From degradation to restoration: addressing the environmental challenges in the Ethiopian highlands Authors: Eyasu Elias, Abreham Berta Aneseyee, Afework Mekeberiaw, Habtamu Hailu, Hailu Hundie, Józef Hernik, Hamere Yohannes Published: 2025 ABSTRACT AbstractThe highlands of Ethiopia are a well-known example of severely degraded landscapes in Africa, with concerns for food shortages. The Ethiopian Highlands Reclamation Study (EHRS) has documented the rates, underlying causes, and costs of land degradation, which resulted in intensive conservation and restoration efforts over the past decades. Despite the extensive effort devoted to the problem, the actual extent of conservation and restoration work has not been adequately quantified and documented. This prompted a crucial inquiry into the current environmental status to quantify the biophysical impacts of 20 years of Soil and Water Conservation (SWC) and restoration in the Ethiopian highlands. Quantitative data on the extent of conservation and restoration measures implemented were generated through a systematic analysis of metadata obtained from Ethiopia’s watershed database, maintained by the Ministry of Agriculture (MoA), with field validation in 56 sample watersheds. The analysis of change in land use/land cover (LU/LC) and the normalized difference vegetation index (NDVI) was employed to examine the extent of vegetation restoration and forest cover change in the Ethiopian highlands from 2002 to 2022. The Google Earth Engine (GEE) was utilized for acquiring and processing satellite image data. The potential soil loss across different land-use types was quantified using the InVEST-SDR (Integrated Valuation of Ecosystem Services and Trade-offs Sediment Delivery Ratio) model, combined with other biophysical data disaggregated by physiographic regions to capture diversities in the biophysical settings. Erosivity, erodibility, digital elevation model, vegetation cover, and land management factors were the input data used for the InVEST-SDR model. The results showed that Ethiopia has achieved unprecedented progress in reversing land degradation over the past two decades, with nearly 28 million hectares of degraded landscapes treated through community mobilization and flagship programs. The forest area has increased from about 2.6 million ha in 2002 to 5.9 million ha in 2022. Conversely, the total area of bare land was reduced from 2.5 million hectares in 2002 to 830,000 hectares in 2022. Similarly, the average soil erosion rate on cultivated fields declined from 69 ton/ha in 2002 to 35 ton/ha in 2022, reflecting the cumulative impact of extensive land restoration and soil and water conservation interventions implemented across the Ethiopian highlands. These benefits underscore the effectiveness of combining community-led restoration, bottom-up conservation efforts with top-down policy support. Numerous Sustainable Development Goals, such as Zero Hunger (SDG 2), Climate Action (SDG 13), and Life on Land (SDG 15), can be achieved in part thanks to Ethiopia’s experience. Future research should focus on detailed watershed-level studies to study agro-ecological variations in conservation outcomes, conduct cost–benefit analyses of water and soil conservation measures, and implement long-term vegetation monitoring and socio-economic surveys to enhance community-led restoration efforts.Supplementary informationThe online version contains supplementary material available at 10.1007/s10661-025-14914-6.Keywords: Community watersheds, Ethiopian highlands, Land degradation, Soil conservation, Enrichment planting Introduction Deforestation and soil erosion are the dominant forms of land degradation in the Ethiopian highlands (Tadesse & Hailu, 2024). Ethiopia is one of the severely degraded environments in Africa, with the consequent implications for food shortages, droughts, and floods. The effects of climate change have further exacerbated these calamities (Amede et al., 2007b; Elias, 2017). Large volumes of data have been generated regarding the rate and extent of land degradation in Ethiopia since the mid-1980s, a period when Ethiopia was severely affected by devastating droughts and famines believed to have resulted from the alarming rates of degradation (Bojo & Cassels, 1995; FAO, 1986; Hurni, 1988; Nyssen et al., 2004). However, the most comprehensive, influential, and frequently cited report on the environmental status of the Ethiopian highlands is the Ethiopian Highlands Reclamation Study (EHRS) (FAO, 1986). The EHRS provided the first systematic estimates of land degradation and erosion rates in Ethiopia and has since played a decisive role in shaping soil and water conservation (SWC) policies and land management strategies in the country. However, the EHRS report presents the major conclusions and projections that have shaped land management policies in Ethiopia, which require updated empirical evidence—a gap that the present study seeks to address. The EHRS indicated an alarming annual deforestation rate of about 1000 km2 (approximately 100,000 ha), primarily caused by the expansion of cultivation and the extraction of wood for household energy. It further estimated a mean annual soil erosion rate of 130 t/ha on cultivated fields, with rates rising to 300 t/ha in areas lacking ground cover (bare lands). Moreover, the report projected that, out of the 54 million hectares comprising the Ethiopian highlands, 27 million hectares were already “significantly” eroded, and about 2 million hectares had been degraded beyond recovery, much of which was expected to be reduced to bare rock by 2010. Although the accuracy of these estimates and projections remains questionable, the alarming data and figures—together with the realization that the widespread droughts and famines of the 1970 s and early 1980 s were primarily attributed to land degradation—prompted successive Governments of Ethiopia (GoE) to launch large-scale restoration and soil and water conservation (SWC) programs. These efforts were supported by donors, most notably through the widely recognized United Nations Food-for-Work (FFW) program, which remains a mainstay of restoration efforts (Haregeweyn et al., 2015; Kebede et al., 2021). However, due to the physical nature of the measurements and the lack of community involvement, the FFW program was deemed unsustainable and ineffective (Campbell, 1991; Elias & Scoons, 1999). Later, in 2003, the Managing Environmental Resources to Enable Transitions to Sustainable Livelihoods (MERET) program was introduced, emphasizing community empowerment in both planning and execution (Nedessa and Wickrema, 2010). The Productive Safety Net/Public Work program (PSNP/PW) was introduced concurrently in 2005. This was the most extensive social transfer programme in Africa, reaching 8.7 million people in degraded landscapes (Desalegn & Ali, 2018; Sabates-Wheeler & Devereux, 2010). The PSNP/PW made cash transfers to labor in payment for the construction of SWC structures and in support of community forestry development (Sabates-Wheeler & Devereux, 2010). In 2007, the GoE, with financial support from the International Fund for Agricultural Development (IFAD), initiated the Participatory Small-Scale Irrigation Development Programme (PASIDP), which was aligned with the PSNP/PW. It involved restoring degraded PSNP watersheds to regulate hydrological cycles for sustainable irrigation development (Alessandra & Bezawit, 2018; Garbero and Chichaibelu, 2019). Drawing on the lessons and experiences of MERET, PSNP/PW, and PASIDP, the GoE developed the Sustainable Land Management Programme (SLMP) in 2008 and the Climate Action through Landscape Management (CALM) programme in 2019 (Teferi et al., 2020). Both SLMP and CALM focused on community watershed management activities that combined the construction of physical structures (i.e., terraces, bunds, and basins) and reforestation through enrichment planting (i.e., tree planting within degraded forests to increase the density and diversity of trees), agroforestry practices involving the planting of fruit and fodder tree/shrubs on individual farm fields, and area closures on degraded hillsides to allow natural