1 Introduction

Soil erosion caused by a loss of vegetation cover in highland areas is the most serious environmental problem worldwide (Gobin et al. 2004; Pimentel 2006; Borrelli et al. 2017). Rainfall intensity, soil erodibility, poor soil conservation practices, and land use land cover (LULC) change is a challenge in Ethiopia, particularly in the highlands where highly rugged topography, and cultivating on steep slope lands, which accelerate soil erosion (Shiferaw 2011; Belayneh et al. 2019; Olika and Iticha 2019; Tsegaye 2019; Aneseyee et al. 2020; Nut et al. 2021). The removal of fertile soil is aggravated by an increase in surface runoff and a decrease in vegetation cover from steep slopes (WoldeYohannes et al. 2018; Kidane et al. 2019). The LULC change and ridged topography are the most important key parameters influencing soil erosion, particularly in Ethiopian highlands (Bewket and Abebe 2013; Gessesse et al. 2015; Moisa et al. 2021; Negash et al. 2021). According to Li et al. (2020), the rate of soil erosion increases as the catchment slope length and gradient increase. In areas where soil erosion is a risk, immediate action is required to implement soil and water conservation strategies (Haile et al. 2006; Molla and Sisheber 2017; Girmay et al. 2020; Moisa et al. 2021). The expansion of agricultural land to the highlands is a major cause of soil erosion in Ethiopia (Sisay et al. 2014; Kidane and Alemu 2015). Earlier studies (Gashaw et al. 2017, 2018, 2019) have reported that the expansion of cultivated land around steep slopes is the major driving force for soil erosion in the Ethiopian highlands. Every year, approximately 3.5 billion tons of top soil is removed due to soil erosion in the Ethiopian highlands, and soil erosion rates in Ethiopia are extreme (Kidane and Alemu 2015).

There are three soil erosion models, which include conceptual-, physical-, and empirical-based model (Santos et al. 1998). The conceptual models emphasized on the process governing system behaviors that makes unique from physical and empirical models (Beck 1987). The conceptual models focused on catchment processes, without considering various interactions (Renschler 1996). The physical-based models focused on the understanding of the physics of erosion process and the parameters used are measurable and known. In physical-based model, the parameters should be calibrated against observed data (Beck 1987). The empirical models are based on simulation of natural processes (Adongo et al. 2019) and mathematically simple to use but they are limited to the area where they have been developed (Santos et al. 1998). The revised universal soil loss equation (RUSLE) model, developed by Renard et al. (1997) was widely used due to relatively easy to use. The compatibility with geographic Information System (GIS) and remote sensing techniques is another advantage of using RUSLE model in soil erosion loss assessment and prioritization of hotspot areas for conservation action (Moisa et al. 2021).

Ethiopia has 60 million hectares of agricultural land (Douglass 1984; Sisay et al. 2014). Approximately 27 million and 14 million hectares of agricultural land have been highly and extremely eroded (Lema et al. 2016; Tilahun et al. 2018). In Ethiopia’s highlands, agricultural land has a high risk for soil fertility decline (Habtamu et al. 2014; Balabathina et al. 2020). In Ethiopia, soil loss from cultivated fields is estimated to be approximately 42 ton/ha/year (Hurni et al. 2008; Bekele and Gemi 2021). According to Hurni et al. (2015) and Atoma et al. (2020), western Ethiopia has the highest rate of soil erosion. The Upper Anger watershed is located in western Ethiopia and is known for its undulating terrain with rolling plains and valleys, agricultural expansion, deforestation, overgrazing, land fragmentation, and poor land management practices, all of which have exacerbated the situation.

