Evaluate and analyze land cover change dynamics, driving force and their implications to biodiversity in the western escarpment of the Rift valley

Land use land cover (LULC) changes are caused by natural and human alterations of the landscape that could largely affect forest biodiversity and the environment. The study aimed to analyze LULC change dynamics and driving forces in the western escarpment of the rift valley of the Gamo Zone, Southern Ethiopia. Digital satellite images, which are downloaded from USGS, were analyzed using ERDAS Imagine (14) and Arc GIS 10.2 software and supervised image classification was used to generate LULC classification, accuracy assessment, and Normalized Difference Vegetation Index (NDVI). Drivers of LULC change were identified and analyzed. Four land classes were identified such as forest, farmland, settlement, and water‐wetlands. Settlement and farmlands have increased by 8% and 6%, respectively. On the other hand, both forest and water‐wetlands decreased by aerial coverage of 9% and 5%, respectively. The overall accuracy of the study area was 92.86%, 94.22%, and 94.3% with a kappa value of 0.902, 0.92, and 0.922, respectively. NDVI values ranged between −0.42 and 0.73. Agricultural expansion (31.4%), expansion of settlement (25.7%) and Fuelwood collection and Charcoal production (22.9%) were the main driving forces that affected the biodiversity of the vegetation in the study area. Integrated land use and policy to protect biodiversity loss, forest degradation and climate changes are deemed necessary.


| INTRODUCTION
Deforestation, forest fragmentation, land degradation and soil erosion are major causes of land use land cover change and it is a major issue in a global environment (Cheruto et al., 2016;Meyer & Turner, 1992); changes are so pervasive such that, when aggregated globally, they significantly affect key aspects of Earth System functioning (Lewis, 2006;Zürich et al., 2005).This directly impacts biodiversity throughout the world (Sala et al., 2000); contribute to local and regional climate change (Chase et al., 2000;Sintayehu, 2018) as well as to global climate warming (Houghton, 2005); are the primary sources of soil degradation (Tolba et al., 1992); and, by altering ecosystem services, affect the ability of biological systems supporting human needs (Hooper et al., 2005;Vitousek et al., 2020).Such changes also determine, in part, the vulnerability of places and people to climatic, economic, or socio-political perturbations (Kasperson & Kasperson, 2001).
Tropical forests are highly threatened by human activities that leading to forest fragmentation and habitat loss with drastic consequences for climate change (Cremonesi et al., 2021;Htun et al., 2011;Laurance, 2007).Researchers predicted that the clearing of half of the world's residual forests would remove 85% of all the species living in them (Jhariya et al., 2012;Kittur et al., 2014;Le et al., 2014).LULC changes are widespread, accelerating, and the trade-offs offset human livelihood (Agarwal et al., 2002).The rapid growth and expansion of urban centers, population pressure, scarcity of land, changing technologies are among the many drivers of LULC change in the world today (Barros, 2004).
Land cover change occurs naturally in a progressive manner, but it could at times be rapid and abrupt due to anthropogenic activities (Halefom et al., 2018;Lambin et al., 2003).Expansion of agricultural land, urbanization, population growth, and land scarcity are among the many drivers of LULC changes (Cheruto et al., 2016;Habte et al., 2021;Kebede, 2018;Tewabe & Fentahun, 2020).The indicators of LULC changes manifest as the current worldwide environmental concerns such as increasing concentrations of greenhouses gases in the atmosphere, loss of biodiversity and conversion and fragmentation of natural vegetation areas (Polasky et al., 2011;Zhang et al., 2018;Obeidat et al., 2019;Abd El-Hamid, 2020).Studies carried out in the different regions in Ethiopia indicated that LULC changes are critical threats to natural forests.The study area is facing rapid deforestation and degradation of forest resources (Wana & Woldu, 2006) due to population pressure that forced the conversion of forest land into other forms (agriculture, settlement, etc.).There is a dearth of LULC change detection studies in the study area.Hence, the aims of the present study is to evaluate and analyze LULC dynamics, driving force and their implications to biodiversity in the western escarpment of the rift valley of the Gamo Zone, Southern Ethiopia.

| Description of the study area
Ethiopia is located in the Horn of Africa and its latitude and longitude is 3 0 and 14.8 00 latitude 33 0 and 48 0 longitude.The study was carried out in the western escarpment of the rift valley of the Gamo Zone, Southern Ethiopia (Figure 1).Topographically, the study area consists of plains and hill sides of the Gamo mountain ridge between 6 05 0 N to 6 12 0 N and 37 33 0 E to 37 39 0 E. The elevation of the area ranging from 1168 m to 2535 m a.s.l and the slope of the forest ranges between 0 and 32 .