regeneration of vegetation. Local communities are mobilized to contribute 45–60 days of voluntary labor annually to build conservation structures on millions of hectares (Abebe & Waktolo, 2021). A micro-watershed (about 560 ha) was adopted as the unit of intervention for landscape restoration by Community Watershed User Cooperatives (CWUCs) for restoration planning, monitoring, and evaluation of interventions (Amede et al., 2007a). National “Community-Based Participatory Watershed Development Planning Guidelines” were also published to standardize the implementation of the SWC measures by the CWUCs (Hadush, 2015). Despite decades of intensive Soil and Water Conservation (SWC) and vegetation restoration across the Ethiopian highlands, the scientific record remains fragmented. Numerous plot-to-micro-catchment studies and foundational assessments demonstrate the local benefits of terraces, bunds, and area closures. More recent research has linked PSNP public works to quantifiable increases in tree cover (Haregeweyn et al., 2015; Hirvonen et al., 2022; Wolka et al., 2023). However, there is no consolidated, peer-reviewed national accounting restoration area across the highlands of Ethiopia; most evidence is geographically narrow and short-term. Consequently, neither the volume of restoration work implemented nor its landscape-level impacts on erosion, sediment yield, and vegetation recovery have been quantified—gaps that this study aims to address. Documenting Ethiopia’s extensive experience of restoration and recording success stories will provide remarkable showcase examples and valuable lessons for other countries implementing similar restoration interventions. Therefore, this study was carried out (1) to document the land restoration outcomes across physiographic regions and land uses over 2002–2022; and (2) to quantify the impacts of the land management practice interventions on soil erosion rates and environmental restoration in the Ethiopian highlands. Materials and methods Description of the study area The Ethiopian highlands are characterized by undulating to rolling high plateaus, rugged volcanic mountains, rolling plains, and deep gorges incised by rivers (Elias et al., 2019; FAO, 1984). Elevation above 1500 m is generally considered highlands (Fig. 1) and covers approximately 54 million hectares, accounting for 48% of the country’s landmass (1,120,000 km2). Due to the landform, topographic features, and potential for agricultural production, the highlands of Ethiopia are classified into five physiographic environments (Fig. 1), including the Tigrian Plateau, the North-Central Highlands, the Southwestern Highlands, the South-Eastern Highlands, and the Hararghe Plateau (Elias, 2016). Taking a more nuanced approach, the land restoration and interventions were assessed across these physiographic environments to provide a more updated and disaggregated understanding of the progress made in terms of landscape restoration. Location of the study area with topographic features within physiographic units of the Ethiopian highlands. NC, north-central highlands; SW, south-western highlands; SE, south-eastern highlands The climate ranges from cold, sub-humid, and afro-alpine highlands at elevations of 2500–3200 m above sea level, while tepid, moist to sub-moist conditions prevail in the mid-altitudes between 1600 and 2400 m above sea level. The 30-year rainfall and temperature weather data obtained from meteorological stations across the highlands show that the highest mean annual rainfall of 1343 mm was recorded for the southwestern highlands, with the lowest annual average rainfall of 689 mm in the Tigraian plateau. The overall average rainfall for the entire highlands is 987 mm/annum, and the mean temperature is 19.3 °C. The mean minimum and maximum temperature are 10.4 °C and 27.6 °C, respectively, recorded for the Tigrian and Hararghe plateaus (Fig. 2). According to the revised soil map of Ethiopia (Ali et al., 2024), the Ethiopian highlands are covered by five Reference Soil Groups (RSGs)—Leptosols (36%), Cambisols (18%), Vertisols (15%), Luvisols (12%), and Nitisols (9%). Mean monthly rainfalls and mean minimum and mean maximum temperatures in the Ethiopian highlands Data sources and methods Quantitative data on the extent of SWC and reforestation measures were generated through a systematic review of metadata obtained from the watershed database of the Ministry of Agriculture (MoA), which is widely used in similar studies (Ahmad et al., 2020; Hasan et al., 2020; Tamene et al., 2022). To explore the extent of vegetation restoration and changes in forest cover, LU/LC change analysis and the normalized difference vegetation index (NDVI) (https://code.earthengine.google.com/) were employed. To detect soil erosion control and reduce sediment yield, the InVEST (Integrated Valuation of Ecosystem Services and Trade-offs; https://naturalcapitalproject.stanford.edu/invest/sediment-retention) model was applied. Metadata analysis from the MoA watershed database Quantitative data on the extent of 20 years of SWC work implemented (2002–2022) by large-scale flagship programmes and high-priority initiatives funded by donor communities were sourced from the watershed database of the MoA (http://slmpkmis.gov.et:8080/jspui/handle/123456789/1 and http://nrm-mdis.moa.gov.et/) (Fig. 3a). This is primary data for the MoA itself, but secondary data for the authors, who organized the datasets into a coherent protocol for analysis. The MoA data were then standardized with conservation and restoration data from different sources (i.e., MERET, PSNP/PW, SLMP, PASIDP) and integrated into a uniform template to ensure compatibility and facilitate seamless integration and analysis. After obtaining spatial and tabular datasets from the MoA watershed database, we conducted a field investigation in 56 sample watersheds (Fig 3b) selected from those undertaken in the restoration land intervention under flagship projects. Within these watersheds, detailed field measurements were made for topographic parameters (DEMs), land use/land cover (LU/LC), and land management practices such as the type of conservation structures constructed (bunds, terraces, micro-bains, check dams, etc.) and their maintenance status, current use, and geographic locations (i.e., GPS points). In addition to validating the nationally reported restoration work, these data served as inputs to the InVEST-SDR model–DEM data (LS factor), vegetation cover (C-factor), and conservation practices (P-factor). Validation was conducted by comparing the reported interventions data with measured field data, which revealed no significant discrepancies between the MoA documentation and field verification. In addition, we reviewed midterm and end-line evaluation reports (MoA, 2019) and published articles associated with biophysical impact assessments under the flagship projects (Alemaw et al., 2024; Berhane et al., 2021; Debie, 2024; Hirvonen et al., 2022; Weldegebriel et al., 2025; World_Bank, 2020). Spatial distribution of land restoration watersheds under the flagship projects (a) and sample watersheds used for field work (b) Soil erosion and sediment export quantification Soil loss across different land use types was estimated using InVESTmodel based on the Revised Universal Soil Loss Equation (RUSLE) (Eq. 1). where RUSLEi, the mean annual soil loss (t/ha/year); R, rainfall erosivity derived from the mean annual rainfall of the area (MJ Mm/ha/year); K, the soil erodibility (MJ mm/t/ha); LS, the topographic factor derived from the digital elevation model of the area; C, the cover factor; and P, land management factor. Erosivity (R) factor The rainfall erosivity factor (R) was derived from the mean annual rainfall using the empirical formula (Eq. 2) provided by Hurni (Hurni, 1985); where R is the rainfall erosivity factor (MJ mm/ha/hour/year), and P is the mean annual rainfall (mm). Thirty years of monthly rainfall data from the Ethiopian Meteorological Institute (Annex 1) were aggregated to determine the mean annual rainfall (P) for each reference period (2002–2009, 2009–2016, and 2016–2022). Rainfall data were collected from meteorological