Although the consequences of soil erosion as a result of declining vegetation cover in elevated areas were severe, it received little research attention. Several studies have been conducted on the impact of LULC change on soil erosion in Ethiopia (Kidane et al. 2019; Gashaw et al. 2019; Aneseyee et al. 2020; Moisa et al. 2021, 2022c; Negash et al. 2021; Negese 2021), but these studies did not address the impacts of vegetation cover losses from elevated areas on soil erosion. The southwestern parts of the country do not get the attention of scholars because of its relatively good in vegetation cover as compared to other parts of Ethiopia. Even though, some areas are covered by vegetation, substantial areas are highly exposed to deforestation driven by agricultural land expansion. The loss of vegetation cover aggravates the problem of soil erosion in southwestern parts of Ethiopia which in turn leads to decline in agricultural productivity and increased the problem of food insecurity. This problem needs research attention to sustain communities’ wellbeing on one hand, and on the other hand, providing robust evidence on the linkage between deforestation and soil loss can alert local administration and decision-makers to actively engaged in natural resources conservation in general and soil erosion protection in particular, which may increase agricultural productivity. Furthermore, this study analyzes the correlation between normalized difference vegetation index (NDVI) and slope with mean soil loss in the study area. Finally, this study indicates the effectiveness of RUSLE model and geospatial technologies to estimate the impact of vegetation cover loss on soil erosion in the highland areas of Ethiopia.

2 Materials and methods

2.1 Descriptions of the study area

This research was conducted in the Upper Anger watershed which is situated in the Abay River basin. The study area was located between the East Wollega Zone and the Horo Guduru Wollega Zone of Oromia National Regional State, Western Ethiopia (Fig. 1). Geographically, the Upper Anger watershed lies between 9°27′30″ and 10°23′00″ N and 36°39′00″ and 37°73′00″ E and the elevation of the study area lies between 1330 and 3171 m above sea level. It covers an area of 1869.1 km2.

Fig. 1
figure 1

Map of the study area

Full size image

2.2 Data sources and data types

In the present study, the required data were obtained from Landsat TM, Landsat ETM + and Landsat OLI/TIRS for the years of 1989, 2002 and 2020, respectively from the U.S. Geological Survey (USGS) website (https://earthexplorer.usgs.gov/). Rainfall data from the National Meteorological Agency of Ethiopia, and soil types from digital soil map of Ethiopia developed by Food and Agriculture Organization (FAO), 30 m resolution of digital elevation model (DEM) from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) website https://asterweb.jpl.nasa.gov/gdem.asp of USGS were used for this study (Fig. 2). Materials used for the study included: Handled GPS and Digital Camera as well as software, such as ArcGIS 10.3, ERDAS 2015, Arc SWAT and Google Earth pro, were used for the analysis.

Fig. 2
figure 2

Methodological flowchart of the study

Full size image

2.3 Data analysis

2.3.1 Land use land cover change analysis

In classifying the images, supervised image classification with the maximum likelihood algorithm was used to categorize different LULC types (Khatami et al. 2016). Images were classified into agricultural land, forest land, grassland, bare land and settlement using the ERDAS Imagine 2015 software.

2.4 Estimation of soil loss

To estimate the amount of soil loss in the Upper Anger watershed, the RUSLE model with integration of the Geographic Information System (Wischmeier and Smith 1978) was adopted to estimate the annual soil loss (Eq. 1).

$$A = R*K*{\text{LS}}*C*P$$
(1)

where \(A = {\text{average soil loss per unit area in ton}}/{\text{ha}}/{\text{year}}\), R = rainfall–runoff erosivity factor in MJ mm ha−1 h−1 year−1, \(K = {\text{soil erodibility factor}}\) in ton ha/MJ/mm, \({\text{LS}} = {\text{slope length and steepness}}\), \(C = {\text{cropping and management systems}}\), \(P = {\text{conservation practices}}\)

2.4.1 Rainfall erosivity (R) factor

The erosivity factor reflects the effects of raindrops on soil erosion (Ganasri and Ramesh 2016; Napoli et al. 2016; Lal and Elliot 2017; Ouyang et al. 2018; Koirala et al. 2019). Therefore, the erosivity factor R was calculated according to the equation given by Hurni (1985), derived from a spatial regression analysis for Ethiopian conditions based on the available mean annual rainfall data (Eq. 2).

$$R = { } - 8.12 + \left( {0.562{*}P} \right)$$
(2)

where P is mean annual rainfall in (mm).

