Responses of soil and water-related ecosystem services to landscape dynamics in the eastern Afromontane biodiversity Hotspot

Abstract

The investigation of soil and water-related ecosystem services (ES) and recognizing the spatiotemporal effects of land-use and land cover changes (LULC) are essential for the formulation of conservation strategies. The research employed the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) and the Revised Universal Soil Loss Equation (RUSLE) models for the assessment of ES. The study was carried out in the Kaffa Forest Biosphere Reserve in Ethiopia, which is part of the eastern Afromontane biodiversity hotspots. The aim of this investigation was to examine and map the temporal and spatial fluctuations in sediment retention, soil erosion, and water yield resulting from LULC modifications between 1986 and 2019, and to provide forecasts for the next three decades. According to the RUSLE analysis, the landscapes exhibited estimated soil losses ranging from zero to 1.5 tons ha-1 yr-1 in 1986, 2009, and 2019, respectively. The mean annual sediment exports for the years 1986, 1999, and 2019 were estimated to be 12.6, 9.9-, and 28.7-tons ha⁻¹, respectively. The water yield of the site experienced a notable increase from 9.8 × 109 m³ in 1986 to 19.6 × 10 9 m³ in 1999, and subsequently rose to 39.3 × 109 m³ in 2019, which is considered to be a disadvantageous to the site. The study found a significant positive correlation between water yield and the expansion of settlement area (r = 0.99, P = 0.015) as well as agricultural land (r = 0.99, P = 0.05). It was also found that significant positive correlation found between vegetation dense area such as forest (r < 0.999, P < 0.001) and shrub & bamboo (r = 0.998, P = 0.036) with sediment retention service. The investigation discovered that there existed tradeoffs between the ES of sediment retention and water yield as the slope increased. The results may be attributed to the presence of dense vegetation cover on the elevated slope regions, rendering them unsuitable for agricultural activities, and the concurrent expansion of arable lands in the lower slope areas, which are flat terrains more conducive to cultivation. The transition from land with more vegetation density to land with lower or no vegetation coverage resulted in an increase in soil loss and water yield, while simultaneously decreasing the sediment retention service. Therefore, the findings can be used as a document to guide decision-makers to design soil-water conservation technologies to enhance the ecological integrity of the biosphere reserve.

Graphical abstract

Highlights

• •We used field data to investigate the effects of land-use changes on carbon balance.
• •The carbon stock decreased from 478.3 megatons in 1986 to 457.3 megatons in 2019.
• •Our analysis explains the relationship between land-use dynamics and climate change.
• •Agricultural land increased, and that drove carbon storage loss.
• •Anthropogenic activities have caused a decline in ecological lands in the biosphere reserve.

1. Introduction

Soil resources play a crucial role in providing essential material goods and ecosystem benefits to humans and other organisms [1]. According to Briak, Moussadek [2], soils offer diverse ecosystem services to human beings. These are nutritional requirements [3] and a potential site for a carbon sink [4]. Soil has the capacity to mitigate the likelihood of flooding by serving as a reservoir for precipitation and subsequently discharging it at a gradual pace. The terrestrial substrates are inhabited by a diverse array of microorganisms such as bacteria, fungi, and invertebrates. The intricate interaction among soil organisms and ecological functions plays a pivotal role in preserving soil health and promoting plant root growth, soil structure, and soil water retention capacity [5,6].

Despite the crucial role of soil in ecosystem functions, the majority of studies on ecosystem services (ES) during the 1990s were centered on the conceptualization and establishment of frameworks [7]. It is noteworthy that until recently, literature on ES did not extensively cover the subject of soil [8]. The adoption of an ES framework in soil science was a direct result of the publication of the Millennium Ecosystem Assessment (MA) report in 2005, as noted by Woznicki, Cada [3] and soils are under the supporting service [9]. The MA report of 2005 indicates a growing trend in publications that consider soil as a form of natural capital and an ES [7,10].

According Robinson, Hockley [9], soil is a natural capital. It is responsible for providing a multitude of ES that are crucial for the preservation of humanity, as highlighted by Pereira, Bogunovic [11]. According to Pimentel, Burgess [12], disservices occur when the rate of soil loss surpasses the rate of soil formation. The loss of soil from terrestrial areas poses a significant challenge to the planet, resulting in adverse effects on agricultural output, ecological integration, and increased sedimentation [13]. In recent times, researchers have exhibited a strong interest in exploring ES related to soil-water resources in various geographical regions, ranging from small sub-watersheds to vast landscape areas. This has been achieved through the utilization of the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model and the Revised Universal Soil Loss Equation (RUSLE), which have been integrated with the Geographical Information System (GIS).

Soil erosion is a natural geological phenomena whereby the soil is displaced from its original formation location by agents such as water or wind formation [14,15]. According to a recent study conducted by Fenta, Tsunekawa [16], water erosion has resulted in an estimated average annual soil loss of 16.5 tons ha⁻¹ yr⁻¹ in Ethiopia. The acceleration of soil erosion rates due to human activities has resulted in a detrimental process, surpassing the rate of soil formation [14,17,18]. The freshwater flow is a crucial ES within a watershed region. Water plays a crucial role in various sectors including domestic, agricultural, industrial, hydropower, and ecological preservation, thereby promoting sustainable development. Recently, there has been a focus on the evaluation and estimation of water yield as a crucial ecosystem service in development activities [[19], [20], [21]]. Climatic variables, including precipitation and evapotranspiration, as well as anthropogenic alterations to land use and land cover change (LULC), have been found to impact both the quantity and quality of water yield quantity [20,22,23]. LULC changes are significant factors that contribute to the fluctuation and depletion of ES and have an impact on their prospective quantity and distribution [[24], [25], [26], [27]]. The expansion of agricultural practices in tropical regions has resulted in soil erosion, which consequently causes a decrease in soil fertility, compromised soil structure, diminished water quality, and various other environmental risks [28]. Furthermore, the high and rapid substitution of forested regions with infrastructure development aggravates the pace of soil erosion [12,29]. Therefore, the preservation of soil-water resources plays a crucial role in safeguarding the ecosystem and its ability to maintain ecosystem functions [30].

According to the MA [31] report of 2005, sediment retention is identified as a regulatory ecosystem service related to soil and water resources. This pertains to the regulation of sediment volume that reaches stream networks, as discussed by Bogdan, Pătru-Stupariu [32]. Consequently, the vegetation offers advantages such as the conservation of water and soil integrity, as well as the mitigation of expenses associated with the removal of sediments from the dams reservoirs’ [[32], [33], [34]]. The sediment retention service provided by the natural landscape is of significant interest to water resource managers [32]. Therefore, evaluating the sediment retention service offered by the Kaffa biosphere reserve, henceforth referred to as Kaffa BR, is beneficial in safeguarding the water and soil resources of the area.

According to Brauman, Daily [35], the requirements of ES escalate in proportion to the expansion of the human population. The preservation of natural resources is of utmost importance in order to prolong the sustained advantages of soil-water ecosystem services. While it may be difficult to isolate ES from human beneficiaries [36], ES has emerged as a conceptual framework that highlights the interdependence between ecosystems and human well-being humans [37]. Comprehending the interconnections among diverse ES is paramount importance in formulating informed decisions and policies concerning matters pertaining to environmental, economic, social, and land use facets [38]. The notion of ES presents an opportunity to establish a connection between the welfare of human beings and the natural resources they rely on, thereby facilitating the development of effective conservation policy. The effectiveness of this approach could be enhanced by employing a comprehensive methodology that examines the interplay between ES and the nature and extent of their relationship with human beings [39,40].