| Climate
The study area has a bimodal rainfall type.Maximum and minimum mean annual rainfall during 1999-2019 was 1141.1 mm and 491.8 mm, respectively.The maximum and minimum mean annual temperature was 33.6 C and 15 C, respectively (Figure 2; National Meteorology Agency, 2019).

| Population
The total population of the study area has increased in the three successive periods (1999, 2009, and 2019; Table 1).

| Data types and sources
Primary and secondary data were used: Ground control points (GCP) for ground truth were collected as primary data using handheld GPS.   3).

| Accuracy analysis
Since image classification without accuracy assessment is incomplete (Lillesand et al., 2003), accuracy assessment for the images was carried out.The accuracy of the classification was assessed using producers, users and overall methods of accuracy assessment.The overall accuracy, as well as kappa statistics, was calculated based on the GCP collected from the identified land-use types.Kappa statics was calculated using Equation 1:

| Land use land covers change detection
The LULC maps of 3 years showing periods with a range of 10 years in between (1999, 2009, and 2019) were generated from the satellite imageries using supervised maximum likelihood classification.To analyze the land cover structural changes in the study area showing in hectares and percentage changes between the periods 1999-2009, 2009-2019, and 1999-2019 were measured for each LULC type.
Change detection was calculated by: where, R = rate of change, Q 2 = recent year forest cover in ha, Q 1 = initial year forest cover in ha and, t = interval year between initial year and recent year.

| Vegetation index
Normalized Difference Vegetation Index (NDVI) is one of the indicators commonly used to detect the vegetation cover.NDVI values were calculated on composite images using band 3 (Red) and 4 (Near Infrared) for Landsat 7, and band 4 (Red) come with band 5 (Near Infrared) for Landsat 8. NDVI which measures the degree of greenness correlates with vegetation crown density which in turn correlates with chlorophyll content and its value is between À1 and 1. NDVI is calculated as: where NDVI = normalized difference vegetation index, NIR = near infra-red band R = red band.

| Land use land covers classification
The Four land classes identified in the study include forest, Agriculture, settlement and water bodies and water-wetlands (Figure 3a (1999), b (2009), c ( 2019)).
T A B L E 2 Remote sensing data of the study.

| Land use land covers change
Results revealed that the extent of land cover changes from forest to Agriculture in the last three decades was rapid.The decline of waterwetlands was not as dramatic as the loss of forests (Table 4).

| Land use land covers change detection
LULC change detection showed that the coverage of settlement and Agriculture increased while both forest and water-wetlands decreased (Table 5).

| Normalized difference vegetation index (NDVI)
The statistics and visual observation of the NDVI images over three successive periods (1999, 2009, and 2019) showed that major land cover changes have been taken in the study area (Figure 4a (1999(Figure 4a ( ), b (2009)), c ( 2019)) and Table 9.

| Drivers of LULC changes
Focus group discussion and Key informant interview reported that agricultural expansion is the leading cause for forest cover change followed by expansion of settlement while Demographic and Economic were indirect causes of LULC changes of the study area (Tables 10   and 11).

