}\mathit{pixels}\left(\mathit{diagonal}\right)}{\mathit{total}\mathit{number}\mathit{of}\mathit{reference}\mathit{pixels}}\times 100\mathrm{\%}$

$\mathrm{User}\mathrm{accuracy}=\frac{\mathit{number}\mathit{of}\mathit{total}\mathit{number}\mathit{of}\mathit{correctly}\mathit{classified}\mathit{number}\mathit{pixels}\mathit{in}\mathit{catagory}}{(\mathit{total}\mathit{number}\mathit{reference}\mathit{pixel}\mathit{in}\mathit{that}\mathit{category}\left(\mathit{ROW}\mathit{total}\right)}\mathrm{x}100\mathrm{\%}$

$\mathrm{Producer}\mathrm{accuracy}=\frac{\mathit{number}\mathit{of}\mathit{total}\mathit{number}\mathit{of}\mathit{correctly}\mathit{classified}\mathit{number}\mathit{pixels}\mathit{in}\mathit{catagory}}{\mathit{total}\mathit{number}\mathit{reference}\mathit{pixel}\mathit{in}\mathit{that}\mathit{category}\mathit{the}\mathit{colume}\mathit{n}}\mathrm{x}100\mathrm{\%}$

$\mathrm{Kappa}\mathrm{statistics}=\frac{\left(\mathit{Ts}\times \mathit{Tcs}\right)-\sum \left(\mathit{column}\mathit{total}\times \mathit{Row}\mathit{total}\right)}{\mathit{Ts}2-\sum (\mathit{columen}\mathit{total}-\mathit{Row}\mathit{total})}$

The classification of 1990 images had an overall accuracy of 96.55% with a kappa coefficient of 95.16% (Table 6a). This is in agreement with findings of B. A. Miheretu and Yimer (2017) and Agide and Singh (2017) who reported the overall accuracy of 85.8% with a kappa coefficient of 83.4% from the northeastern highlands of Ethiopia. Similarly, overall accuracy of 85.51% with a kappa coefficient of 0.81 was reported from the study conducted by Belete et al. (2021) at Bita District, South Western part of Ethiopia. Hassen et al. (2016) suggested three possible ranges of groupings for the kappa coefficient: a value > 80% represents a stronger agreement, a value between 40% and 80% represents moderate agreement and the value below 40% depicts poor agreement. The classification of 2005 images had an overall accuracy of 93.39% with a kappa coefficient of 91.14% (Table 6b) while overall accuracy of 2020 images was found to be 94.69% with a kappa coefficient of 92.75% (Table 6c). Thus, classification of images of all the three periods agrees well with the reference data.

3.3. Normalized difference vegetation index

The NDVI values in 1990, 2005 and 2020 images ranged from −0.17 to 0.67, −0.4 to 0.67 and −0.1 to 0.62, respectively. The result of 1990’s NDVI classification shows that the values for settlements and bare land; cultivation land, shrub land, and forest land ranged from −0.17 to 0.10,0.10 to 0.20, 0.20 to 0.34 and 0.34 to 0.67, respectively (Figure 3a). The result of 2005’s NDVI classification value shows that value of NDVI less than 0.2 indicates urban and bare land, from 0.20–0.32 sparse and unhealthy forest or shrubs and from 0.32–0.67 dense and healthy forest (Figure 3b). NDVI Values of the 2020 image less than 0.2 indicated urban and bare lands while those ranging from 0.15 to 0.22 indicated sparse and unhealthy forest or shrubs and those higher than 0.20 to 0.62 indicate dense and healthy forests (Figure 3c).

Figure 3. NDVI classification of 1990 (a), 2005(b) and 2020 (c).