stations located inside or near the sample watersheds to ensure representativeness and data reliability. To capture the spatial variability in rainfall throughout the Ethiopian Highlands, these data were further classified by physiographic region (Table 1). This approach captures both temporal and spatial patterns of rainfall while ensuring that the computed means are representative of the project intervention areas. ArcGIS spatial analysis, utilizing the Inverse Distance Weighting (IDW) tool, was employed to generate spatial raster data on erosivity for the InVEST model (Annex 2). The spatial data generated from this analysis is provided in Annex 3. Table 1Mean annual rainfall (P) and R factor across physiographic regions and reference periods (after Hurni, 1985)Physiographic regionMean annual rainfall (P)Rainfall Erosivity factor (R) value20022009201620222002200920162022Tigrian plateau104892512041425593525681805North-central highlands1865163419752186104991911111229Southwestern highlands19231893227723481086106812841324South Eastern highlands128413381775237672675610021340Hararghe plateau1221117912371317691668700745Mean14681394169319308297879561089 Mean annual rainfall (P) and R factor across physiographic regions and reference periods (after Hurni, 1985) Soil erodibility (K factor) Soil erodibility is the susceptibility of soil particles to detachment by the force of raindrops and transportation by the runoff. The soil erodibility factor was adapted for the Ethiopian highlands by focusing on easily observable soil erodibility factors of soil types differentiated by soil colour (Hurni, 1985). The K-values assigned based on soil colour are well matched with the dominant soil types (see Annex 4). The most important soil characteristics that determine soil erodibility are soil particle (sand, silt, and clay), soil organic matter content, soil profile permeability, and soil structure (Wischmeier & Smith, 1978). Data on these properties of the major Ethiopian soils were obtained from the National Soil Information System (NSIS) database (https://nsis.moa.gov.et/). This data was subjected to IDW spatial analysis using the ArcGIS environment to derive the raster K factor (Fig. 4a). Spatial soil erodibility factor (K) (a) and DEM used for the InVEST model (b) Digital elevation model (DEM) The topographic factors (LS) of erosion were derived from a 30-m resolution DEM obtained from the USGS Earth Explorer (https://earthexplorer.usgs.gov). From this DEM, slope gradient (S), slope length (L), and flow accumulation were extracted. The raster inputs used for the InVEST model were resampled to match the cell size of the DEM (Fig. 4a and b). Analysis of change in land use/land cover (LU/LC) LU/LC changes were examined over four periods to capture key timelines associated with land restoration policies, strategies, and environmental shocks such as drought that significantly affect the land degradation–restoration continuum. Period 1 (2002–2009) represents the transition from Food-For-Work to the MERET programme and the design of the PSNP/PW. Period 2 (2009–2016) is marked by the design and implementation of the first phase of the SLMP and the formulation of the climate resilient green economy (CRGE) strategy, which includes climate-resilient soil and water conservation practices through mass labor mobilization. Drought, conflict, and the spread of the COVID-19 pandemic during Period 3 (2016–2022) limited community labor mobilization for environmental protection and tree planting campaigns, which in turn affected the dynamics of land use/land cover (LU/LC). The Google Earth Engine was utilized for acquiring and processing satellite image data (Pham et al., 2020; Robinson et al., 2017), using https://code.earthengine.google.com/. The Landsat and Sentinel-2 satellite images of Level 2 surface reflectance (SR) were obtained from Landsat 5 and 7 for the years 2002 and 2009, respectively, and from Landsat 8 for the reference periods of 2016 and 2022. Landsat satellites capture imagery over the exact location (185 km2) once every 16 days, allowing any specific location in the Ethiopian highlands to be observed twice a month. To avoid seasonal fluctuations of land cover due to weather and phenology, only the months in the dry season (i.e., November, December, January, February, and March) were considered. Then, the composite mean monthly NDVI value was calculated from the satellite observations for each month in the dry season of the year. The use of monthly mean C-factor values is proposed by Durigon et al. (2014) and widely applied. Supervised classification was performed using multiple algorithms, including Random forest (RF), support vector machine (SVM), and regression trees (CART), to identify the most suitable classifier for the task. In the end, the Smile Random Forest Classifier was chosen because it outperformed the other classifiers in the area classification. This classifier utilized the decision rules of the trained algorithm to categorize the image into distinct land cover classes (Rumora et al., 2020). Accuracy assessment To assess the reliability of the land use/land cover classification, an accuracy assessment was carried out using a confusion matrix approach. Reference points were derived from high-resolution satellite imagery and field observations of the SLM intervention sites using a stratified random sampling technique. A total of 27,162 reference points were collected, of which 30% (8171 samples) were reserved for independent validation. The confusion matrix compared classified categories with the corresponding ground-truth data to determine classification accuracy. Standard accuracy metrics, including overall accuracy and the Kappa coefficient, were estimated to determine the level of agreement between the classified maps and the reference data. Biophysical table In the InVEST model, a biophysical table provides quantitative input data describing the physical and biological characteristics of different land cover types. The cover (C) factor and land management (P) factors were used in the CSV biophysical table (Table 3), which illustrates the extent of vegetative cover for each land use type and conservation practice, respectively. The cover factor (C) in various land use types compares soil loss from a reference plot against that from a clean-tilled, continuous bare fallow, with values ranging from 0 to 1; zero indicates dense forest cover, while 1 indicates bare fallow (Gunten, 1993). Table 3The biophysical table with C and P factors assigned for different land use types used in the InVEST modelLand use types2002200920162022CPCPCPCPOpen Forest*0.040.150.030.120.020.100.010.09Woodland0.050.390.040.360.030.310.020.26Shrub/bush0.120.350.110.320.100.310.090.18Cropland0.200.490.220.430.240.310.260.21Grassland0.080.280.090.260.080.130.070.14Bare land0.870.500.820.440.810.370.790.30Wetland0.150.280.130.210.120.230.100.14C cover factor for different land use types determined NDV (Eq. 4), P conservation practices assigned based on the types of SWC practices implemented in each year, as shown in Table 2*“Open forests” are forests experiencing degradation, characterized by low tree density requiring enrichment planting and protection as part of sustainable land management The biophysical table with C and P factors assigned for different land use types used in the InVEST model C cover factor for different land use types determined NDV (Eq. 4), P conservation practices assigned based on the types of SWC practices implemented in each year, as shown in Table 2 *“Open forests” are forests experiencing degradation, characterized by low tree density requiring enrichment planting and protection as part of sustainable land management The C-value was estimated by means of NDVI, which is computed by processing Landsat images using Google Earth Engine API (Application Programming Interface) (https://code.earthengine.google.com/) as shown in Eqs. (3) and (4). The estimated NDVI values were then converted to the C factor as proposed by Durigon et al. (2014). where NIR is the spectral reflectance of the surface in the near-infrared band, and RED is the spectral reflectance of the surface in the