Seven metrological stations (Shambu, Anger Gute, Sibu Sire, Nekemte, Haro, Chagni and Debremarkos) with mean annual rainfall of 31 years were used to calculate the R factor. The annual rainfall was interpolated from each station using Inverse Distance Weighting (IDW) interpolation techniques (Fig. 3).

Fig. 3
figure 3

Annual rainfall and R factor map

Full size image

2.4.2 Soil Erodibility (K) factor

The K factor is used to estimate the vulnerability of soil to erosion (Prasannakumar et al. 2012; Atoma et al. 2020; Nasidi et al. 2020). The susceptibility of soil to erosion varies based on the soil types (Gayen et al. 2020). The value of K was estimated using soil color which adopted by Hurni (1985) for Ethiopian condition (Table 1). The FAO digital soil map of Ethiopia were used for this study, which was obtained from Oromia Water Works Design and Supervision Enterprise (OWWDSE) and used to develop the soil map for soil erosion risk analysis (Fig. 4). Then, erodibility value (K factor) is assigned for each of the soil types based on their colors according to (Hurni 1985; Gelagay and Minale 2016; Gashaw et al. 2017; Esa et al. 2018; Desalegn et al. 2018).

Table 1 Soil types and K factor
Full size table
Fig. 4
figure 4

Soil types and K factor map

Full size image

2.4.3 Slope length and steepness (LS) factor

In this study, the slope length and steepness factor were determined using digital elevation model (DEM) with 30 m resolution (Fig. 5). Based on FOA (2006) slope gradient classes (Table 2), we re-classified slope class of the Upper Anger watershed by considering great topographic variation (1330–3171), and LULC dynamics. Accordingly, the slope of the study area was classified as: flat to almost flat terrain (0–2%); gently flat to undulating terrain (2–10%); rolling terrain (10–15%); hilly terrain (15–30%), and steep dissected to mountain terrain (> 30%).

Fig. 5
figure 5

LS factor map of the study area

Full size image
Table 2 Slope gradient classes (FOA 2006)
Full size table

The LS is calculated using (Eq. 3). LS is the ratio of observed soil loss related to the soil loss of standardized plot (22.13) as developed by Moore and Burch (1986), and used by (Ostovari et al. 2017).

$${\text{LS}} = {\text{Flow }}\;{\text{accumulation}}\, X \left( {\frac{{{\text{Cell }}\,{\text{size}}}}{22.13}} \right)^{0.4} X \left( {\frac{{\sin \,{\text{Slop}}}}{0.0896}} \right)^{1.3}$$
(3)

where Flow accumulation represents the contribution of an area accumulated upslope for a given cell, LS is the combination of the slope length and slope steepness factor, cell size refers to the size of the grid cell (for this study, the specific DEM is 30 m pixel size) and the sin slope is the slope degree value in sin.

2.4.4 Cover management (C) factor

Cover management (C factor) is used to estimate the relative influence of the management approach on the conservation plan (Fayas et al. 2019; Almagro et al. 2019) which was calculated from LULC types in the study area. The Upper Anger watershed was classified into five LULC types and the C factor value was assigned (Table 3; Fig. 6) based on the existing literature (Moisa et al. 2021; Negash et al. 2021).

Table 3 C factor and P factor of the study area
Full size table
Fig. 6
figure 6

C factor map of the study area

Full size image

2.4.5 Erosion control practice factor

The P factor signifies the ratio of soil erosion from a land treated with a specific conservation measure to its equivalent soil loss from up and down slope tillage (Esa et al. 2018; Belayneh et al. 2019). The P factor for different land use categories was assigned based on the existing literature (Fig. 7) as previously used by (Fayas et al. 2019; Olika and Iticha 2019; Prasannakumar et al. 2012).