The Kaffa BR is located in the “Eastern Afromontane Biodiversity Hotspot” and is known for its rich biodiversity. However, there is a lack of adequate information regarding the forest ES in the Kaffa BR [41]. The current research designed to evaluate the effects of LULC alterations on soil water-related ES in the Kaffa BR. The InVEST model is a commonly utilized tool for measuring and evaluating ES and providing assistance in the decision-making process [24,42]. Hence, it is imperative to carry out case studies utilizing the InVEST model at the biosphere level in nations such as Ethiopia, where there exists a paucity of data, in order to address pressing requirements and acquire fresh insights into the ES provided by the forest reserve. Furthermore, the discovery may provide insights on how to formulate effective management approaches for the preservation of soil and water resources. Gaining a comprehensive understanding of the dynamics of ES change across spatial and temporal scales among LULC classes can prove advantageous. Consequently, the current investigation was formulated with the aim of examining and mapping the spatiotemporal fluctuations of ES pertaining to soil-water interactions between the years 1986 and 2019, as well as evaluating the effects of LULC alterations on the aforementioned ES variations. The study had a specific focus on three objectives: (i) evaluating the water provisioning services provided by the biosphere reserve, (ii) determining the potential of the sediment retention service and the annual rate of soil erosion, and (iii) analyzing the impact of land use and land cover change trajectories on the trade-offs and synergies of soil-water-related ecosystem services.

2. Materials and methods

2.1. Description of the study site

The Kaffa BR is located in the southern part of the country (Fig. 1A and B). The geographical coordinates of the biosphere reserve, situated between 6°54′ to 8°10′ N and 35°59′ to 37°18’ E, as depicted in Fig. 1C, cover an area of 744,919.18 ha, was declared by UNESCO in June 2010 [43]. The biosphere reserve has the main rainy season from June to September, while the small rainy season occurs from February to April. The mean annual rainfall of the area ranges from 1500 to 2000 mm. The mean yearly temperatures fall between 15 °C and 24 °C [44].

Fig. 1.

(A) The map of Ethiopia; (B) The map of the former SNNPR's administrative regional state; and (c) The map showing the geographical location of the study area, sample plots, major rivers, and roads existed in the Kaffa Biosphere Reserve in southwestern Ethiopia.

Over half of the remaining mountain forests in Ethiopia are comprised within the biosphere reserve. The biosphere reserve is primarily characterized as a region of high biodiversity, encompassing a variety of natural habitats including Afromontane evergreen forest, bamboo, grassland, wetlands, plantations, and coffee forests [[45], [46], [47]]. A total of 244 plant species were documented, with 106 of them being classified as woody plants and distributed across 74 genera and 38 families. Coffea arabica L., a widely distributed plant species in the biosphere reserve, has its origins in Ethiopia. Furthermore, the biosphere reserve harbors a diverse range of fauna, comprising 300 avian species, 300 mammalian species, and numerous other species of reptiles, amphibians, and fish [43,[48], [49], [50], [51]]. The biosphere reserve serves as a crucial source of freshwater for the surrounding area. However, the area covered by native habitats is declining due to human activities [52].

2.2. Ecosystem modeling and mapping methods

The study used the InVEST model in conjunction with other essential analytical instruments to estimate the ES of Kaffa BR ecosystems [53]. The utilized model enables the user to assess various forest ecosystem services (FES) through the application of LULC change scenarios. The InVEST model comprises multiple sub-models that facilitate the assessment and mapping of ecosystem [54]. Consequently, the aforementioned model facilitates the process of measuring, mapping, and evaluating ecosystem services generated within the scope of a given landscape or watershed. By utilizing the InVEST and RUSLE models, the sediment yield/retention, soil loss, soil erosion, and water yields ES were estimated and mapped in the Kaffa BR. The findings of this study offer a more comprehensive evaluation of the ES of the Kaffa BR than any prior research. The investigation utilized the InVEST model (version 3.8.9) to assess the water yield (water provision service) and sediment export (soil conservation service) of the study site.

LULC data sources

The research employed a supervised approach to LULC classification, utilizing the maximum likelihood classifier technique. The conversion process involved transforming every image into a standardized Universal Transverse Mercator projection with a 37 N designation. The images were geographically referenced to the datum of Ethiopia utilizing the World Geodetic System 1984 (WGS-1984). This practice can ensure in maintaining uniformity between the datasets while conducting the analysis. Appendix 1 provides an overview of fundamental attributes pertaining to the utilized image, including the date of accusation, resolution, as well as paths and rows of the Landsat images.

The scope of this investigation encompassed a time period spanning 33 years, specifically from 1986 to 2019, and included projection analysis of LULC for the years 2034 and 2049. The process of image classification was executed utilizing the software ArcGIS, while the projection was performed within the IDRISI Selva. The various landscapes present within the biosphere reserve were classified into eight LULC types. The raster data utilized in the InVEST model for evaluation of ES pertaining to soil-water interactions had a spatial resolution of 100 m. The authors Mengist, Soromessa [55] provided a comprehensive account of the methodology employed for the preparation of the LULC map, as well as the corresponding values for overall accuracy and Kappa coefficient. The present study obtained raster-formatted input data on LULC for historical periods (1986, 1999, and 2019) and future projections (2034 and 2049) from Mengist, Soromessa [55], which is our previous work from the same research project site and can be seen for further detail.

2.3. Soil erosion and sediment retention

2.3.1. Parameterization of RUSLE model

Wischmeier and Smith [56] introduced the parameter of the Universal Soil Loss Equation (USLE) and RUSLE by Renard, Foster [57], as shown in equation (1). The parameters are simple and commonly used as the approach is needed limited input data [58]. The model is applicable for estimating the average annual soil loss in forested and agricultural areas. The parameter in question is widely utilized in Ethiopia for the purpose of estimating soil losses, as evidenced by several studies. It is regarded as an easy, simple, and clear method [[59], [60], [61], [62], [63]].

The RUSLE model requires five inputs: R, K, LS, C, & P

$$A=R*K*LS*C*P$$ (Eq.1)

where: A is the average annual soil loss for a selected study area, which is represented in (ton/h ⁻¹ y ⁻¹);

• R factor is the rainfall-runoff erosivity (MJ mm ha⁻¹ h⁻¹ y⁻¹),

• K factor represents the soil erodability (a ton⁻¹ ha⁻¹h⁻¹ha⁻¹mj⁻¹mm⁻¹),

• LS factor is the slope length and steepness of the area,

• C factor is the cover management factor it is expressed in a range of 0–1 and,

• P factor is the management practice that is expressed in the value of 0 and 1.

The data sources were the LULC maps of the site, rainfall data of the region in raster format from the Ethiopian national meteorological agency.

2.3.1.1. Rainfall-runoff erosivity (R)

The R-factor is a parameter that takes into consideration the erosive power of rainfall [56]. The erosive potential of rainfall is linked to its distribution, amount, and intensity [64]. However, soil erosion is primarily determined by the intensity of rainfall, as highlighted by Blanco-Canqui and Lal [65]. The R-factor of the model is obtained through the analysis of precipitation intensity data, however, there is a dearth of information regarding rainfall intensity in Ethiopia at large and within the specific study area. Thus, the R factor was computed utilizing the alternative empirical equation (2) that was formulated specifically for the Ethiopian highlands [66]. Consequently, the R-factor was calculated based on the mean annual precipitation of the location. The grid-based rainfall data spanning from 1986 to 2016 were obtained from the National Meteorological Services Agency (NMSA) of Ethiopia. The dataset was presented in a tabular format spanning the temporal range from 1986 to 2016 [67]. The spatial resolution of the raster data sets is 4 × 4 km, enabling the inclusion of multiple raster points in the analysis. To ensure consistency with the other inputs utilized in the model, the R-factor underwent resampling to a resolution of 100 × 100 m within ArcMap.

$$R=-8.12+0.56*P$$ (Eq.2)

where P is the mean annual rainfall (mm) and R is the rainfall erosivity factor (MJ mm ha^_1 h^_1 yr^_1).

2.3.1.2. Soil erodibility factor (K)

The soil erodibility factor refers to the capacity of soil to withstand erosion, as measured by the quantity of soil lost per unit of rainfall-runoff erosivity [68]. This refers to the soil's responses to the impacts of a raindrop. The erodibility of the soil is related to its physicochemical properties, including soil texture, infiltration capacity, shear strength, soil organic content, and chemical compositions [69]. However, the absence of a comprehensive soil map for the site necessitated the production of the soil map from the digital river basin maps of Omo-Ghibe and Baro-Akobo [67].