red band. The land management factor (P) was determined based on data on SWC practices generated from the watershed database discussed under the source of data section (http://slmpkmis.gov.et:8080/jspui/handle/123456789/1 and http://nrm-mdis.moa.gov.et/). The geospatial watershed database provides information on the type and area extent of SWC measures implemented across the Ethiopian highlands. Using this data, the areas treated with SWC measures, such as contour bunds, hillside terraces, and gully control structures (i.e., check dams), were quantified. Based on this data, the P values were assigned values ranging from 0 to 1, as suggested by Hurni (1985) and Ganasri and Ramesh (2016). Conversely, for non-conserved watersheds, a P value of 0.9 was assigned, considering that soil erosion continues unabated (Table 2). Table 2The P-factors assigned for different SWC measures implementedNo.SWC practicesP factorLand use typesSources1Fanya juu terracing0.37Cultivated lands Ebabu et al., (2022) 2Stone bunds0.30Cultivated fields Tian et al., (2021) 3Soil bunds0.40Cultivated fields Kebede et al., (2021) 4Hillside terracing0.30Communal hillsides Wei et al., (2016) 5Bench terracing0.21Communal grazing lands Ebabu et al., (2022) 6Micro-trenches and basins0.36Cultivated and other LU Ebabu et al., (2022) 7Area closure0.30Communal hillsides Escobar-Soldevilla et al., (2023) 8Areas with no conservation0.90All land use types Hurni, (1985) The P-factors assigned for different SWC measures implemented Based on the above information, a biophysical table was prepared to quantify soil erosion and sediment export across different land use types, using the C and P factors (Table 3). Results Extent of land area restored through flagship projects Over the past 20 years, Ethiopia has restored 28 million hectares of degraded land through the construction of physical SWC structures and the establishment of area closures and community woodlots across the highlands. This was achieved through a combination of mass labor mobilization (MLM) and flagship programs, such as PSNP/PW, SLM, and MERET, among others (Fig. 5). Of this restoration, 60% (14.4 million hectares) was achieved through MLM. In contrast, flagship projects accounted for the remaining share. Each year, approximately 20 million community members voluntarily contribute 45–60 days of free labor to construct level soil and stone bunds, fanya juu terraces, and gully rehabilitation measures (mainly check dams), along with tree planting campaigns to establish community woodlots, complemented by targeted interventions from flagship projects. The projects provide technical expertise, training for farmer-technicians, and essential tools (e.g., Jalo, GPS, string, measuring tape, shovels, hoes) as well as construction materials (cement, gravel, bamboo mats). Total land area restored and the contribution of mass–labor and flagship projects In severely degraded areas—typically steep slopes with slopes above 15%—flagship projects supplement voluntary labor with wage incentives. Restoration in these sites requires advanced engineering, including the construction of hillside terraces and gully rehabilitation structures, such as gabion and sandbag check dams. Donor-funded projects therefore apply an 80:20 cost-sharing arrangement: 80% of the work is remunerated at person-day work norms established at the national level, while the community contributes the remaining 20%. To scale up land restoration and integrate volunteer labor with institutional, financial, and technical support, community mobilization and flagship projects have worked in synergy. Among the projects, PSNP accounts for 42% (~5 M ha) of the total project-supported restored area (~12 M ha), followed by SLM (36%), MERET (9%), and CALM (8%), while PASIDP is limited in scope and coverage (Fig. 5). Land restoration across different physiographic regions The data in Table 4 shows that land restoration and conservation interventions are prioritized according to physiographic units with distinct phases of restoration depending on the severity of degradation. The largest share of land restoration—over 11 million hectares, representing 37% of all restored land—was achieved in the north-central highlands (notably North Gondar and North and East Shewa). This was followed by the Tigrian plateau, which accounted for over 6 million hectares (22%) of the total restored land area in the Ethiopian highlands. The combined effects of topography, climate, and land-use pressure have produced a highly favorable environment for severe land degradation in these physiographic regions. The sub-moist to sub-humid climate regimes, characterized by brief but intense rainfall events, promote high erosivity, while the rocky plateau and escarpment landscapes increase the risk of erosion. Long-standing land-use practices, such as intensive cultivation of steep hillslopes, removal of crop residues, and extensive deforestation, have exacerbated deforestation and soil erosion in these physiographic regions, exacerbating these natural stressors. As a result, these physiographic regions align well with the areas of high to very high degradation severity in the country, as shown in the Land Degradation Monitoring and Assessment (LDMA) map of Ethiopia (RCMRD, 2018). In response, these physiographic regions have been consistently prioritized as national hot spots for land restoration since the 1970s. Following the devastating 1973–74 drought and famine that claimed the lives of millions of people, large-scale programs—including Food-for-Work, MERET, the PSNP public works, SLMP/RLLP, and, more recently, the CALM—have focused conservation, afforestation, and watershed rehabilitation efforts in these severely degraded highland zones, particularly across Amhara, Oromia, and Tigray regions. Table 4Total land restoration (million ha) across physiographic regions and reference periodsPhysiographic region2002–20092009–20162016–20222002–2022Tigrain Plaeau0.651.250.752.65North-central highlands3.455.953.9513.35Soth-western highlands0.732.521.7645.014South-eastern highlands0.562.051.574.18Hararghe Plateau0.461.120.992.57Total5.8512.899.02427.76 Total land restoration (million ha) across physiographic regions and reference periods In contrast, the southwestern and south-eastern highlands (Arsi–Bale highlands) have received less priority for large-scale land restoration and conservation programs, and hence represent 18% and 14% of the total restored area of the country, respectively. In the southwestern highlands, the dominant land-use system is agroforestry (enset-coffee tree-based homesteads), which integrates food crops, perennial cash crops, maintains continuous ground cover, enhances soil organic matter, and reduces erosion, thereby contributing to the lower degradation levels observed. The Hararghe Plateau, on the other hand, is among the most severely degraded regions and has received substantial restoration attention since the Food-for-Work interventions of the 1970 s and the PSNP/PW-based incentives for restoration of the 1980s. Despite a long history of restoration efforts, the plateau accounts for only about 9% of the total restored area nationally, primarily due to its limited geographic extent compared to other physiographic units. It is therefore important to recognize the differences in the geographical scope of these regions when comparing total restored areas across physiographic zones. The importance of policy and external shocks in land restoration As observed by (Hishe et al., 2024), the national policy context and external shocks, such as political unrest and armed conflict, can have profound impacts on land restoration implementation at the field level. While our discussion highlighted the period 2009–2016, the two adjacent periods (2002–2009 and 2016–2022) also mark distinct phases in Ethiopia’s restoration trajectory, each influenced by policy, institutional, and socio-political contexts. During the 2002–2009 period, the Government of Ethiopia, supported by development partners, scaled up watershed-based interventions through the MERET (Managing Environmental Resources to Enable Transition) and PSNP/PW (Productive