| DISCUSSION
The  At the satellite image of 1999, forest was the dominant LULC type making up 31% of the study area followed by agriculture (28%), settlement (26%), and water-wetlands (15%).In 2019, the overall land class has changed.Agriculture and settlement equally occupied the largest portion (34%) of the study area.The remaining portions were occupied by forest and water-wetlands.The conversion of forest to agriculture and forest to settlements increased progressively, while water-wetlands to agriculture and water-wetlands to forest was very small percentage due to fluctuation of both lakes water volume.This might be small-scale irrigation by pumping water from the lakes and rivers production of fruit and vegetables (Desalegn et al., 2014;Hamere et al., 2017;Twisa & Buchroithner, 2019;Dibaba et al., 2020;Team, 2021).Hailemariam et al. (2016) also showed that urban settlements and Agriculture expansion gained the most in the area compared to other LULC types, while forest areas exhibited a decreasing trend.Demand for food and grazing land for the growing population appears to be the driving factors (Hassen & Assen, 2018;Mekuria et al., 2018;Song et al., 2014).
The quantitative results of change analysis of 30 years with three time periods (1999-2009, 2009-2019, and 1999-2019)  The accuracy assessment was performed using land-use maps, ground truth points and Google Earth.Three periods (1999, 2009, and 2019) land use classification have shown, user's accuracy and producer's accuracy are greater than 85%, as well the overall accuracy of 92.86%, 94.22%, and 94.3%, respectively.These values indicate that the LAND SAT images and the methodologies used were accurate.
The Kappa coefficient was also calculated, with a value of K = 0.9, which indicates that the classification is almost perfect since it is greater than 0.8.9).
NDVI analysis has proven that there had been changes in vegetation cover between 1999 and 2019 images and higher values were recorded in 1999.
The key informant interviewee (KII) identified five direct and four indirect factors as important drivers for LULC changes in the study area.Agricultural expansion (31.4%) and expansion of settlement (25.7%) were the top direct drivers of LULC changes in the study area.
In addition, some of the KII reported population growth (37.1%),Economic (28.6%), and Biophysical (22.9%) were indirect causes of LULC  changes.In agreement with our findings, these causes were reported in Ayele et al. (2018) in the Libokemkem district of South Gonder, Ethiopia; Mekuria et al. (2018) in the Munessa-Shashemene, Southcentral highlands of Ethiopia.Fuelwood collection and charcoal production were the main degradation drivers for the African continent (Hosonuma et al., 2012;Mande, 2020;Solomon et al., 2018).According to Defries et al. (2010);Fisher (2010), existing deforestation in Africa is still largely driven by small-scale subsistence agriculture.
Our focus group discussion (FGD) in the different kebeles have indicated that population pressure had a great impact on forest dynamics.The demographic data of the study area over the past three decades had increased.Accordingly, in the Gamo highland there is land scarcity.Hence, there is migration to low land for settlement and farming.This causes clearing dense forest in the low land area.The work of Lambin et al. (2003) shows that the impact of human population pressure causes the accelerated conversion of natural habitats into agricultural and settlement areas to meet the mounting demand for food and housing.In the FGD, elders pointed out that the unemployed and youths are dependent on the selling of fuelwood as an immediate source of income during decline or failure of crop production as a result of drought years.This is a common survival strategy of rural populations in the events of degradation, drought, and rainfall variability across Africa (Campbell, 1990).In Ethiopia, resettlement and villagization programs during the Military Government (1977)(1978) had made a significant contribution to the expansion of settlements and agriculture.Due to the low policy enforcing capacity of the then government landless farmers cleared forests and occupied as much land as possible to increase the chances of land ownership.
Studies in other parts of the country also reported population pressure as a major driver of LULC changes (Aerts et al., 2016;Dessie & Kleman, 2007;Zhang et al., 2005).

| IMPLICATION TO BIODIVERSITY AND CLIMATE CHANGES
During the period (1999) considered in this study, the land use land cover changes were dramatic.Agricultural land and settlement increased, while the forest cover and water-wetlands showed a declining trend.Agricultural land increased from 11% in 1999 to 21% in 2019 with annual expansion rate of 10%, whereas forest cover declined by 11% in 1999, 15% in 2009, and 21% in (Gebretsadik, 2016;Karki et al., 2018;Tekalign et al., 2018).

| CONCLUSION
There were four land classes in the study area which include forest, Agriculture, settlement and water-wetlands.The changes observed in 2009 and 2019 were more rapid than in 1999 due to the expansion of small-scale irrigated farmlands for fruit and vegetable production.Moreover, field observations, KII and focus group discussion confirmed that the main cause of LULC changes in the study area was the expansion of agriculture and settlement.On the other hand, demographic, economic and biophysical conditions have been indirect driving forces of LULC changes.
Linking participatory forest management with an institution and strong monitoring policies, green legacy and creating awareness in the local population are hoped to improve the current status of forest biodiversity and the environment of the study area.Furthermore, promoting nonagricultural economy to the unemployed youths and creating forest reserve areas with a buffer zone, which helps to minimize land use and land cover conversion.
Ethio-Region, (b) Gamo Zone, (c) study area (surrounded woreda, Lake Abaya and Chamo, rivers and roads all and dry weathered).[Colour figure can be viewed at wileyonlinelibrary.com]

2. 3 |
Land-use change assessment (1999-2019) Digital satellite images were processed, classified and analyzed using ERDAS Imagine (14).Computations of the area and changes in land use categories were made using Arc GIS 10.2 analytical tools.Preprocessing of satellite images was done to create a more faithful representation of the original scene.An intensive preprocessing such as geo-referencing, layer-stacking, resolution merge, and sub sets were carried out to Ortho-rectify the satellite images into UTM coordinates (WGS, 1984) and to remove disturbances such as haze, noise, steep slope effect, and radiometric variation between acquisition dates.A stacked satellite image of the study area was extracted by clipping the Area of Interest (AOI) layer of the Gamo shapefile in ERDAS 14 software.The satellite image was classified using the supervised image classification technique which employed pixel-based supervised image classifications with the maximum likelihood classification algorithm (Clevers, 2009) to produce LULC maps of the study area.Appropriate band combinations were obtained and the signatures were used for the supervised classification.Land cover change detection for the study area was monitored at three intervals: 1999-2009, 2009-2019, and 1999-2019.The supervised classification produced four land classes, namely, forest lands, Farmlands, settlement, water-wetlands (Table