The K-factor was computed by utilizing the soil texture and organic matter content of the soil within the 0–30 cm depth range. In 2021, the organic carbon content of the sampled soil was analyzed at the Tepi soil laboratory located in Ethiopia. The values range from 1.0 (most eroded) to 0.01 (almost non-erosive) [70]. The K-factor was calculated utilizing equation (3) that was adapted from Williams [71] and performed in Arc GIS 10.7.1 using a raster calculator tool.

KRUSLE = fcsand × fcl−si × forgC × fhisand … … … … … … … … … … …..… … … … … … ….(Eq.3)

Where,

• fcsand refers to a factor, which has provided soil erodibility of low and high value for soils of high coarse-sand and little sand contents, respectively;

• fcl-si is providing soil erodibility factors, which is low-value for soils rich in high clay to silt ratio;

• forgC is a factor that decreases soil erodibility with organic carbon in high content and,

• fhisand is a factor that reduces soil erodibility with extremely high sand contents. For the detailed procedure, we refer to Hategekimana, Allam [72], and Girma and Gebre [62].

2.3.1.3. Slope length and slope steepness (LS)

The dominant topographical factors that dictate the intensity of water erosion are the slope of the earth's surface and the length of the embankment [73]. The LS value was executed using equation (4), a commonly employed formula in numerous research works conducted in Ethiopia [59,60,74].

$$LS=\left({\left(flowacc*cellsize/22.13\right)}^{0.5}*{\left(\frac{\mathit{Sin}Ɵ*0.01745}{0.0896}\right)}^{1.3}\right)$$ (Eq. 4)

where LS is the combined Length and Slope-factors, 22.13 is the standard plot length in RUSLE; 0.5 is slope length exponent; θ is a slope in degrees (i.e., the slope of DEM × 0.01745).

2.3.1.4. Land use and land cover factor (C)

The LULC factor (C) represents the ratio of soil loss that occurs under a specific LULC in relation to the soil loss resulting from cultivation and slope under the same rainfall conditions (Morgan, 1994). The assessment of erosion hazard holds significant importance. Inaccuracies in LULC classification have the potential to result in substantial overestimations or underestimations of soil erosion. Consequently, it is imperative to possess a contemporary and precise land use and land cover (LULC) map for the purpose of scrutinizing the C-value. The C-value was generated utilizing the land use and land cover (LULC) maps from 1986, 1999, and 2019.

The C-factor values of the identified land-use types were assigned using the values from related studies (Appendix 2) [59,60,66,69,75,76]. The classified image format was changed into a vector format and subsequently, the editing menu of ArcGIS software was utilized to assign the corresponding C-value for each LULC class. The C-factor maps that were created were finally transformed from polygon to raster format, with a cell size of 100 m, in order to align with the other model inputs.

2.3.1.5. Land management practice (P) factor

The P-factor denotes the importance of implementing soil erosion control and management strategies to mitigate soil loss in highland regions caused by rainfall [56]. The P-value is subject to variation between zero and one, contingent upon the specific soil management activity being considered. According to Wischmeier and Smith [56] classification scheme, land-use types categorized into two distinct groups: agricultural and non-agricultural. In agricultural land, the P-value exhibits an ascending trend within the range of 0.1–0.33. The significance of agricultural or cultivated land varies based on the specific land management practices employed. Nevertheless, the non-agricultural land-use categories have a P-value of 1. As per the report by Hurni, Zeleke [77] report, it has been observed that the highland regions of Ethiopia lack substantial measures for soil and water conservation. In the past, when soil-water conservation technologies were not available, it was suggested in the literature to create a P-factor map using the land-use and slope maps of the area [56,59,78]. Nevertheless, there exist certain limitations associated with the utilization of this methodology. The methodology resulted in a decreased P-value for agricultural and cultivated land, while an increased P-value was given for other land uses, encompassing natural vegetation such as forests, grasslands, wetlands, and similar types. In contrast, the RUSLE model employs a multiplication of the five factors to assess the extent of soil loss, which leads the natural covers with high P-values experienced more soil erosion [59]. Hence, the present study has addressed the aforementioned limitation by formulating the P-factor value through the utilization of Land Use and Land Cover (LULC) maps, as presented in Appendix 2.

The P-factor for each LULC category utilized in this research was derived from previous studies conducted in Ethiopia [59,74,75,79]. The P-factor maps underwent a conversion process to achieve a map resolution of 100 × 100 m, in order to align with the remaining inputs of the soil erosion model. The estimation of the yearly soil erosion in each cell was derived from the multiplication of the five designated input parameters.

2.3.2. Sediment delivery ratio (SDR)

The RUSLE equation provides an estimation of soil loss; however, it does not assess the sediment yields that occur beyond the immediate location. The utilization of SDR was implemented for the computation of sediment yield [3]. The term SDR denotes the quantity of fine sediment generated and transported to the stream. The hydrological connectivity of the region is determined through the methodology initially proposed by Vigiak, Borselli [80]. Appendix 3 contains information regarding the format of the model inputs and the sources of data. The methodology utilized to calculate the connectivity index (IC) was derived from Borselli, Cassi [81] procedures, as shown in equation (5).

$$IC={\mathit{log}}_{10}\left(\frac{D_{up}}{D_{dn}}\right)$$ (Eq.5)

where $D_{up}$ is the upslope component and $D_{dn}$ is a downslope component [82].

Second the SDR value for pixel i is computed from the IC following equation (6) [80]:

$$SDRi=\frac{SDRmax}{1+\mathit{exp}\left(\frac{{IC}_0-{IC}_i}{K_b}\right)}$$ (Eq.6)

where SDR_Max refers to the maximum theoretical value, which is defined as the maximum fine sediment proportion (<1000 μm) that can be transported into the stream [80]. The relationship of SDR–IC is shaped by using the calibration parameters of IC₀ and K_b, which is defining the sigmoid function. For the detailed procedure, we recommend consulting the works of Hamel, Chaplin-Kramer [83], and Bogdan, Pătru-Stupariu [32].

2.3.3. Sediment yield

The sediment yield emanating from a designated pixel denoted as “i" is a function of the soil loss and the SDR factor, which is sed_export_i (Ei) (ton ha^{− 1}yr^{− 1}) as seen in equation (7) [84]:

$$Ei={RUSLE}_i*{SDR}_i$$ (Eq.7)

The total sediment yield, sed_export (ton ha^{− 1}yr^{− 1}), resulting from sheet wash erosion in a catchment can be determined by calculating the sum of sediment yield from all pixels using equation (8) [83].

$$E=\sum _{i=1}{E}_i$$ (Eq.8)

2.3.4. Sediment retention services

The sediment retention value was calculated by using: R_i*K_i*LS_i(1-C_i*P_i)*SDRi [84]. The size of sediment retained by the vegetative habitats was compared to an alternative scenario using the analysis of the actual LULC classification. The alternative assumes that the current natural vegetation cover was completely “removed” and converted into bare ground.

2.4. Water yield model

The model helps to facilitates quantification of the water-provisioning ES of the forest or watershed. The water yield refers to total runoff per unit area in a river basin over a certain period [26,85]. The InVEST water yield model facilitates the estimation of total water production by utilizing the water balance. The water supply can be defined as the disparity between the amount of precipitation and the amount of water lost through evapotranspiration from a given grid cell [21]. The resolution and sources of the inputs is located in Appendix 4. The comprehensive process for determining the water yield, which encompasses the parameter settings, was derived from the works of Gao, Jiang [86]; Lang, Song [21]; Yang, Liu [85]; and Sharp, Tallis [84].

2.4.1. Precipitation (P) and reference evapotranspiration (ET₀)

The determination of the water yield per annum Y(x) value for the specified landscape (x) in each pixel was determined using equation (9):

$$Yxj=\left(1-\frac{AET\left(xj\right)}{P\left(x\right)})\times P(x\right)$$ (Eq.9)

where AET (xj) refers to the actual annual evapotranspiration for pixel x and P(x) is the annual precipitation value for pixel x, and AET (xj)/P(x) is the approximation of the Budyko curve [84].