Safety Net Programme/Public Works) programmes. These efforts accelerated the implementation of soil and water conservation and vegetative restoration measures, effectively doubling the cumulative area restored from approximately 3 million hectares to 6 million hectares. Importantly, this phase laid the groundwork for the subsequent design of the national Sustainable Land Management (SLM) programme, which was launched in 2008. The second period (2009–2016) represents the most rapid expansion and the most significant increase in restoration coverage. The total restored area more than doubled again, increasing from 6 million hectares to 13 million hectares. This was enabled by three reinforcing factors: (i) nationwide adoption of annual mass-labor mobilization campaigns, which engaged 15–20 million people each year in terracing, tree planting, and gully rehabilitation; (ii) strong institutionalization of the SLM programme as a flagship national initiative; and (iii) the gradual formalization of community participation mechanisms, later consolidated under the watershed user cooperative association proclamation (No. 1223/2021), which legally recognized the ownership rights of communities over restored communal lands. The third period (2016–2022) is characterized by a steady slowdown in restoration activities, especially after 2018, due to Ethiopia’s experience of political unrest, conflict in the northern regions, and the global COVID-19 pandemic, all of which disrupted labor mobilization campaigns, restricted community gatherings, and diverted national priorities. Although some restoration activities continued, the pace was significantly reduced compared to the earlier momentum. Physical structures constructed The results show that the flagship programmes and community-based watershed management efforts have enabled the construction of physical SWC structures on 22.54 million ha (Table 4). Among the physical structures, soil and stone bunds, fanya juu terraces, and deep trenches are implemented on individual farm fields, and together they constitute 56% of all SWC structures. Hillside terraces (15%), water retention structures (12%), and bench terraces (8%) are constructed on communal land (Fig. 6). Various types of check dams (gabion, brushwood, loose stone, and sandbag) are built in the gullies, erosion channels, and small streams to reduce the flow velocity of runoff and trap sediment. The water retention structures (i.e., basins and infiltration pits) are commonly built on watershed outlets, gully beds, and in tree plantation sites to collect water for use during the dry season. A disaggregated analysis of the distribution of physical conservation structures by physiographic region indicates that 52% were implemented in the north-central highlands, followed by 17% in the southwest and 16% in the Tigrian highlands. Physical SWC measures implemented in the Ethiopian highlands In comparison, the south–eastern highlands and the Hararghe plateau accounted for 10% and 5%, respectively (Table 5). The north–central highlands are prioritized due to the severe erosion previously documented by the EHRS report. This region, including areas such as Wollo, Gondar, and North Shewa, represents parts of Ethiopia that were heavily impacted by successive droughts and famines in the mid-1980s and early 1990 s, which is believed to be due to land degradation. The Tigrian plateau and the north–central highlands have been the primary targets for the flagship programs for land restoration since the inception of the Food-for-Work programs in the 1970 s, in response to the long history of land degradation and severe soil erosion in these physiographic regions. Table 5Extent of physical SWC measures implemented by physiographic region in millions of ha (2002–2022Physiographic regionBench terraceHill-side terraceFanya Juu terraceSoil bunds and deep trenchesStone bundsCheck damsWater retention structures**TotalTigrian plateau0.150.720.130.611.050.090.903.65NC highlands0.942.231.462.692.540.051.8211.73SW highlands0.370.250.420.940.810.051.093.93SE highlands0.220.080.110.670.430.040.622.17Hararghe HL0.180.070.070.320.280.030.111.06Total1.863.352.195.235.110.264.5422.54% of all SWC815102322.51.526*Micro-basins (33%); infiltration pits (29%); eyebrow basin (20%); and half-moon basins (18%). NC North Central, SW Southwestern, SE South Eastern highlands of Ethiopia. The percent (%) covered by a given SW structure is calculated as a percentage of the total area covered by all physical structures Extent of physical SWC measures implemented by physiographic region in millions of ha (2002–2022 *Micro-basins (33%); infiltration pits (29%); eyebrow basin (20%); and half-moon basins (18%). NC North Central, SW Southwestern, SE South Eastern highlands of Ethiopia. The percent (%) covered by a given SW structure is calculated as a percentage of the total area covered by all physical structures Vegetative measures for landscape regreening The vegetative measures comprise enrichment planting in the open spaces of sparse (open) forests and community woodlands (i.e., mixed tree species and shrub stands established on communal hillsides). Agroforestry integration involves planting fruit and fodder trees within individual farm fields, located within homestead areas. Through these regreening efforts, 866 million trees have been planted, with the majority (98%) being enrichment plantings (Table 5). The emphasis of enrichment planting in the sparse forests and community woodlands was on fast-growing mixed exotic trees (e.g., Grevillea robusta, Cupressus spp., and Eucalyptus spp.) to increase forest density, as well as for environmental protection. Community woodlots are areas within watersheds where communities establish a mixture of various tree/shrub species on degraded lands (usually hillsides) to meet their needs for fuelwood and construction materials. Area closures (i.e., degraded hillsides that are protected from human and livestock interference) are the most dominant forms of vegetative restoration implemented. Over 2 million hectares of degraded land have been treated through area closures, leading to significant improvements in vegetative cover (Table 6). Table 6Vegetative SWC measures implemented in the Ethiopian highlands (2002–2022)Type of practiceMillions of seedlingsArea covered in M haArea closures on degraded hillsides-2.07Enrichment planting in a sparse forest8411.25Agroforestry integration on farm fields251.76Total8665.08Source: MoA SWC database; analysis by authors Vegetative SWC measures implemented in the Ethiopian highlands (2002–2022) Source: MoA SWC database; analysis by authors Land use/land cover change analysis Data on spatial and temporal changes in land use/land cover types for the four reference periods are presented in Fig. 7 and Table 7. The data indicate that there have been significant changes in land use, particularly between 2009 and 2016. The area of forest land expanded from 2.6 million ha (5.9%) in 2002 to nearly 6 million ha (13%) in 2022, while the area of woodlands grew by 28% (Table 6). The establishment of area closures played a key role in the threefold reduction of bare land, decreasing from 2.5 million ha in 2002 to nearly 830,000 ha in 2022 (Table 7). On the other hand, cropland expanded from 17.4 million hectares (39%) in 2002 to nearly 19 million hectares (41%) in 2022. The LU/LC change matrix (Annex 5) shows that crop area is expanding primarily at the expense of shrublands, grasslands, and woodlands, which are converted mainly to crop lands, settlements, or forests under the restoration programs. Certain areas of bare land (4.8%) were converted into cropland (Annex 5) due to the SWC interventions that reclaimed abandoned lands for productive use. Land use/land cover map of the Ethiopian highlands (2002–2022) Table 7LU/LC change during the reference periods in the Ethiopian highlandsLand use types2002200920162022Area (ha)%Area (ha)%Area (ha)%Area (ha)%Forest land2,656,9175.903,035,9536.745,818,21912.915,871,76313.03Wood land4,712,62410.466,134,53913.616,281,23613.946,294,85713.97Shrub/bush land13,401,58029.7412,691,74528.176,668,74214.806,642,61314.74Crop