Fig. 7
figure 7

P factor map of the study area

Full size image

2.5 Normalized difference vegetation index (NDVI)

The NDVI was calculated from Landsat images of multi-spectral bands and used to calculate the extent to which vegetation covers the earth’s surface (Tran et al. 2017). Band 4 was used to measure near-infrared bands on Landsat 5 and 7, and band 5 was used on Landsat 8. The red bands of the Landsat data were measured using band 4 for Landsat 8 and band 3 for Landsat 5 and 7 (Asare et al. 2020). In this study, NDVI was calculated using red bands and near-infrared images from Landsat 5 in 1990, Landsat 7 in 2000, and Landsat 8 in 2020. High reflectivity in the near-infrared (NIR) region indicates healthy vegetation, while low reflectance in the red band indicates stressed vegetation (Mahajan and Bundel 2016). This index's formula is presented in (Eq. 4)

$${\text{NDVI = }}\frac{{\text{ NIR - Red}}}{{\text{NIR + Red}}}$$
(4)

The NDVI scale spans from − 1.0 to 1.0 (Mahajan and Bundel 2016). For health and dense vegetation, the NDVI values are always between 0.2 and 0.9 (Ahmed and Akter 2017). Land cover such as rock, water, and barren plains, on the other hand, are represented by values less than 0.1 (Ju and Masek 2016).

3 Results and discussion

3.1 Land use land cover of the study area

Land use land cover classes in the study area were classified into agricultural, forest, grassland, bare land and settlement during the study period (1989, 2002 and 2020) (Fig. 8). The results show that, agricultural land was the most dominant LULC types with an area of 846.9 km2 (45.3%), 1245.8 km2 (66.7%) and 1554.3 km2 (83.2%) during 1989, 2002 and 2020, respectively. This clearly revealed that agricultural lands were growing significantly over the study period. A recent study by Moisa et al. (2022a) in the Anger River Sub-basin reported the increasing trends of agricultural land between the year 1990 and 2020. The results revealed that forest land shows a decreasing trend from 1989 to 2020 due to the encroachment of agricultural land and other underlying factors. Results show that forest cover in 1989 was about 380.2 km2 (20.3%) and declined to 342.0 km2 (18.3%) in 2002 and then rapidly reduced to 200.3 km2 (10.7%) in 2020 (Table 4). Bare land and settlement cover classes showed an increasing trend. For instance, the bare land has been increased from 3.9 km2 in 1989 to 18.4 km2 by the year 2020. Similarly, the settlement increased from 1.3 km2 in 1989 to 3.6 km2 and 5.6 km2 in the year 2002 and 2020, respectively.

Fig. 8
figure 8

LULC types of the study area between 1989 and 2020

Full size image
Table 4 LULC change of the study area
Full size table

3.2 Land use land cover conversion between 1989 and 2020

The land use transfer matrix (LUTM) method was used to analyze the LULC change over the study period (1989 to 2020) as recently used by Moisa et al. (2021). The results show that about 530 km2 of grassland land cover was converted to agricultural land, while about 241.5 km2 of forest cover was converted to agricultural land between 1989 and 2020 (Table 5). The high conversion of forest cover to agricultural land has been reported by Negassa et al. (2020) in Komto protected forest priority area from the years 1991 to 2019. Other study by Moisa et al. (2022b) found that vegetation covers substantially declined due to agricultural expansions. The spatial distribution of the major LULC conversion is presented in (Fig. 9).

Table 5 LULC change matrix of the study area (km2)
Full size table
Fig. 9
figure 9

Spatial distribution of LULC conversion map between 1989 and 2020

Full size image

3.3 Topography of the study area

Similar to other highland areas in southwestern parts of Ethiopia, the study area was dominated by hilly terrain with an area of 718.3 km2. Next to hilly terrain, study area situated in gently flat to undulating terrain and rolling terrain by area of 477.5 km2 and 383.5 km2 respectively (Table 6; Fig. 10).