The Budyko curve was computed using equation (10), originally proposed by Budyko [87] for LULC types and subsequently modified by Fu [88] and Zhang, Hickel [89], which was employed in this study.

$$\text{This}\text{was:}\frac{AET\left(xj\right)}{P\left(x\right)}=\frac{1+\omega xRxj}{1+\omega xRxj+\frac{1}{Rxj}}$$ (Eq.10)

The term Rxj denotes the Budyko dryness index value associated with a particular land use j at a given pixel x and $\omega$ x is a parameter that characterizes the ratio of water storage accessible to plants to the expected precipitation over the year [21]. The evapotranspiration value was computed using the mean monthly and annual temperatures extracted from the gridded raster data of the study area. The value is computed using the Hamon equation (11) [90].

$$PEDHamon=13.97*d*D^2*W_t$$ (Eq.11)

where d is the total days in a month, D refers to the mean monthly daylight hours in each year, and Wt is a measure of the concentration of saturated water vapor, which can be computed using the following formula, equation (12):

$$W_t=\frac{4.95*e^{0.062*T}}{100}$$ (Eq.12)

where T refers to the mean monthly temperature of the area in degrees Celsius.

The annual potential evapotranspiration (PET) map was computed for the observed years by aggregating the monthly PET values of each grid cell [84].

2.4.2. Soil depth and plant available water content (PAWC)

In 2020, soil samples (0–30 cm) were collected from 155 plots at the study site to prepare the soil texture. The procedure for soil analysis was conducted at the Tepi soil laboratory located in Ethiopia [91], and subsequently, the results were interpolated using Arc Map.

The plant available water capacity (PAWC) denotes the disparity between the water content at field capacity and the water content at wilting point [84,85]. The PAWC was derived from the methodology of Zhou, Liu [92], which incorporates the soil's physicochemical properties of the soil. The aforementioned approach has been employed in several associated studies [22,23,85]. The formula is represented as follows, equation (13):

$$PAWC=54.509-0.132+sand-0.003+{\left(sand\right)}^2-0.055*silt-0.006*{silt}^2-0.738*clay+0.007*{clay}^2-2.688*OM+0.501*{OM}^2$$ (Eq.13)

2.4.3. The sub-watershed and biophysical table

The biosphere reserve was subdivided into 19 distinct watersheds through the application of a hydrological analysis tool within ArcMap 10.5. The biophysical value attributed to each identified LULC type can be found in Appendix 5. The term LULC-desc refers to the designated nomenclature of every LULC category. The LULC_veg variable denotes the AET equations that assign a value of 1 to LULC categories characterized by vegetation, with the exception of wetlands, and a value of 0 to all other LULC types. The term Kc denotes the evapotranspiration rate of the plant as per the FAO 56 guidelines [93].

The root depth is often the depth at which 95 % of the root biomass occurs. The remaining land cover types that were not vegetated were found to have a minimum value that was greater than zero [84]. The vegetated LULC of the numeric values for root depth was determined in accordance with the methodology outlined by Canadell, Jackson [94]. Grid-based maps depicting the depth of roots in millimeters were generated by utilizing the Africa Soil Information System (AfSIS) database [95]. The African soil profiles database was established through the collection of soil samples from around 28,000 distinct locations. The soil sample covered 40 African countries from 2008 to 2014 aiming to address the existing gap in soil information on the African continent [96].

2.4.4. The Z parameter

The Z parameter is a constant value that serves to characterize precipitation patterns, and its values typically fall within the range of 1–30. The magnitude of the value is greater in regions that experience frequent occurrences of precipitation. The Z value in the present investigation was determined to be 19.3 through the use of equation (14).

$$Z=\frac{\left(\omega -1.25\right)*P}{AWC}$$ (Eq.14)

where the value of P is Precipitation and AWC is the Available Water Capacity of the study area [84].

2.5. Model validation

The process of validating models is an essential component of soil-water ecosystem services research. In order to ensure the accuracy of model outputs, it is necessary to validate them against observed data. The use of data collected from plot or field-based sources is more precise in this regard [97,98]. The study findings relating to soil loss, sediment erosion, sediment retention, and water yield in the biosphere reserve were evaluated against the observed data recorded in 2019. The observed suspended sediment concentrations (g/ml) and streamflow (m³/s) were obtained from the Ministry of Water, Irrigation, and Electricity (MoWIE), of Ethiopia in 2019 [99]. The model performance was assessed using various metrics such as the root means square (RMSE), Nash-Sutcliff model efficiency (NSE), coefficient of determination (R2), and percent bias (PBIAS), as documented in previous studies [[99], [100], [101]]. The study involved a comparison between model estimates of runoff, sediment, and water yield derived from measured data obtained from MoWIE and simulated data for each event.

The InVEST SDR model output is measured in units of ton ha-¹ yr-¹, while the observed streamflow data is measured in m³/s and the suspended sediment concentration is measured in g/ml. The determination of sediment load and suspended sediment concentration was carried out through the utilization of equation (15) [102].

$$SC=b\times Q^c$$ (Eq.15)

where SC (is sediment loss in ton/day), Q (is streamflow rate in m³/s), and b & c are regression constant, which are determined from the sediment concentration and streamflow.

2.6. Statistical analysis

The study's objectives were taken into consideration when determining the criteria for evaluating the model's performance. The acceptable evaluation criteria were determined to be R2 > 0.50, NSE >0.30, RMSE >0.60, and PBIAS <10, as established by previous research conducted [101,103,104].

The correlation coefficients (Pearson, two-tailed) were utilized to express the associations between soil loss, sediment erosion, and sediment retention with land-use/cover types in various years, highlighting the trade-offs and synergies. The statistical analysis was conducted utilizing SPSS version 26 [105].

3. Results

3.1. Model validation

The simulated model output was validated with the observed data and the result were fitted. The findings indicate that the model's performance exhibited a satisfactory correlation between the measurements taken in the field and the model's outcomes regarding sediment loss in the sampled regions (R² = 0.996, p < 0.05). The results indicate that the mean values of PBIAS, RMSE, and NSE were 5.9 %, 0.69, and 0.96, respectively, as depicted in Fig. 2A. The comparison of water yield results between observed and simulated from 1986 to 2019 was presented in Fig. 2B. The statistical analysis has verified that the utilization of observed simulation did not exhibit any significant deviation, indicating the appropriateness of the water yield sub-model in InVEST for the evaluation of annual water yield amounts in response to land use and land cover alterations.

Fig. 2.

(a) Comparison of the observed and predicted values of the SDR model, and (b) the water yield model using the estimated and observed data. The light blue color refers to the predicted value, and the pink color refers to the values of the observed or field data.

3.2. LULC change and dynamics

The LULC types of the biosphere reserve were categorized into eight LULC types based on previous studies at the site. Native habitats such as forest, grassland, shrub & bamboo, and wetland were converted into human-dominated land-use types [55]. The transformation of natural ecosystems, including forest, grassland, shrub and bamboo, and wetland, into anthropogenic land-use categories is a prevalent phenomenon. Table 1 indicates a reduction in the extent of forestland, grassland, shrub, and wetlands by 19.1 %, 7.8 %, 3.8 %, and 8 %, respectively, between the years 1986 and 2019 (Table 1).

Table 1.

The areal size and the change matrix of LULC in the Kaffa BR, Ethiopia.

	LULC_2019
LULC-1986	LULC types	Agricultural land	Barren land	Forest	Grass land	Settlement	Shrub land	Water body	Wetland	Grand Total	Loss
	Agricultural land	124,523.7	750.6	13,839.5	414.9	7497.4	10,191.0	2531.3	13,093.1	172,841.6	48,317.9
	Barren land	804.7	400.7	11.4	11.7	604.8	3397.0	36.1	314.8	5581.2	5180.5
	Forest	117,236.0	70.5	216,329.5	169.1	2932.4	11,829.1	5746.5	20,640.1	374,953.1	158,623.7
	Grass land	7117.8	236.9	881.8	3126.6	476.7	2006.7	205.6	3334.7	17,386.9	14,260.3
	Settlement	4742.8	101.6	488.4	42.7	749.5	834.4	83.8	290.6	7333.8	6584.3
	Shrub land	24,500.2	4649.3	438.2	520.3	4320.4	18,556.1	1185.5	2040.6	56,210.6	37,654.5
	Water body	3768.0	230.5	19,118.4	5.0	387.9	2224.2	4380.5	696.4	30,810.9	26,430.4
	Wetland	47,495.7	85.5	6615.8	197.4	1523.5	3153.4	239.8	20,311.3	79,622.5	59,311.2
	Static (persist)									388,377.8
	Grand Total	145,257.9	5986.9	40,401.3	3742.8	1145.9	18,055.4	148,297.2	6695.5	369,582.9
	Gain	205,665.1	6124.8	41,393.7	1361.2	17,743.2	33,635.8	10,028.5	40,410.3
	Net gain	157,347.2	944.3	−117,230.0	−12,899.1	11,158.9	−4018.7	−16,401.9	−18,900.9
	NP	1.3	2.4	−0.5	−4.1	14.9	−0.2	−3.7	−0.9

Bolded diagonal numbers represent proportions of each LULC that were static (persisted) between 1986 and 2019.