land17,419,16838.6617,526,19438.8918,563,87841.2018,598,36741.27Grass land5,418,95312.034,274,6989.494,574,81310.154,506,84410.00Bare land2,500,0485.552,491,7435.53946,1742.10830,7431.84Wetland19,9510.0440,5470.0961,2010.1460,7690.13Water bodies387,5090.86396,1520.88396,0900.88395,7200.88Settlement98,1600.22130,4650.29196,8090.44198,3590.44 LU/LC change during the reference periods in the Ethiopian highlands The land use change matrix across physiographic regions (Annex 6) and for four reference periods suggests the strong interplay between degradation processes and restoration interventions across Ethiopia’s diverse physiographic settings. This includes cropland expansion, vegetation recovery, and forest area expansion, as well as a decline in bare land with variability across physiographic regions. The north-central and southwestern highlands showed the most dynamic land use/land cover transitions, with significant expansions of forest and woodland occurring at the expense of shrubland. In contrast, the Hararghe Plateau exhibited relatively modest and localized shifts, with more minor net changes across most land-use categories. In all physiographic regions, shrubland consistently declined, indicating its conversion into cropland, settlements, or forestland under restoration programs. Cropland and settlement areas steadily increased, reflecting demographic pressure and agricultural expansion. The expansion of vegetation recovery in the Hararghe highlands is presumably due to Catha edulis plantations—a bushy tree species cultivated for its succulent leaves, which are consumed as a stimulant throughout the region and exported to Djibouti and the Middle East (Elias et al., 2022). Recovery in Tigray is primarily due to the widespread practice of area closures over the past many decades. Accuracy assessment and kappa coefficient Accuracy assessment results show high classification reliability, with an overall accuracy of 90% and a Kappa coefficient of 0.90 (Table 8), values that demonstrate strong agreement between the classified images and the reference data, and fall within the acceptable and robust range for remote sensing studies. Table 8Accuracy assessment and kappa coefficient of land use/land cover change analysisLand use typeFLWLSHLCLGLBLWLWBSsumUser’s accuracyForest land (FL)81782151081861289691.18Woodland (WL16803171896141689090.22Shrubland (SL)1768191791646089491.61Cropland (CL)61316801168197989589.5Grassland (GL)18115178217191390291.02Barren land (BL)31361298093221789490.49Wetland (WL)167617143811151790689.51Waterbody (WB)3193221331823789492.06Settlement (S)21413119181382990092.11Sum8988948719199198818828979108071Producer’s accuracy908994878991919191 Accuracy assessment and kappa coefficient of land use/land cover change analysis The NDVI results There was a gradual increase in the NDVI values from 0.30 (2002) to 0.38 (2022), indicating positive changes in vegetation cover and environmental health (Fig. 8). This shows a significant regreening of the landscape through an increase in vegetation cover and a corresponding reduction in bare land areas, which is in line with the LU/LC analysis. Table 9 details the tree planting initiatives of flagship programs, such as PSNP and SLMP, which have helped plant nearly a billion tree seedlings, primarily between 2009 and 2016. Additionally, area closures and agroforestry integration on farm fields have likely contributed significantly to the environmental greenness. Reforestation, in the form of enrichment planting with millions of seedlings, was conducted in open gaps of sparse forests and community woodlots, thereby enhancing tree cover and increasing the density and structure of forests. As a result, open forests are shifting to more closed forests (i.e., continuous canopy cover of trees with their crowns overlapping). NDVI map of Ethiopia showing a much greener landscape in recent years Table 9Enrichment planting (millions of seedlings planted) by flagship programmesFlagship Programmes2002–20092009–20162016–2022TotalSLMP10.4318.93271.04300.40PSNP45.644.88351.59442.07MERET38.6259.990.0098.61Total94.65123.8622.63841.08 Enrichment planting (millions of seedlings planted) by flagship programmes Soil erosion and sediment yield by land use type and by physiographic regions Table 10 presents soil loss by land use types across reference periods. The data show that soil loss rates have declined significantly across land use types from 2002 to 2022. The mean soil loss across all land use types declined from 32 to 24 t/ha/year, demonstrating a highly significant (p < 0.001) effect of restoration measures. The lowest mean soil loss reduction was for forest land (from 2 to 1.8 t/ha) and woodland (from 6 to 4.5 t/ha/year). At the same time, grassland and shrub/bushlands experienced moderate but significant declines, with soil loss reduced from 19 to 13 t/ha/year and from 29 to 22 t/ha/year, respectively, reflecting the role of vegetative cover in mitigating soil erosion. Croplands showed the most significant improvement in soil loss, reduced by nearly 50% (i.e., from 69 to 35 t/ha/year) due to the large-scale implementation of soil and water conservation measures, such as soil/stone bunds and fanya juu terraces. The most prone land use type to erosion, bare land, on the other hand, only slightly decreased (from 91 to 85 t/ha/year). Table 10Soil loss by land use type and across reference periods in the Ethiopian highlands (mean ± standard error)Land use type2002200920162022P valueForest2.00 ± 0.101.91 ± 0.101.85 ± 0.091.80 ± 0.09 < 0.01Woodland6.21 ± 0.325.00 ± 0.264.67 ± 0.244.47 ± 0.23 < 0.001Grassland19.03 ± 0.9720.00 ± 1.0214.09 ± 0.7212.82 ± 0.65 < 0.001ShrubLand28.80 ± 1.4627.05 ± 1.3724.00 ± 1.2222.09 ± 1.12 < 0.001Cropland69.11 ± 3.5367.30 ± 3.4445.97 ± 2.3535.21 ± 1.80 < 0.001Bare Land91.02 ± 4.6387.32 ± 4.4485.81 ± 4.3684.51 ± 4.29 < 0.001Wetland9.00 ± 0.468.50 ± 0.438.00 ± 0.417.40 ± 0.38 < 0.001Mean32.26 ± 0.5231.01 ± 0.4525.91 ± 0.4024.04 ± 0.29 < 0.001P value < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 Soil loss by land use type and across reference periods in the Ethiopian highlands (mean ± standard error) Figure 9a–d demonstrates the value of vegetation cover in soil erosion control—areas with dense vegetation cover, such as forests (Fig. 9a) and woodland (Fig. 9b), experience significantly lower soil erosion rates compared to bare lands (Fig. 9c) and crop lands (Fig. 9d), especially at initial stages when there is no or little vegetation cover. With conservation intervention, however, there are considerable positive environmental trends in soil erosion rates over time. For example, in Fig. 9c, both the extent of bare land and the rate of soil erosion declined over time (2002–2022). This can be attributed to the effective expansion of area closures, enrichment planting, and gully stabilization (i.e., check dams) on formerly bare lands. Similarly, in Fig. 9d, the initially higher rates of soil erosion in croplands declined despite cropland expansion due to the implementation of conservation practices that helped stabilize the soil, reduce runoff, and increase water infiltration. The relationship between land use type and soil erosion: soil erosion in forest land (a), soil erosion in woodland (b), soil erosion in bare land (c), and soil erosion in crop land (d) The data show that there are significant variations in soil erosion among the physiographic units, with the north-central highlands experiencing the highest mean soil loss (41.00 t/ha) in 2022, followed by the Tigrian plateau (30.20 t/ha). This is in line with the fact that rugged terrain with steep side slope escarpments is familiar (Table 11). Figure 10 indicates the variation in soil loss severity across different physiographic regions, highlighting notable improvements in recent years. The sediment export data is a mirror image of the erosion from different land use types, with the highest mean sediment load in the bare land (12.37 t/ha), followed by crop land (3.50 t/ha), and an overall mean of 3.86 t/ha/year (Table 12). Considering the 54 million hectares in the entire highlands, the total sediment export that ends up in water bodies is approximately 208 million