3.4. Trends of land use land cover change, 1990–2020

The LULC trend analysis made for the past three consecutive decades from 1990–2020 has shown that the study area was subjected to considerable LULCCs (Figure 3 and Table 7). Forest land showed a dramatic decline in contrast to settlement and bare land which significantly increased to become the second and third major land covers by 2020 (Table 7). From 1990 to 2005, cultivation, bare land and settlement experienced significant increase while forest land and shrub land experienced a dramatic decline. This is in agreement with Belete et al. (2021) which reported an estimated annual rate of change of forest of − 1.21%, −0.72% and − 1.95% in the first, second and third periods. The land cover change analysis made at south central Ethiopia by Mariye et al. (2022) also indicated a considerable decline in forest cover from 43.1% in 1973 to 13.1% by 2000 and in accordance with the present finding. From 2005 to 2020, settlement, bare land and shrub land had their area experienced a positive change while forest land and cultivation experienced a negative change (Table 7). Mariye et al. (2022) also reported an overall expansion of settlement to 55,269% with an increasing rate of 12.3 ha/year from 1973 to 2018 which is in line with the present finding. According to Tsegay and Wana (2019) in central high land Ethiopia about 58.4% of total forest land was cleared between 1957 up to 2017 which is in line with the present finding. The land use land cover change analysis made in Huluka watershed by Ogato et al. (2021) over 38 years (1979–2017) also indicated an increase in the settlement area by 351% at an average rate of 16.20 ha/year and agrees well with the present finding.

Table 7. Land use land cover of the study area, 1990–2020

LULC Class	Area (ha)			Area (%)			Change (ha)
	1990	2005	2020	1990	2005	2020	1990–2005	2005–2020	1990–2020
Forest	9,220	5580	2702	38.64	23.38	11.32	−3640	−2878	−6518
cultivation	12,168	14,329	10,811	51	60.05	45.31	2161	−318	−1357
settlement	866	1395	5,589	3.62	5.85	23.42	469	4194	4723
shrubs	1,335	768	783	5.6	3.32	3.42	−567	15	−552
Bare land	273	1790	3,978	1.14	7.50	16.67	1517	2188	3705
Total	23,862	23,862	23,862	100	100	100

3.5. Assessment of local community perception

The average annual income of majority of the sampled respondents was very low and this is something generally accepted as the common cause of land cover change and forest degradation. Households leaving around Suba Sabeta Forest sell at least one type of forest product to earn additional income (pers. obs.) and this is in agreement with Duguma et al. (2019) which states that the poor livelihood conditions force communities to utilize all possible sources of income in their surrounding legally or illegally. Malagnoux (2007) further states that poor people are obliged to exploit whatever resource they access which is in accordance with the present finding. About 87.5% of the respondents lived in the area above 15 years while only 12.5% were recent settlers and about 91.6% (88) of the respondents agree and understand the LULCC in the study area.

About 78.62% of respondents perceived that Suba Sabeta Forest is declining which may even get worse if urgent intervention measures are not taken (Table 8). In contrast, 8.36% of the respondents mentioned that the forest cover was increasing which might be due to the community observing annual plantation activities carried out as part of the green legacy initiative while 12.7% replied that there was no change in forest cover. The respondents identified poverty (79.08%) and population growth (20.92%) as the underlying causes of the LULCC in the study area while firewood collection (49.8%), charcoal production (31.8%) and settlement (13.8%) were identified as the proximate causes (Table 8). Further, majority of the responses indicated that forest cover decline (72.8%) and soil erosion (18%) were among the concrete results of the LULCC in the area.

Table 8. Responses on the land use land cover change in the area

Questions	Response	Frequency	Percent
Trend of forest cover	Increasing	8	8.36
	Decreasing	76	78.62
	No change	12	12.7
Proximate causes	Fire wood	48	49.8
	Charcoal production	31	31.8
	Settlement	13	13.8
	Timber	3	3.3
	Cultivation	1	1.3
Underlying Causes	Poverty	76	79.08
	Population growth	20	20.92
Associated Impacts	Disappearance of species	70	73.22
	Soil erosion	17	18
	Production decline	9	8.8
	Total	96	100

Source: own household survey, 2022.