The loss and gain values refer the proportion of the LULC that experienced gross loss and gain in each class, respectively.

All the figures in the table are in hectares in thousand.

*Refer the sum of diagonals of static LULC.

Net change = gain–loss.

The predominant land-use category exhibiting a notable increase rate was agricultural land, followed by areas used for human settlement. Between 1986 and 2019, there was a 22 % increase in agricultural land and a 3.9 % increase in settlement area. The research results indicate that the location has undergone a significant land use and land cover transformation, as evidenced by the data presented in Table 1.

3.3. RUSLE analysis of soil loss, and sediment export and yields

3.3.1. Trends of soil losses in the Kaffa BR

The study results demonstrated that the mean yearly soil loss varied from no loss in flat topography to 471.5, 1731.8-, and 2421.4-tons ha⁻¹ yr⁻¹ in areas with steepness inclines and lacking vegetation coverage in the years 1986, 1999, and 2019, respectively. The data presented in Appendix 7 Fig. 1 indicates that the average yearly soil erosion rates for the years 1986, 1999, and 2019 were 0.32, 0.47, and 0.84-ton ha⁻¹ yr⁻¹, respectively. According to estimates, the aggregate annual soil losses across the landscape of Kaffa BR were 2.4, 3.5, and 6.3 thousand tons per year in 1986, 1999, and 2019, respectively. The classification approach developed by Prasannakumar, Vijith [28] has categorized the rate of soil loss into three severity classes, namely low (0–1.5 tons ha⁻¹ yr⁻¹), moderate (1.5–5 tons ha⁻¹ yr⁻¹), and high (>5 tons ha⁻¹ yr⁻¹). The study took into account the average yearly soil loss and its classification into distinct categories of soil erosion based on the field's state. At present, the majority of the study sites' sections exhibit dense vegetation cover, which have contributed to a reduction in soil erosion.

In 1986 and 2019, the low erosion intensity category encompassed 96.2 % and 89.7 % of the area of the biosphere reserve, respectively. The remaining exhibited an annual rate of soil loss exceeding 1.5 tons ha⁻¹ yr⁻¹. Despite the relatively low degree of soil loss, the proportion of land exhibiting high levels of soil loss (greater than 5 tons ha⁻¹ yr⁻¹) was found to be 3.9 % in the year 2019, representing an increase from the preceding two years as outlined in Table 2. The significant soil losses observed can be attributed to the combination of various factors, including steep topography, high precipitation rates, and the expansion of anthropogenic land uses such as agriculture, urbanization, and transportation infrastructure.

Table 2.

The average yearly intensity of soil loss, measured in both percentage and hectare units.

Severity	Soil loss (ton ha−1 yr−1)	1986		1999		2019
		ha	%	ha	%	ha	%
Low	<1.5	56,5016	96.2	51,9762	94.9	46,8225	89.7
Moderate	1.5–5	14,205	2.4	17,469	3.2	33,610	6.4
High	>5	7967	1.4	10,580	1.9	20,240	3.9

3.3.2. Change in sediment export and yields

In 1986, the biosphere reserve exhibited a maximum sediment export rate of 24.4- and 7.3-tons ha⁻¹ yr⁻¹ from barren land and shrub & bamboo areas, respectively. The average sediment export in the given year was 5.2 tons, as indicated in Table 3. The aforementioned approximations do not take into account a hydrological feature that is classified as a non-erodible type of terrain. The dominant sources of sediment export were found to be areas of land characterized by a lack of vegetation, such as barren land, shrub & bamboo, and agricultural land. In contrast to barren land, the persistent anthropogenic activity on arable land may introduce variability in the assessment of sediment exports. The term “barren land” relates to soil surfaces that have been eroded and degraded, resulting in a lack of vegetation throughout the year where the highest possible amount of erodible soil in relation to sediment export from other land use types.

Table 3.

The outcomes of the InVEST model concerning the average and total sediment export for every land-use category and year.

	1986		1999		2019
LULC	MSE	TSE	MSE	TSE	MSE	TSE	TSE_change
Agriculture	3.4	592,342.4	5.2	905,037.0	0.5	7385.6	584,956.8
Barren land	24.4	136,619	1.2	8603.9	1.4	8980.9	127,638.1
Forest	1.8	652,418.1	0.5	166,547.2	1.8	454,176.3	198,241.8
Grassland	1.0	18,050.7	0.5	4330.1	0.3	1290.3	16,760.4
Settlement	2.5	19,682.6	1.2	16,334.5	3.1	62,101.6	−42,418.9
Shrub & bamboo	7.3	411,358.1	1.5	180,839.4	1	50,363.4	360,994.6
Wetland	1.4	109,888.9	0.7	42,753	0.6	36,543.4	73,345.4
Total	–	1,940,359.7	–	1,324,445.1	–	620,841.5	–
Mean	5.2	242,545	1.4	165,555.6	1.1	77,605.2	164,939.8

MSE refers to the mean sediment export (t/ha/year) and TSE stands for the total sediment export in tons.

The SDR model is performed for three temporal instances that correspond to the extant LULC maps of 1986, 1999, and 2019. The SDR was calculated over a period of 33 years, spanning from 1986 to 2019 and the recorded values for the years 1986, 1999, and 2019 were 12.6, 9.86, and 28.74 tons/ha, respectively. The findings indicate that there was a decline of 21.7 % in sediment exports between 1986 and 1999. However, there was an increase of 36.9 % in sediment exports from 1999 to 2019, as illustrated in Appendix 7 Fig. 2.

3.4. Soil loss and sediment export verses LULC and slopes

This study showed correlation between changes in LULC and an increase in the proportion of soil loss. The amount of soil erosion is positively correlated with the expansion of human-dominated landscapes. For instance, the mean sediment export in 1986 was 3.4 tons ha⁻¹ and increased to 5.2 tons ha⁻¹ in 1999, when the agricultural land was increased by 2000 ha (Table 3). On the contrary, there is a reduction in soil erosion as the vegetation cover increases. The rate of soil loss exhibited an inverse relationship with the expansion of natural habitats, including forest, grassland, shrub and bamboo, and wetland. The vegetation covers found in native habitats serve to mitigate the impact of rainwater and protect the soil from erosion caused by precipitation. During the period spanning from 1999 to 2019, there was a notable increase in the rate of land conversion and deforestation for the purpose of expanding agricultural land. The outcome of this phenomenon has led simultaneously a high level in both soil erosion and sediment transportation, as illustrated in Fig. 3. The spatial variability of soil loss rate was contingent upon several factors, including the slope's steepness and length, climatic conditions, soil physicochemical properties, the land use land cover, and the management factors. The agricultural land, forestland, and shrub & bamboo exhibit high mean values of soil loss owing to their larger areal sizes, whereas barren land experiences high soil loss values due to the high C-factor value, as depicted in Fig. 3. The findings indicate that there is a positive correlation between human interference and soil loss, while conversely, there is a negative correlation between vegetation cover and soil loss. The relationship between sediment export, vegetation coverage, and human interference, however is not linear.

Fig. 3.

The annual rate of soil loss and sediment export for each LULC type across various years.

The results of the correlation analysis revealed that barren land (r = 0.89, P = 0.027) and wetland (r = 0.89, P = 0.027) exhibited a statistically significant positive relationship with total soil loss and sediment retention, respectively. Among the eight LULC types, agricultural land (r = 0.89, P = 0.039) had a significantly positive correlation to the total soil erosion.