Table 6 Slope classes name and their corresponding area coverage
Full size table
Fig. 10
figure 10

Slope and slope range of the study area

Full size image

3.4 Land use land cover change over slope range in 1989

From the result of LULC change, agricultural land was increased on highland (> 30%) slope than gentle slope or low land (0–2%) in 1989 by an area of 50.3 km2 and 13.8 km2, respectively. This is the main reason for the decline in vegetation cover and increase of soil erosion rate (Table 7).

Table 7 LULC types with slope range in 1989
Full size table

3.5 Land use land cover change over slope range in 2002

From the result of LULC change, agricultural land was increased on high land (> 30%) slope than gentle slope or low land (0–2%) in 2002 by an area of 116.5 km2 and 14.3 km2 respectively. This is the main reason for decline of vegetation cover and increment of soil erosion rate (Table 8).

Table 8 LULC types with slope range in 2002
Full size table

3.6 Land use land cover change over slope range in 2020

From the result of LULC change, agricultural land was increased on highland (> 30%) slope than gentle slope or lowland (0–2%) in 2020 by an area of 187.2 km2 and 18.9 km2 respectively. This is the main reason for decline of vegetation cover and increment of soil erosion rate (Table 9).

Table 9 LULC types with slope range in 2020
Full size table

3.7 Estimation of annual soil loss of Upper Anger watershed

The results revealed that the mean annual soil loss in the Upper Anger watershed increased from 44.8 ton/ha/year in 1989 to 66.4 and 87.9 ton/ha/year in 2002 and 2020, respectively (Fig. 11). The annual soil loss of the Upper Anger watershed is greater than the indicated tolerable range by Hurni (1985) for different agro-ecological zones of Ethiopia, i.e., 2 ton/ha/year to 16 ton/ha/year. The rate of soil erosion increases as the slope length and gradient of the catchment increase. The highest key parameters that affect soil erosion are LULC change and ridged topography. It is necessary to implement soil and water conservation strategies in areas where soil erosion is a risk. As the result, decline of vegetation cover for agricultural expansion to steep slope was the main causes for substantial increasing of soil loss rate from time to time in the study area. Moisa et al. (2021) confirmed that, LULC change has an impact on soil erosion.

Fig. 11
figure 11

Annual soil loss map of the study area

Full size image

Results revealed that about 68% of the study area were classified under sever and very sever soil loss severity range in 2020 while it was about 44% in the year 1989 (Table 10). This clearly indicates that there is high LULC change in the study area which was driven by agricultural expansions. Moisa et al. (2021) confirmed that changes in LULC have an effect on soil erosion. Other study by Gemeda et al. (2021) reported that LULC change is one of the major driving forces for the increasing trends of temperature in southwestern parts of Ethiopia. The LULC change highly contributes to annual soil loss. The spatial distribution of soil loss severity varies across the study area and very severe soil loss was observed in the southeastern parts of the watershed (Fig. 12). The annual rainfall and R factor map showed that the southeastern parts of the watershed receive high amount of annual rainfall, while the northwestern receive low amount of annual rainfall.

Table 10 Severity classes and area of soil loss in the study area
Full size table
Fig. 12
figure 12

Soil loss severity map of the study area

Full size image

3.8 Soil loss rate per slope from 1989 to 2020

The mean of soil loss in the study area increased from low land (0–2%) slope to high land (> 30%) slope from 1989 to 2020 by 2.25 ton/ha/year to 207.6 ton/ha/year, respectively (Table 11). Expansion of agricultural to high land area by decreasing vegetation cover is the main factor for increasing soil erosion over the study period. Beyene (2019) confirmed that soil erosion increased with slope.

Table 11 Mean soil loss and slope range
Full size table

3.9 Land use land cover change and soil erosion rate

The relationship between LULC types and mean soil loss was calculated in ArcGIS using zonal statistics. As a result, soil erosion in agricultural land increased in the study area from 1989 to 2020. By reducing vegetation cover, agricultural land was expanded to elevated land. Soil loss was increased from 75.9 ton/ha/year in 1989 to 98.5 ton/ha/year in 1990, and 103.8 ton/ha/year in 2020 (Table 12). The findings of this study are consistent with Moisa et al (2021) and Negash et al. (2021) in the case study areas of Temeji and Chogo watersheds, respectively.