F
I G U R E 2 Maximum and minimum mean annual temperature and rainfall.[Colour figure can be viewed at wileyonlinelibrary.com]T A B L E 1 Total population of the study area.
LULC changes are influenced by a number of driving factors.In the study area, human activity is often mentioned as the major driver of LULC Changes.For a better understanding of LULC changes data were collected including field observation, focused group discussion (FGD) and key informant interview (KII).KII and FGD were selected based on the recommendation of local community leaders and agriculture extension workers.The participants included elders (male and female), agriculture extension workers and youth jobless.The informants were asked for their consent to participate in the discussion were then given clear information about LULC changes in the study area.Data were analyzed using IBM SPSS version 20.

F
I G U R E 3 Land use land cover change from 1999 to 2019.[Colour figure can be viewed at wileyonlinelibrary.com]T A B L E 4 Land use land covers change (1999-2019).
classification of satellite images show a clear conversion of land covers into farmland and settlement.The LULC categories have shown that forest land class has progressively decreased while agriculture and settlement increased from 1999 to 2019.This might be due to an increased population in the study area and fewer job opportunities for youth.Similar results were reported by Kassa and Forech (2020), Ariti et al. (2015), and Mengistu et al. (2012) showing that farmlands in the Rift valley of Ethiopia have expanded.Muzein (2006) has shown that more than 80% of the total terrestrial productive land in the Ethiopian Central Rift valley was lost to agriculture.T A B L E 5 Land use land covers change detection from 1999 to 2019.
and a change matrix from 1999 to 2019 revealed the extent of changes that occurred in different LULC classes throughout the three decades.In general, agriculture and settlement increased progressively while natural forest and water-wetlands consistently decreased over the study periods.The change results revealed that the reduction of forests over the second(2009-2019), and third   (1999-2019)  study periods was more than 9% while water-wetlands increased 5%.The total area of forest converted to agriculture between the second and third study period amounts to À200 ha while water-wetlands converted to agriculture between the first to second study period amounts to À1002.8 ha.The forest cover of the study area decreased from 31% to 20% between 1999 and 2019.Similar trend was observed in the water-wetlands land class (Figure5).Findings from other studiesZeleke and Hurni (2001);Rudel et al. (2002);Lu et al. (2004);Ramankutty et al. (2006);Dessie and Kleman (2007) argued the increasing trend in agriculture and settlement to fulfill growing food demand and to solve youth questions on income generation, housing, and job.
, near-infrared band; R, red.F I G U R E 5 LULC change detection of the study area.[Colour figure can be viewed at wileyonlinelibrary.com] 2019 with annual decreasing rate of 5%.The decline of forest cover is likely to cause plant biodiversity loss and affect human wellbeing and climate change in the study area.The variations in plant biodiversity loss across the study sites had shown a clear linkage between land-use change and biodiversity loss.Land-use change, which involves clearing of the natural vegetation, changes the diversity and dominance of the plant species.The studies in other parts of Ethiopia show that land use land cover change and habitat loss are the major drivers of biodiversity depletion in Ethiopia Characteristics of land cover classes.
Overall accuracy of the study area(2009).Overall accuracy of the study area (2019).
T A B L E 6 Overall accuracy of the study area(1999).F I G U R E 4NDVI of the study area from 1999 to 2019.[Colour figure can be viewed at wileyonlinelibrary.com]DINGAMO ET AL.
Zhang et al. (2018)12)andZhang et al. (2018)argued that overall accuracy values greater than 0.8 indicate in the Landsat and the methodologies used to have high accuracy.The results revealed that the upper threshold value of NDVI was approximately 0.73 and the lower threshold value was À0.42.The pixels showing an NDVI value above the threshold were identified as vegetated areas, while low NDVI values represented nonvegetated areas.For nonvegetated areas, researcher found that water bodies (lake Abaya and rivers) were represented by low NDVI values, ranging from À0.28 to À0.42, while the pixels having NDVI values in the range of 0.51-0.73were considered as vegetation cover (Table Underlying causes of LULC changes.NDVI result of the study area.