The variability of soil loss and sediment erosion in Kaffa BR is contingent upon the gradient of the slope. According to the data presented in Table 4, the majority of the biosphere reserve, specifically 93.6 %, exhibits a slope that is less than 5°. Conversely, a very small proportion of the landscape, less than 1 %, is located on slopes exceeding 30°. In contrast to regions with high and medium slope gradients, low slope regions exhibited a significant impact on the overall yearly sediment export and soil loss. The study findings indicate a linear correlation between an increase in slope and a reduction in soil loss and sediment erosion within the landscape due to the availability of dense vegetation cover in those upper slope areas. The upper slope areas, basically mountainous land, are less preferred by the local community for farming and thus they are covered with Afromontane Forest.

Table 4.

The mean values of soil loss and sediment export in relation to slope classes.

Slope in degree	Severity classesa	Area	%	SL (tons/ha)	SE (tons/ha)
0–5	Very slight	697182	93.6	111.2	16.97
5–15	Slight	26917	3.6	3.4	0.5
15–30	Moderate	14697	2	4.8	0.7
30–50	Severe	5563	0.8	3.1	0.5
50+	Very severe	590	0.1	0.4	0.1

^aThe classification of severity was determined utilizing the works of FAO [106], Haregeweyn, Tsunekawa [74], & Degife, Worku [107].

3.5. Soil erosion probability zones

The classification of the biosphere reserve region was based on the forest ecosystem category established by Prasannakumar, Vijith [28], which categorized the area as low (<1.5 tons ha⁻¹ yr⁻¹), moderate (1.5–5 tons ha⁻¹ yr⁻¹), or severe (>5 tons ha⁻¹ yr⁻¹). This information can be found in Appendix 7 Fig. 3 and Appendix 6. The present investigation involved the delineation of approximately 19 watersheds through the utilization of their respective drainage systems, which were systematically labeled from W1 to W19. The erosion hazard maps were reclassified into low, moderate, and severe categories, as depicted in Appendix 7 Fig. 3. Out of 19 watersheds within the site, nine watersheds were classified as having low erosion risk, four were classified as having moderate erosion risk, and six were classified as having severe erosion risk.

In 2019, watersheds that experienced significant soil loss and sediment exports, comprising 18.8 % or 139,829 ha, exhibited average values of 7.34- and 2.77-tons ha⁻¹ yr⁻¹, respectively. Certain watersheds within this cohort were located in topographically complex and uneven terrain, serving as the upper source of the rivers. The remaining areas exhibit significant anthropogenic impacts on land use, including the proliferation of cereal crop farming. This indicates a greater need for soil-water conservation measures in comparison to other watershed regions. The Gojeb River, being the main tributary of the Ghibe River, has exhibited a significant level of soil export that may lead to substantial silt accumulation in the Ghibe hydropower plants.

The remaining areas within the biosphere reserve, comprising 81.2 % of the total, exhibited annual soil loss and sediment export at moderate to low levels, with the former being the more prevalent. The mean soil erosion rates observed in the aforementioned watershed groups were 2.26- and 0.74-tons ha⁻¹ yr⁻¹, and the sediment export rates of 0.21- and 0.25-tons ha⁻¹ yr⁻¹, respectively. In contrast to regions situated at lower elevations, it was observed that landscapes located at higher elevations did not exhibit a substantial degree of soil erosion and sediment transportation as the vegetation coverage was dense. This finding corroborates the absence of a correlation.

3.6. Sediment retention by LULC types

Good vegetation coverage provides a valuable service in retaining sediment [108]. The InVEST model employs a theoretical scenario to approximate the quantity of soil erosion that has been prevented, which assumes that all lands have been cleared to the level of bare soil. The model calculates the sediment retention services by taking into account the variation in sediment exports from the biosphere reserve under the current LULC conditions, assuming that all lands are devoid of vegetation. The sediment retention services exhibited variability based on the LULC categories, like to the patterns observed in soil loss and sediment export. According to the data presented in Fig. 4B, the forest was responsible for 40.2 %, 37.6 %, and 26.6 % of the total sediment retention in the years 1986, 1999, and 2019, respectively. Similarly, the agricultural land contributed 9.8 %, 11.7 %, and 16.3 % to the total sediment retention during the same years. The agricultural land classification encompasses agroforestry techniques, such as coffee, root crop, and tea cultivation, which may contribute to enhanced sediment retention capability. The grassland, bare land, and settlement areas exhibited the lowest sediment retention service. With regards to the quantity of sediment retained per unit area (hectare), the forest exhibited the greatest retention rate (>125 tons ha⁻¹ yr⁻¹), followed by shrub and bamboo (72.4, 108.4-, and 26.1-tons ha⁻¹ yr⁻¹) and agricultural land (69.9, 75.6-, and 61.4-tons ha⁻¹ yr⁻¹) in the years 1986, 1999, and 2019, respectively (Fig. 4A).

Fig. 4.

(A) Mean sediment retention services per LULC in the studied years and (B) total sediment retention services per LULC in the studied years.

Fig. 5A–C illustrates the maps showing the spatial distribution of sediment retention service within the Kaffa BR. The cumulative sediment retention service of all LULC categories was found to be 0.12, 0.11, and 0.12 Pg yr⁻¹ in the years 1986, 1999, and 2019, respectively. Regions with dense vegetation covers, such as forests, shrubs, wetlands, and grasslands, were found to exhibit the highest sediment retention services owing to their significant water-retention capacity.

Fig. 5.

Maps showing the sediment retention services of the Kaffa biosphere reserve in the southwest Ethiopia (a) in the year 1986, (b) in the year 1999, and (c) in the year 2019.

These areas accounted for 45 %, 51 %, and 28 % of the total soil retention in 1986, 1999, and 2019, respectively (Fig. 5A–C). The study conducted a correlation analysis to investigate the relationship between land use types and changes in sediment retention service over time. The results indicated a significant positive correlation between forest (r < 0.999, P < 0.001) and shrub & bamboo (r = 0.998, P = 0.036) land use types and sediment retention service. Nevertheless, the remaining LULC exhibited a weak association with the sediment retention service for example, bare land had weak (r = 0.24, p < 0.001).

3.7. Spatial patterns of water yield in the biosphere reserve

The aggregate yearly water yield capacity of the examined region exhibited an upward trend, rising from 9.81 × 10⁹ m³ in 1986 to 19.6 × 10⁹ m³ in 1999, and subsequently escalating to 39.3 × 10⁹ m³ in 2019.

The variability in water yield across watersheds can be attributed to variations in precipitation levels, evapotranspiration rates, LULC characteristics, and land area, as illustrated in Fig. 6A–C. The analysis of the water yield's spatial distribution revealed that the regions with the highest quantities were predominantly located in the southern, southwestern, central, and southeastern parts of the biosphere reserve. In 1986 and 1999, the study reveals that the watershed 8 exhibited the lowest mean water yield per hectare values of 11.77 × 10² m³ and 11.42 × 10² m³, respectively, while the highest values of 18.7 × 10² m³ and 18.4 × 10² m³ were observed in watershed 14, respectively (Fig. 6 A & 6 B). In 2019, there was a notable increase in the overall water yield within the biosphere reserve, particularly in the southern and western regions. This increase was observed in the watersheds that are indicated by the colors of red and light brown (Fig. 6C). The cause of this increase can be attributed to the significant removal of natural vegetation in these areas, which was done to expand agricultural and settlement areas. According to Fig. 6A–C, the mean water yield contribution exceeded 15 × 102 m³ ha⁻¹.

Fig. 6.

The spatial distribution of water yield in the Kaffa BR (A) in the year 1986, (B) in the year 1999, and (C) in the year 2019. The maps display watershed labels through the utilization of numerical values ranging from one to forty-four.