tonnes per year. The spatial trend analysis indicated that soil loss and sediment export had declined, as shown in Fig. 10. Table 11Estimated soil loss by physiographic regions of Ethiopia and reference yearsPhysiographic regionsMean soil loss (t/ha)2002200920162022P valueTigrian plateau40.00 ± 2.1239.00 ± 2.0131.95 ± 1.8830.20 ± 1.63 < 0.001North-central highlands45.20 ± 1.9043.00 ± 1.8142.00 ± 1.4341.00 ± 1.21 < 0.001Southwestern highlands24.00 ± 0.8023.00 ± 0.2319.00 ± 0.2116.00 ± 0.20 < 0.001South-Eastern highlands21.37 ± 0.9120.60 ± 0.7214.80 ± 0.5415.00 ± 0.33 < 0.001Hararghe plateau30.26 ± 0.6729.45 ± 0.4521.80 ± 0.3018.00 ± 0.22 < 0.001Mean32.26 ± 1.2231.01 ± 1.1025.91 ± 1.3124.04 ± 1.10 < 0.001P value < 0.001 < 0.001 < 0.001 < 0.001 Estimated soil loss by physiographic regions of Ethiopia and reference years Soil loss trends by physiographic regions of the Ethiopian highlands Table 12Sediment export in different land use types in the Ethiopian highlandsSediment export (t/ha) by land use typeLand use type2002200920162022P valueForest land0.14 ± 0.010.15 ± 0.020.09 ± 0.010.05 ± 0.01 < 0.01Woodland0.56 ± 0.410.78 ± 0.430.56 ± 0.220.48 ± 0.11 < 0.01Grassland1.10 ± 0.811.73 ± 0.911.17 ± 0.210.76 ± 0.11 < 0.001Shrub/bush2.90 ± 0.563.29 ± 0.712.04 ± 0.561.70 ± 0.31 < 0.001Cropland5.32 ± 1.105.97 ± 1.154.08 ± 1.053.50 ± 0.91 < 0.001Bare land17.05 ± 1.7123.45 ± 2.2315.95 ± 1.4312.37 ± 1.02 < 0.001Wetland0.01 ± 0.011.04 ± 0.301.24 ± 0.410.65 ± 0.47 < 0.01Mean3.87 ± 0.395.20 ± 0.523.59 ± 0.392.79 ± 0.28 < 0.01P value < 0.01 < 0.01 < 0.01 < 0.01 Sediment export in different land use types in the Ethiopian highlands Discussion Greener landscape now than in the past Both the LU/LC analysis and the NDVI results confirm that the highlands of Ethiopia are greening more now than in previous decades, primarily due to conservation and reforestation interventions across 28 million hectares. Our findings on land use change favoring forest cover and vegetation recovery are consistent with previous studies. Abera et al. (2020) reported that Ethiopia has undertaken one of the most extensive land restoration programs in Africa, with massive investments in sustainable land management (SLM) that promoted the adoption of Soil and Water Conservation (SWC) technologies (Woldearegay et al., 2017). Restoration efforts of this scale not only rehabilitate degraded landscapes but also enhance key ecosystem services, particularly carbon sequestration (Bastin et al., 2019). Moreover, afforestation on degraded lands has been shown to trigger positive ecological shifts, including improved microclimates and increased biodiversity (Myers–Smith et al., 2011). Discussion with community watershed user cooperatives (CWUCs) members indicated that they encourage fast regeneration of the closures by means of building terraces and micro-basins for soil and moisture conservation. According to Biratu & Asmamaw, 2016, collective participation of farmers in soil and water conservation activities makes them increase their household income. The LU/LC change matrix reveals significant changes in land cover types, particularly between 2009 and 2016, indicating remarkable vegetation regeneration resulting from conservation and restoration interventions. There was a doubling of forestland, a significant expansion of woodland, and a threefold reduction of bare lands during 2002–2022. These outcomes can be attributed to the integration of agroforestry, area closures, and enrichment planting in open forests and shrub lands, driven by flagship programs such as the PSNP and SLMP, as well as massive tree planting campaigns under the Green Legacy Initiative (GLI). In addition, the country has been advancing the REDD + programme (Reduced Emissions from Deforestation and Forest Degradation, plus conservation) to promote sustainable forest management and enhance forest carbon stocks (Ebissa et al., 2023). These results are in agreement with previous studies on vegetation restoration in the Ethiopian highlands (Hirvonen et al., 2022; Lambert & Deyganto, 2024; Weldegebriel et al., 2025). For example, a study in the Omo-Gibe basin of Ethiopia’s humid highlands showed that landscape restoration activities, such as area closures, resulted in a 35%–100% reduction in bare land area, more than doubling the forest and woodland area, suggesting a major ecological restoration (Wolka et al., 2023). The national forestry initiative under the CRGE framework aims at protecting existing forests through participatory forest management (PFM) and re-establishing forests through aggressive tree planting for their ecosystem services (carbon sink, erosion control, hydrological regulation) (FDRE, 2015). As a result, the previously reported alarming deforestation rate of 100,000 ha per year in the 1980 s (FAO, 1986) has declined to just 27,000 ha per year (Ebissa et al., 2023; EFD, 2021). Additionally, the Green Legacy Initiative (GLI) of Ethiopia is the continent’s most extensive tree-planting campaign. It has emerged as a key driver of these improvements, contributing to the expansion of forest cover, control of soil erosion, and adaptation to climate change in the Ethiopian highlands (Kassa et al., 2022; Lambert & Deyganto, 2024). The tree seedling survival rate is generally considered high in the Ethiopian highlands, ranging from 68% to 100%, depending on the species, site conditions, and management practices. For instance, a 2024 study in Northern Ethiopia reported a seedling survival rate of around 68% (Berhe et al., 2024), while the GLI seedling survival rate averaged 83% due to careful site selection and management (Beyene & Shumetie, 2023). A recent study by Roba et al. (2025) demonstrated significant gains in forest cover and a reduction in barren lands in the Upper Awash Basin, further underscoring the effectiveness of the GLI in reversing land degradation. Other studies comparing the current environmental status of the Ethiopian highlands with historical aerial photos and satellite imagery, as well as terrestrial photographs, also show that the northern Ethiopian highlands are greener now than at any time in the past, mainly due to terracing and reforestation efforts, especially since the 1980 s (Kropáček et al., 2019; Nyssen et al., 2014, 2016). Soil erosion control and sediment yield reduction Following the release of the EHRS report (FAO, 1986), soil and water conservation has been a central policy priority in Ethiopia, achieved through mass community mobilization and flagship programs. Extensive conservation structures—such as soil and stone bunds, fanya juu terraces, and micro-basins—were implemented on crop fields alongside agroforestry practices as erosion rates have declined markedly. Compared to the EHRS baseline (FAO, 1986), the current soil erosion rate in croplands (54 t/ha) is less than half the 1986 estimates (130 t/ha), while soil erosion in bare lands (87 t/ha) is more than threefold lower than the 300 t/ha previously reported as a result of gully rehabilitation structures constructed. Likewise, the mean sediment export has dropped dramatically, from 35 t/ha (1.9 billion tons) to 2.79 t/ha (≈208 million tons) across the highlands. We argue that these reductions were achieved through the accelerated construction of bunds and terraces, coupled with tree/grass planting, and area closures on bare lands, mainly during 2009–2016. Other studies also reported similar trends. For example, (Haregeweyn et al., 2015) reported an average reduction of 65% in soil erosion where SWC measures were implemented. In comparison, the establishment of area closures on hillsides resulted in an average reduction of 71% in runoff and 69% in soil erosion (Haregeweyn et al., 2015). Another basin-wide study in the Omo-Gibe River basin demonstrated that soil bunds effectively reduce surface runoff by 80–92% and reduce soil erosion by approximately 96% on crop lands (Wolka et al., 2021). These findings indicate that soil and stone bunds are commonly constructed by farmers, who widely build them on crop fields, and have become an important component of farmland management in Northern Ethiopia (Nyssen et al., 2009). Based on these observations, we argue that EHRS’s soil erosion estimates for cultivated fields (130 t/ha) were significantly overestimated, which warrants further investigation. Subsequent studies (Bojo & Cassels, 1995; Hurni, 1988) provided more reliable data, showing lower erosion rates (70 t/ha/yr) when measurements were scaled appropriately. Overestimation in the EHRS is likely attributed to data extrapolations from small runoff plots that were aggregated for the entire highlands (Bojo & Cassels, 1995; Elias & Scoons, 1999). Limitations of the InVEST SDR model and NDVI The findings on soil loss and export from the InVEST-SD outputs model demonstrate its effectiveness and reliability in estimating soil erosion and sediment delivery in the Ethiopian highlands. The model integrates biophysical inputs (rainfall, soil erodibility, land cover, topography, and conservation practices) and applies a spatially explicit framework, which makes it particularly suitable for data-scarce regions like Ethiopia (Gashaw et al., 2021). The model has been widely applied for estimating soil loss and sediment delivery at the watershed scale and has been successfully used in diverse geographic contexts (Aneseyee et al., 2020; Tamire et al., 2022; Tsegaye et al., 2024). A limitation of the InVEST-SDR model, however, is that it was primarily designed to capture sheet and rill erosion processes, and does not explicitly represent gully erosion. In many highland landscapes in Ethiopia, on the other hand, gully erosion contributes substantially to the total soil loss and sediment yield. While the model simplifies some processes by not fully accounting for gully erosion or channel processes, it has proven reliable for capturing relative spatial differences and identifying areas of erosion. Therefore, the data and figures presented in this paper are reliable first-order estimates that are well-suited for informing policy dialogue, guiding management interventions, and identifying priority areas for land restoration in accordance with the EHRS report (FAO, 1986). Future research, however, should emphasize that detailed validation through field measurements, including setting up runoff plots, sediment, and hydrological monitoring at the gauging stations, and systematically comparing treated and untreated catchments across physiographic regions, remains necessary to improve accuracy and strengthen confidence in absolute values. While these steps are beyond the scope of the current study, they represent an important direction for future research to strengthen the reliability of soil erosion and sediment yield estimates in the Ethiopian highlands. The NDVI has been widely adopted as a proxy for vegetation greenness and canopy cover; however, it also has certain limitations. According to Huang et al. (2021), a significant limitation of the NDVI is that it is prone to atmospheric effects, sensor differences, and variations in illumination conditions, which can introduce variability and reduce comparability across sensors and time periods. Unlike land-use classifications, which can be validated against ground-truth reference data, NDVI lacks precise black-and-white reference data for validation (Shao et al., 2019). NDVI is often used as a dependent variable to predict other vegetation attributes that cannot be detected directly with remote sensing imagery (Box et al., 1989; Loranty et al., 2018). Acknowledging these caveats not only strengthens the credibility of current findings but also highlights the need for complementary methods—such as ground-based vegetation surveys and improved parameter calibration—to refine future research. Conclusion The paper provides a comprehensive assessment of land degradation and restoration trajectories in the Ethiopian highlands by systematically exploring both temporal and spatial dynamics. As a point of departure, the study explores spatial diversity by disaggregating restoration patterns across distinct physiographic regions with highly variable topographic and agro-climatic conditions, LU/LC, and farming system settings. The temporal dynamics capture critical policy and institutional milestones, such as the formulation of the Community Watershed User Cooperative proclamation, which accelerated land restoration between 2009 and 2016. Additionally, they account for external shocks including conflict and the COVID-19 pandemic that slowed community action in subsequent years. Combined with the methodological integration of large-scale input datasets for the InVEST-SDR model, including land use/land cover change, conservation and management measures, and rainfall data disaggregated by physiographic region, these analyses produced robust, policy-relevant evidence on vegetation restoration, soil loss reduction, and sediment retention. With nearly 28 million hectares of degraded landscapes restored through community mobilization and flagship programs, the findings show that Ethiopia has made unparalleled strides in reversing land degradation over the past 20 years. The amount of bare land has decreased threefold, woodlands have grown, and forest cover has doubled. These achievements show the synergy between bottom-up community-based natural resource management and top-down policy support. From a policy perspective, the study reinforces the central role of a community-based approach, underscoring the role of Community Watershed User Cooperatives as stewards of restored landscapes and highlighting the importance of embedding restoration in broader frameworks of rural livelihoods. Donor-financed flagship programs, working in conjunction with community mass mobilization, have played a pivotal role in scaling up land restoration efforts by providing technical and financial support that ensured sustained implementation across the Ethiopian highlands. Concerted restoration work remains particularly critical in degradation hotspots in the north-central highlands and the Tigray plateau, where vast areas of land continue to suffer from severe degradation despite the tremendous progress achieved. Ethiopia’s experience underscores the importance of a holistic approach to environmental conservation, combining flagship programs and a community-based natural resources management approach. On a broader scale, Ethiopia’s large-scale land restoration initiatives are contributing toward the fight against climate change and are directly aligned with some key SDGs: SDG 2—Zero Hunger (increase productivity of the land); SDG 13—Climate Action (forest expansion, restoration of degraded landscape to combat climate change and its impacts), and SDG 15—Life on Land (protection and sustainable use of terrestrial ecosystems, combat land degradation and desertification), among others. Supplementary information Below is the link to the electronic supplementary material. (DOCX 259 KB) Acknowledgements The authors would like to thank the Ethiopian Meteorological Institute for providing the climatic data. The Ministry of Agriculture should be thanked for providing the conservation data. We want to thank the University of Agriculture in Krakow, Poland, for providing professional language correction. Author contribution E. E Conceptualization, Investigation, Methodology, Writing – Original Draft, Supervision. A.B E Methodology, prepared all figures, revised and edited the manuscript, A.M Remote sensing data analysis, Data analysis, and supervision. H.H. Field data collection and supervision. H.H. Data collection, validity of the data, and editing and supervision. J.H. editing and supervision. H.Y. Data analysis and supervision, editing and supervision. Funding Not applicable Data availability The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request. Declarations Ethical responsibilities of authors All the authors have read, understood, and complied as applicable with the statement on “Ethical Responsibilities of Authors” as found in the Instructions for Authors. Animal ethics and consent to participate Not applicable. Consent for publication The authors certify that the publisher is permitted to publish this work. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials (DOCX 259 KB) Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.