Table 12 LULC types with mean soil loss from 1989 to 2020
Full size table

3.10 Correlation analysis of mean soil loss with NDVI and slope

Soil erosion in the study area has a negative relationship with the NDVI with R2 = − 0.89 and a positive relationship with slope with R2 = 0.99 (Table 13). The results show that a significant decrease in vegetation cover for agricultural expansion to large land areas was the primary reason for an increase in mean annual soil loss from time to time in the study area.

Table 13 Correlation of soil loss with NDVI and slope
Full size table

3.11 Correlation analysis of annual soil loss with rainfall erosivity

In the study area, soil loss showed a positive correlation (Fig. 13) with the rainfall erosivity factor (R2 = 0.93). This indicates that as the amount of rainfall increases there is an increment in annual soil loss that significantly affects the environment in general and agricultural production in particular in Ethiopian highlands particularly during the crop growing season (June to September).

Fig. 13
figure 13

Correlation between soil erosion and rainfall erosivity

Full size image

3.12 Regression analysis of soil loss rate with NDVI and slope

Descriptive statistics were used to summarize the data, primarily mean annual soil loss, while inferential statistics in the form of a one-way analysis of variance (ANOVA) were used to investigate the effect of vegetation cover and slope. The independent variables in the ANOVA were NDVI and slope, while the dependent variable was annual soil loss (Fig. 14). According to the R2 values for each year, NDVI and slope as a predictor of annual soil loss. The primary causes for the rising of annual soil loss were the declining of NDVI and an increasing of slope. Other investigations have yielded similar results (Zhou et al. 2014; Akinyemi et al. 2019). The coefficient of determination (R2) between annual soil loss, NDVI, and slope was determined for the years 1989 to 2020 (R2 = 0.99). In general, the p values in (Table 14) show that an increase in agricultural output. The annual soil loss of the Upper Anger watershed increased by 99.8% due to a decline in vegetation cover from elevated areas for agricultural expansion based on adjusted R2, with the remaining percent possibly increasing due to other factors.

Fig. 14
figure 14

Slope, NDVI and soil loss regression analysis of the study area

Full size image
Table 14 Regression analysis of the study: soil loss, NDVI and slope
Full size table

4 Conclusion

Water erosion is the most common and serious problem in Ethiopian highland particularly in western parts of the country due to decline in vegetation cover from steep slope. In the present study, the causes and effects of slope and decline of vegetation cover on soil loss are calculated using multiple regression analysis. The loss of vegetation cover from elevated areas has a significant impact on soil erosion. Soil erosion has a significant impact on food security by reducing agricultural yields due to the loss of fertile soil. In this study, an integration of RUSLE model and a geospatial technique are used to estimate the impact of deforestation on soil erosion in highland areas of Ethiopia. In this study, different parameters, such as rainfall, soil types, slope length and steepness, cover management, and conservation practices, are considered. In the study area, the NDVI and slope are computed to determine the cause-and-effect relationships with soil loss. Our results conclude that a decline in vegetation cover is a key factor on soil erosion in the study area. Due to the declining trend of vegetation cover and continuous LULC change, the mean annual soil loss in the study area is substantially increased from 44.8 ton/ha/year in the year 1989 to 87.9 ton/ha/year in the year 2020. Our study concludes that the increasing trend of annual soil loss was associated with the declining of vegetation cover over the study period. Moreover, the amount of annual soil loss is influenced by slope steepness and LULC conversion. Our findings clearly identify erosion hotspot areas that require further research and policy intervention. Hence, conservation measures, such as re-afforestation, area closure, agroforestry practices, and community mobilization, to strengthen the ongoing participatory watershed management should be encouraged by all concerned stakeholders.