3.8. Water yield variation among LULC types

The study's findings have validated the existence of discrepancies in the average yearly water yield value among different LULC categories. The agricultural sector exhibited the highest mean water yields, reaching a maximum of 652 m³. This was followed by forestlands with a mean water yield of 427 m³, and subsequently by shrubs & bamboo and wetland in 2019. Nonetheless, the average water yields obtained from bare land (6.6 m³), grassland (13.1 m³), water bodies (41.1 m³) and settlement (61.1 m³) were comparatively meager. During the period of 1986–1999, it was observed that solely the average water yields derived from shrub and bamboo exhibited a rising pattern. Likewise, an upward trend was observed in agriculture, grassland, and waterbody from 1999 to 2019.

The study employed the Pearson correlation coefficient to ascertain the effects of LULC changes on the fluctuations in water yield. The results indicate a noteworthy positive correlation between change in water yield and modifications in both settlement area (r = 0.99, P = 0.015) and agricultural land (r = 0.995, P = 0.05). Nevertheless, the alterations that transpired in the remaining LULC did not exhibit a noteworthy correlation with the fluctuation in water yield. Overall, the expansion of agricultural and settlement areas within the biosphere reserve has resulted in temporal variations in water yield.

3.9. Sediment retention rate and water yield across the slope

Table 5 indicates that there exists an inverse correlation between the escalation of slope and the provision of water and retention of sediment ecosystem services. The findings indicate that regions characterized by lower slopes exhibited the most substantial impact on the overall sediment retention and water yield in comparison to areas with moderate and elevated slopes. The rationale behind this phenomenon is that regions with reduced inclinations exhibit greater surface area and are utilized for agricultural purposes, thereby resulting in increased runoff and soil erosion at the site. The findings indicate a linear inverse correlation between the increase in slope and both sediment retention and water yield.

Table 5.

The mean value of sediment retention and water yield across slope classes.

Slope in degree	Severity classes*	Area (ha)	%	SR	WY
0–5	Very slight	697182	93.6	545	1287.6
5–15	Slight	26917	3.6	15.8	47.7
15–30	Moderate	14697	2	22.0	22.1
30–50	Severe	5563	0.8	18.2	7.01
50+	Very severe	590	0.1	3.2	0.8

4. Discussion

4.1. Impacts of landscape changes on soil loss and sediment export

The phenomenon of soil loss exhibits a significant correlation with LULC [109]. The Kaffa Biosphere Reserve exhibited an average annual soil loss of 2190.9 tons per year, which can be attributed to the expansion of agricultural land and the reduction of natural habitats such as forests, grasslands, wetlands, and shrub lands. The observed outcome was a 6.7 % increase in soil loss as compared to the soil loss recorded in 1986. The aforementioned statement aligns with the research discoveries of Aneseyee, Elias [61]; Yohannes, Soromessa [110], Chimdessa, Quraishi [111], Kidane, Bezie [112], and Hassen and Assen [113], which indicated that alterations in LULC led to an increase in soil loss and sediment erosion rate.

The study revealed that soil erosion was prevalent in the watersheds located in the eastern and northern regions, as well as a small portion in the southern area of the biosphere reserve, in terms of spatial distribution. The main factors contributing to the significant erosion of soil were the influences of precipitation (the R factor) and changes in LULC. According to the RUSLE model, the length and slope factor did not have a significant effect on the rate of soil erosion and sediment export in the Kaffa BR. This finding is in contrast to previous research conducted in Ethiopia, which has consistently identified slope as the primary contributor to soil erosion [60,61,107,114]. The possible reason for the higher soil loss and soil erosion from areas with lower slopes could be the availability of less vegetation cover and higher human interference through cultivation. In contrast, the higher slopes areas showed little soil loss and erosion, despite being subjected to higher amounts of precipitation. The reason for this phenomenon can be attributed to the presence of good vegetation cover that serves as a protective barrier against the erosive effects of rainwater on the soil.

The evaluation of soil erosion and sediment export helps the identification of relevant soil and water conservation methodologies. According to Hurni [115], the soil loss tolerance (T value) refers to the maximum soil loss that can take place from a given land area without causing any substantial degradation of the soil. It is an important guideline for implementing soil-water conservation measures [76,116,117]. The estimated rate of soil erosion, which was <1 ton ha⁻¹ yr⁻¹ is comparatively lower than reports from other regions of the country. The soil erosion rate in the biosphere reserve was found to be lower than the rates reported in previous studies conducted in previous studies conducted in the Wabe catchment (165 tons ha⁻¹ yr⁻¹) by Sahle, Saito [118]; in Lake Hawassa (37 tons ha⁻¹ yr⁻¹) by Degife, Worku [107], in Winike watershed (23.17ton ha⁻¹ yr⁻¹) by Aneseyee, Elias [61], in Chemoga watershed (93 ton ha⁻¹ yr⁻¹) by Bewket and Teferi [76], and in Gojeb watershed (216.13ton ha⁻¹ yr⁻¹) by Choto and Fetene [119].

In comparison to the range of soil formation rates observed in the land units of Ethiopia, which typically fall between 2 and 22 tons ha⁻¹yr⁻¹, the biosphere reserve exhibits a negligible mean soil loss amount [115]. The rationales for this phenomenon were attributed to the topographical features of the area, wherein approximately 97 % of the region exhibits slopes that are less than 15°. Additionally, the agricultural practices in the region were predominantly characterized by agroforestry methods. Furthermore, the highland regions of the biosphere reserve were found to possess an adequate amount of vegetation cover. The lower slopes exhibit relatively flatter terrain, yet experience greater soil erosion compared to their counterparts in the upper slope regions. The observed phenomena may be attributed to various factors, including the development of gullies [120], the proliferation of agricultural practices [16,109], and the accumulation of runoff in the lower slope regions leading to significant soil erosion [121]. The findings suggest that soils that possess ample forest and grass coverage are comparatively less susceptible to erosion in comparison to uncultivated and cultivated lands [122].

4.2. Impacts of land-use/cover change on sediment retention and water provision

The changes in land utilization within the examined area have instigated fluctuations in sediment retention and water yield ecosystem services. Numerous studies elsewhere in the world have investigated the effects of LULC changes on water supply [26,114,[123], [124], [125], [126]], as well as sediment erosion and/or retention services [61,119,127]. The expansion of agricultural land with a high proportion of agroforestry practices increased the sediment retention service as well as the water yield, which is a disservice on-the site, in the Kaffa BR. Nevertheless, as the size of cropland expands in the future, this may lead to a decline in water yield quality and an increase in soil loss. The change in LULC in the biosphere reserve are in line with similar research conducted in other regions, which have documented a decline in water quality as a result of the expansion of human-dominated land-use-like such as cultivation [26,128].

The sediment dynamics within the biosphere reserve were influenced by a combination of climatic factors, topography, soil properties, vegetation cover, and anthropogenic activities that caused an increase in agricultural land and settlement areas [129]. Vegetation plays a crucial role in the conservation of soil and water resources [130]. The presence of ample vegetation cover across the majority of the biosphere reserve has effectively dissipated the energy of rainfall-runoff, thereby preventing soil erosion and subsequent sediment transport [131,132]. The presence of vegetation cover on a slope surface can contribute to the development of a rough nature for erosion and improves the soil resistance capacity by enhancing the soil physicochemical properties [133]. The study region is a forest biosphere reserve in the country, where the vegetation foundation is good and is robust and offers optimal benefits in terms of sediment retention services and reduction of water yield that is exported to the main rivers.

4.3. Contribution of native habitat to sediment retention and water provision services

The diverse advantages that natural ecosystems offer to human societies have been a subject of interest for scholars in recent times, who have endeavored to assess the value of ES [134]. The Kaffa BR offers a multitude of advantages that span from regional to international levels. Nonetheless, the InVEST model has not been extensively utilized to investigate the ES of the area, as there is a dearth of comprehensive research on the study site. The investigation utilized the sediment delivery ratio and water yield sub-models of InVEST in order to map and analyze the ecosystem services. The InVEST water yield model was employed in various regions of Ethiopia to calculate the water provisioning ecosystem services. For instance, a study in the Berressa watershed showed that the water yield has augmented by 32.29 million m³ within 45 years [114], which was higher compared to our study result (9.88 × 10⁹ m³ ha⁻¹ from 1986 to 2019). The observed variation can be attributed to disparities in the study's target years, parameter values for precipitation and evapotranspiration, and seasonal factor (Z), which is a necessary input for water yields in the InVEST model, as well as differences in the rates of LULC change.