The respondents strongly agree and agree that settlement, extraction of forest for construction, fire wood collection, charcoal production and lack of forest conservation policy were the reasons for the decline in forest cover (Table 9) and this agrees well with Duguma et al. (2019) that reported illegal cutting of trees and expansion of farm land as main causes. About 80.71% of the respondents believe that lack of alternative source of energy, using forest wood as a building material; unscientific forest exploitation, knowledge gap/low awareness, and heavy dependency on biomass energy for household consumption and illegal logging were the challenges to the forest resource.

Table 9. Responses on factors of decline and challenges of the forest

variable		Strongly Agree		Agree		Disagree		Strongly Disagree		I don’t Know
		Freq.	%	Freq.	%	Freq.	%	Freq.	%	Freq.	%
Factors of decline	Settlement	29	30.13	24	25	14	15	17	18	12	13
	Construction	21	21.76	57	59	11	11	3	3	5	5
	Fire wood collection	44	45.61	33	34	11	11	3	3	6	6
	Charcoal production	41	42.26	29	30	14	15	6	6	6	6
	Lack of effective policy	12	12.13	67	70	6	6	6	6	6	6
Challenges	Lack of alternative energy source	36	37.66	57	59.41	2	2.09	1	0.84	0	0
	Use as building material	11	11.3	62	64.85	11	11.72	2	2.51	9	9.62
	Poor exploitation	9	9.62	56	58.16	12	12.55	16	16.74	3	2.93
	Increased settlement	16	17.15	68	70.71	8	8.37	1	0.84	3	2.93
	Infrastructure development	3	3.35	37	38.49	38	39.33	16	17.15	2	1.67
	Lack of suitable action	27	27.62	59	61.09	8	8.37	2	2.09	1	0.84
	Lack of awareness	15	15.9	57	59.83	12	12.97	5	5.02	6	6.28
	Dependence on biomass energy	27	27.62	66	68.62	2	2.09	1	0.84	1	0.84
	Illegal logging	18	19.25	73	75.73	2	2.09	1	0.84	2	2.09

Source: household survey, 2022.

Results of landsat image analysis, NDVI and local community responses consistently revealed a drastic decline in Suba Sebeta forest cover in contrast to increased settlement over the study period. The results further indicated increased cultivated land during the first decade which might have partly contributed to the decline in Suba Sebeta Forest cover. Image analysis further showed that bare land significantly increased to become the third major land coverby 2020 (Table 7) and this was supported with the information generated from KIIs and FGDs. The expansion in bare land might be at the expense of the forest cover decline which is in line with Tsegay and Wana (2019) who reported theloss of 18.7% forest cover to barren land in central Ethiopia. Majority of the respondents attributed the decline in Suba Sebeta Forest to fire wood collection (45.61%) and Charcoal production (42.26%) (Table 9), and this was confirmed with personal observation which revealed that households leaving around Suba Sabeta Forest sell at least one type of forest product to earn additional income. This was also supported with Duguma et al. (2019) which states that the poor livelihood conditions force communities to utilize all possible sources of income in their surroundings. Further, results of the KII and FGD also indicated poverty, illegal settlement, lack of necessary awareness on sustainable use of the forest resources, as well as lack of coordination and effective land use policy as major problems contributing to the decline in the forest cover.

4. Conclusions and recommendations

The findings revealed significant patterns of changes in the LULC classes with differing magnitudes and rates. Forest land witnessed a drastic and alarming decline in contrast to settlement and cropland which showed a sharp rise over the whole study periods and during first study period, respectively. Thus, controlling illegal settlement, provision of alternative energy sources, creation of necessary awareness on sustainable use of the forest resources, introduction of participatory forest management and implementation of appropriate land use policy may be recommended to sustain the remnant Suba Sebeta Forest and its ecosystem services.

Acknowledgements

The authors would like to thank Ambo University for material support during field data collection. We wish to express our deepest gratitude to peasant association members and interviewees for their wonderful support in data collection. Other institutions and individuals who contributed to the project in one way or the other also deserve special thanks.

Disclosure statement

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

Data availability statement

Necessary data have been included in the manuscript.