Compared to other types of natural habitats, forest areas provide a greater proportion of sediment retention services. This concept is corroborated by the research conducted by Woznicki, Cada [3], and Wu, Liu [30]. According to their report, land areas exhibiting favorable vegetation conditions were found to have a greater sediment retention ES compared to other categories of land use. Therefore, the preservation of soil-water resources can be effectively achieved through the implementation of vegetation conservation and restoration strategies [135]. Wetlands and forest areas have similar vegetation cover value; however, the wetland areas generate lower soil loss and sediment export due to flat topography. This suggests that wetlands possess the capacity to disrupt sediment connectivity between upstream and downstream land areas [3].

4.4. Management implications under the proposed model

The evaluation of soil loss, sediment retention, water yield, and comprehension of the main contributing factors are essential to the identification and execution of soil and water conservation measures. The findings of the study have the potential to provide valuable insights for policymakers and conservationists in their efforts to investigate the extent and severity of soil loss, soil erosion, and water yield. Consequently, it is imperative to map the spatiotemporal statuses of soil-water ES in order to provide policymakers at various levels, ranging from local to national, with the necessary information to facilitate efforts aimed at controlling soil erosion and water resource. Despite the relatively low erosion rate observed during the period under investigation, the results of the study indicate that human activities within the ecological mosaics of the biosphere reserve have contributed to an acceleration of soil erosion and sediment yield. If the anthropogenic rates persist in their upward trend, the degradation of land and ecosystem services will correspondingly escalate. Thus, the implementation of suitable land management strategies is necessary to significantly reduce soil erosion. The success of resource conservation programs is contingent upon the favorable reception of conservation measures by the local communities. Furthermore, it is imperative that conservation strategies prioritize the preservation of indigenous resource conservation technologies and mitigate any adverse impacts on the livelihoods of local communities. Resource degradation is intrinsically linked to poverty [107]. The utilization of ES mapping has the potential to facilitate the decision-making process. Therefore, the attainment of sustainable development is an essential requirement for the responsible implementation of ES mapping [136].

Throughout the designated period of analysis, it was observed that certain regions within the biosphere reserve experienced significant soil erosion. However, safeguarding larger areas that are susceptible to severe soil loss was deemed impractical due to financial and technical constraints [74,97]. Hence, it is advisable to concentrate on the minor sites that exhibit significant soil erosion, as they necessitate minimal resources for implementing effective soil and water conservation techniques.

4.5. Challenges and uncertainties

The study encountered certain constraints. The InVEST model required limited data amount for evaluation, and our data amount was sufficient for water yield analysis, as stated by Daneshi, Brouwer [54]. Nonetheless, the model did not account for the interrelationships between surface and groundwater. According to Su and Fu [137] research, it is postulated that the water generated within a given watershed region, after accounting for evaporation, ultimately reaches the outlet of said watershed region. Furthermore, the computation of the water yield in the biosphere reserve was based on precipitation data. The aforementioned approach fails to account for the effects of intense precipitation events and fluctuations in precipitation patterns across different seasons and years, as noted by years [26].

Furthermore, the investigation failed to account for the impact of forest age and the influence of tree species on water-soil-related ES [138]. The InVEST SDR model exclusively accounts for sheet erosion, which primarily arises from low-gradient terrain, in its assessment of sediment retention and soil erosion functions. Thus, the model fails to account for other forms of erosion [137]. The resampling of the 250 m soil depth into 100 m raster data [114,139] could potentially introduce additional uncertainty into this study.

An increase in slope has no linear relationship with sediment retention and water yield. Areas at lower elevations contributed to maximum sediment retention and water yield than the other slope categories. This was associated with good vegetation conditions in the higher slope areas, where the rainfall is maximum. On the other hand, crop cultivation is common in lower elevation areas that accelerated soil loss. The findings of our study suggest the importance of developing appropriate technological solutions for the purpose of soil-water conservation, which could be reforestation/afforestation, implementation of physical soil-water conservation structures in bare land, and assisting regeneration using area closure. The study can therefore deliver quantitative information for concerned stakeholders to design strategies for the conservation and management of watersheds with severe soil losses in the biosphere reserve. It can be said that the InVEST and RUSLE models can estimate the ES in the biosphere reserve area. The study on investigating the inter-link that existed between LULC types and soil and water-related ecosystem services could apply to other related biosphere reserves or watersheds where time-sequenced digital LULC data is available.

5. Conclusions

The primary aim of the present study was to explore the impacts of land-use land cover changes on annual water yield, soil loss/soil erosion, and sediment retention services. The evaluation and mapping of the spatiotemporal consequences of LULC changes on those ES are crucial for the formulation of appropriate soil and water conservation methodologies. Comprehending and assessing the effects of LULC changes on the hydrological and soil erosion aspects can aid stakeholders in devising and executing effective measures to safeguard soil-water resources. The implementation of appropriate technologies has the potential to facilitate an immediate and efficient restoration of the ecosystem. The present investigation employed a combined methodology that involved the utilization of the InVEST sediment delivery ratio and water yield model, along with correlation analysis, to assess the individual impacts of various LULC categories and slopes on alterations in sediment retention, soil loss, and hydrological factors.

In general, the findings of the investigation emphasized that the biosphere reserve exhibited a lower degree of soil erosion, which fell below the threshold of acceptable levels owing to the favorable vegetation coverage, particularly in the elevated regions. Moreover, the prevalence of agroforestry techniques in the area resulted in minimal soil erosion. The presence of native habitats with good vegetation coverage has resulted in high sediment retention services that reduce the volume of surface runoff.

This research work can provide quantitative data that can aid stakeholders in formulating strategies aimed at preserving and administering watersheds that experience significant soil erosion within the biosphere reserve. The utilization of the InVEST and RUSLE models has the potential to provide an estimation of the ES within the confines of the biosphere reserve region. The research pertaining to the examination of the interconnectedness between LULC categories and ecosystem services related to soil and water has the potential to be extrapolated to other comparable biosphere reserves or watersheds, provided that there is access to time-series digital LULC data.

The assessing soil loss, sediment retention, water yield, and comprehension of the main contributing factors are essential to the identification and execution of soil and water conservation measures. The findings of the study have the potential to provide valuable insights for policymakers and conservationists in their efforts to investigate the extent and severity of soil loss, soil erosion, and water yield. Consequently, it is imperative to map the spatiotemporal statuses of soil-water ES in order to provide policymakers at various levels, ranging from local to national, with the necessary information to facilitate efforts aimed at controlling soil erosion and water resource. Despite the relatively low erosion rate observed during the period under investigation, the results of the study indicate that human activities within the ecological mosaics of the biosphere reserve have contributed to an acceleration of soil erosion and sediment yield. If the anthropogenic rates persist in their upward trend, the degradation of land and ecosystem services will correspondingly escalate. Thus, the implementation of suitable land management strategies is necessary to significantly reduce soil erosion. This requires the need of implementing appropriate land management strategies that could remarkably lessen soil loss by maintaining vegetation cover at the reasonably high rate might be sensible as a measure against potential severe erosion following Jiang [140] findings. Conservation measures need a positive response from the local communities and should be part of the resource conservation program. In addition, conservation techniques should value the local resource conservation technologies and reduce livelihood challenges for local communities. Resource degradation is intrinsically linked to poverty [107]. Thus, ES mapping can support the decision-making process.

Throughout the designated period of analysis, it was observed that certain regions within the biosphere reserve experienced significant soil erosion. However, safeguarding larger areas that are susceptible to severe soil loss was deemed impractical due to financial and technical constraints [74,97]. Hence, it is advisable to concentrate on the minor sites that exhibit significant soil erosion, as they necessitate minimal resources for implementing effective soil and water conservation techniques.

Data availability statement

Data will be made available on request.

CRediT authorship contribution statement

Wondimagegn Mengist: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. Teshome Soromessa: Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Visualization, Writing – original draft, Writing – review & editing. Gudina Legse Feyisa: Conceptualization, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing.

Declaration of competing interest

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

Appendix A. Supplementary data

The following is the Supplementary data to this article: