Soil loss due to surface runoff is a natural phenomenon accelerated by anthropogenic activities. The study attempted to evaluate soil loss, sediment export, and deposition as influenced by changes in land use and land cover (LULC) in the Bontanga watershed. The InVEST-SDR model integrated with RUSLE was used in soil loss assessment. Results revealed that agricultural land produced 11,365.39 tons/year of soil loss in 1997, followed by 17,476.85 tons/year in 2002. In 2013, agricultural land experienced a soil loss of 5,391.98 tons/year, which finally increased to 91,274.53 tons/year in 2022. Agricultural land exported 56.16% of sediment, 13.39% of dense forest, and 13.30% of grassland. Dense forest deposited 41.54% of the sediment load, 30.49% of mixed shrub and grassland, and 10.85% of grassland. Over a long period, agricultural land is anticipated to contribute soil loss of 2,347,414.04 tons/year and sediment export of 388,497.56 tons/year. Sediment deposition amounting to 1,048,258.78 tons/year is anticipated to be deposited within the agricultural field. Both MAE and MAPE statistical measurements indicate a good model prediction performance for soil loss and sediment export. Understanding where sediments are produced and delivered will guide decision-makers, land use planners, and watershed managers in monitoring and planning the Bontanga watershed.

  • The evaluation spans from 1997 to 2022.

  • The study employed the InVEST-SDR model integrated with RUSLE.

  • Agricultural land produced extreme soil loss amounting to 91,274.53 tons/year in 2022.

  • 56.16% of the sediments downstream of the watershed were exported by agricultural land.

  • Over a long period, agricultural land is predicted to generate an annual soil loss of 2,347,414.04 tons/year.

Soil erosion through surface runoff is a natural process accelerated by anthropogenic activities such as deforestation, intensive agricultural land development, urbanization, overgrazing, and wildfire (Bocco 1991; Marondedze & Schütt 2020). In addition, AbdelRahman (2023) reported that deforestation, large-scale agricultural land expansion, excessive overgrazing, and urbanization are primary factors driving soil erosion and desertification in semiarid regions. Overgrazing contributes to soil degradation by depleting vegetation cover and increasing ground compaction, subsequently decreasing soil infiltration capacity and fostering surface runoff, thereby exacerbating soil erosion (Kosmas et al. 2015). Deforestation, whether for agricultural land expansion, logging, or urban development, results in the destruction of the protective vegetative cover of the land, subsequently causing an increase in soil erosion (Khodadadi et al. 2023). Several researchers are increasingly focusing on the complex relationship between anthropogenic activities, climate change, land use and land cover (LULC) changes, and the potential for soil loss (Ozsahin et al. 2018; Huang et al. 2020; Belay & Mengistu 2021). Water-induced soil erosion can result in substantial sediment deposition in water bodies (Teng et al. 2019). Yaekob et al. (2022) reported that the impacts of soil erosion extend beyond the depletion of fertile soil in the agricultural land, and also led to increased water pollution, sedimentation of lakes, streams, reservoirs, and rivers. In turn, this process results in the obstruction of waterways, leading to flooding and declines in fish and other aquatic species. Water erosion is impacted by many factors such as climatic variables, LULC changes, anthropogenic activities, topographic characteristics of the land such as slope length and steepness, and soil type and its texture (Vanmaercke et al. 2014; Aneseyee et al. 2020; Kanito et al. 2023). The acceleration of land use change, driven by the intensification of agricultural land expansion, rapid population growth, and the absence of effective land conservation strategies, stands as a prominent factor contributing to the increased rates of soil erosion (Abebe & Sewnet 2014). Human-induced alterations, such as contour farming, strip cropping, terracing, and subsurface drainage, have proven effective in controlling soil erosion (Majoro et al. 2020). These traditional farming practices are reported as significant factors suitable for mitigating soil erosion in erosion-prone areas (Sardari et al. 2019). Many studies have shown that soil erosion rates are particularly higher in agricultural lands compared to areas that are well conserved (Tadesse et al. 2017; Aneseyee et al. 2020). Furthermore, several studies have demonstrated that vegetation cover plays a crucial role in reducing surface runoff velocity, mitigating factors contributing to soil erosion, and preserving the natural hydrological response of the watershed (Wang et al. 2021). Considering the on-site and off-site effects of soil erosion, it is crucial for soil loss dynamics in the watershed to be examined for the sustainability of the watershed.

The high soil erosion in the Bontanga watershed has resulted in 10.80% reduction in the storage capacity of the Bontanga reservoir, primarily due to siltation (Adongo et al. 2020). The study utilized RUSLE and LULC change map of year 2016 to estimate annual soil loss and its spatial distribution across nine (9) reservoir catchments in northern Ghana. However, these studies do not quantify how the dynamics of LULC change lead to varying levels of soil loss, sediment export, and deposition into the watershed. Consequently, there is a need for a time-series LULC change study to be conducted that can accurately fill this gap.

Additionally, there is a significant gap in the literature concerning the integration of soil loss estimates, sediment export, sediment deposition, and their spatial-temporal distribution in the context of land use and land cover changes within the Bontanga watershed. This study intends to comprehend relationships between LULC change on (i) soil loss, (ii) sediment export, and (iii) sediment deposition within the Bontanga watershed. According to Aneseyee et al. (2020), soil loss estimation varies based on the method employed, whether through direct measurement or model estimation. Many soil loss models, e.g., the RUSLE equation, are adept at predicting sheet and rill soil loss, but they do not possess the capability to forecast sediment export and sediment deposition within the watershed (De Vente et al. 2013). Soil loss computed using the RUSLE model may not necessarily be exported or deposited within the watershed; therefore, the model's effectiveness in estimating sediment transport and sediment deposition is limited (Boardman 2006). In this study, the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) sediment delivery ratio (SDR) model was employed to overcome the limitations of the RUSLE model.

Many studies have employed the InVEST-SDR model to examine the impacts of anthropogenic activities, climate change, and land use change practices on sediment dynamics. Bendito et al. (2023) utilized the InVEST-SDR model to evaluate the impact of land use and soil conservation on soil erosion and sedimentation processes in a semiarid basin of the Brazilian Savanna. The results indicated that anthropogenic activities within the watershed led to an increase in soil loss and a decrease in soil retention. Similarly, Keshtkar et al. (2022) assessed the impact of LULC change on soil retention service (SRS) in agricultural-urbanized landscape of the Jajrood basin in Northern Tehran, using the InVEST-SDR model. The findings revealed that LULC changes had both negative and positive effects on soil retention service. Additionally, Guo et al. (2023) adopted the InVEST-SDR model to evaluate soil erosion and its driving factors in the Huaihe region, China. The model output revealed a gradual increase in soil loss and soil retention in the Huaihe region.

Moreover, the study conducted by Qiao et al. (2023), on assessment of current and future soil erosion on changing LULC patterns within the Yihe river basin in north China, revealed that soil erosion intensity exhibited higher rates in mountainous areas and lower rates in flat plains, while agricultural land was identified as the principal contributor of the largest amount of soil erosion within the watershed.

Considering the importance of watershed ecosystem service, we utilized the InVEST-SDR model integrated with the RUSLE model to analyze soil loss originating from the upper pixel, sediment export at the watershed outlet, and sediment deposition within the watershed across various land use and land cover classes over a span of 25 years. Results of the analysis will guide decision-makers, land use planners, and watershed managers to design improved strategies for reducing sediment loads through changes in land use and management practices.

Study area

Bontanga watershed is located between latitudes 9°25′00″ and 9°35′30″ N and longitudes 1°0′0″ W (Figure 1). The total area covered by the watershed is 165 km2. There are many communities residing upstream of the watershed, which include Kpalsogu, Zangbalung, Sakuba, Dalun, Tibung, and Wuba, with their main activities being subsistence agriculture and fishing.
Figure 1

Location map of the Bontanga watershed.

Figure 1

Location map of the Bontanga watershed.

Close modal

The study area has two distinctive rainfall seasons: a wet season, which starts from April or May through September or October, and a dry season, which extends from November through March (Alhassan et al. 2014). The total average annual rainfall for the area is about 1,100 mm. The study area has minimum and maximum temperature ranges from 15 to 42 °C and is at an altitude spanning from 110 to 250 meters amsl. At the tail end of the watershed, there is earthfill dam constructed to harvest 25,000,000 m3 of water for irrigation and aquaculture (Alhassan et al. 2014). According to Zakaria et al. (2014) and Ghana Irrigation Development Authority (2010), more than 525 small-scale farmers are benefiting from the irrigation scheme.

Data source

In order to understand the significance of geomorphometry of the watershed components, a 30 m spatial resolution digital elevation model (DEM) and LULC of 25 years corresponding to 1997, 2002, 2013, and 2022 generated from Landsat 5 (TM), Landsat 7 (ETM+), Landsat 8 (OLI), and Landsat 9 (OLI-2) with various bands acquired from the link (https://earthexplorer.usgs.gov/) were used. The rainfall data for 1997, 2002, 2013, and 2022 was collected from the Savanna Agricultural Research Institute (SARI), and rainfall raster maps were generated using the Kriging interpolation method. Soil data were acquired from the Ghana Irrigation Development Authority (GIDA) in Tamale. A study also used Geographic Information System (GIS) and Integrated Valuation of Ecosystem Services and Trade-Offs (InVEST-SDR) models developed by Natural Capital Project, in partnership with Stanford University, University of Minnesota, Chinese Academy of Sciences, The Nature Conservancy, World Wildlife Fund, Stockholm Resilience Centre and the Royal Swedish Academy of Sciences to quantify and assess spatial-temporal variations in sediment generation and delivery to the river stream. Both the InVEST and USLE models use DEM as the primary input raster layer for computation of the slope length and steepness factor (LS factor), which is one of the most important input parameters in the model and affects the output of estimated soil loss. Another important key input parameter in the model is the land use and cover type (Jakubínský et al. 2019). Table 1 presents the satellite imagery data source.

Table 1

Satellite imagery data source

YearSpace craft and sensor IDData sourceSensor IDPath/rowAcquisition dateCloud coverImage resolution (m)
1997 Landsat 5 USGS TM 194/53 10/12/1997 5% 30 
2002 Landsat 7 USGS ETM + 194/53 17/12/2002 5% 30 
2013 Landsat 8 (OLI) USGS TIRS 194/53 05/11/2013 5% 30 
2022 Landsat 9 (OLI-2) USGS TIRS-2 194/53 22/11/2022 5% 30 
YearSpace craft and sensor IDData sourceSensor IDPath/rowAcquisition dateCloud coverImage resolution (m)
1997 Landsat 5 USGS TM 194/53 10/12/1997 5% 30 
2002 Landsat 7 USGS ETM + 194/53 17/12/2002 5% 30 
2013 Landsat 8 (OLI) USGS TIRS 194/53 05/11/2013 5% 30 
2022 Landsat 9 (OLI-2) USGS TIRS-2 194/53 22/11/2022 5% 30 

TM, Thematic Mapper; ETM + , Enhanced Thematic Mapper Plus; OLI, Operational Land Imager; OLI-2, Operational Land Imager 2; TIRS, Thermal Infrared Sensor; TIRS-2, Thermal Infrared Sensor 2.

Research methods

In this study, the InVEST-SDR model was employed to quantitatively analyze soil loss, sediment export, and sediment deposition generation and conveyance to the river stream within the Bontanga watershed for the years 1997, 2002, 2013, and 2022. The InVEST-SDR model relies on five essential RUSLE factors shown in Table 2,. The technical flowchart for estimation of soil loss, sediment delivery and deposition using the InVEST-SDR model is shown in Figure 2.
Table 2

Summary of the principal model equations used to generate RUSLE factors used in estimation of soil loss, sediment export and sediment deposition in the InVEST-SDR models

RUSLE factorEstimated valuesEquation usedReferenceData sourceLocation
Rainfall erosivity factor (R) MJ mm–1ha–1h–1yr–1 1,156.50–3,466.08a  Bagherzadeh (2014); Asma Belasri (2016)  Own computation Appendix 1 Table S1 
Soil erodibility factor (K) t–1h–1MJ–1 mm–1 0.0474–0.0487a  Tu & Mitani (2011); Arnold et al. (2012)  Own computation Appendix 1 Figure S1 
Slope length and steepness factor (LS) 0.00–50.858a  Bagherzadeh (2014); Khosrokhani & Pradhan (2014); and Asma Belasri (2016)  Own computation Appendix 1 Figure S2 
Crop management factor (C0.0293–0.075a 
where
 
Almagro et al. (2019)  Own computation Appendix 1 Figures S7–S10 
conservation support practice factor (P1.00     Panagos et al. (2015); Abdulkareem et al. (2019)    
RUSLE factorEstimated valuesEquation usedReferenceData sourceLocation
Rainfall erosivity factor (R) MJ mm–1ha–1h–1yr–1 1,156.50–3,466.08a  Bagherzadeh (2014); Asma Belasri (2016)  Own computation Appendix 1 Table S1 
Soil erodibility factor (K) t–1h–1MJ–1 mm–1 0.0474–0.0487a  Tu & Mitani (2011); Arnold et al. (2012)  Own computation Appendix 1 Figure S1 
Slope length and steepness factor (LS) 0.00–50.858a  Bagherzadeh (2014); Khosrokhani & Pradhan (2014); and Asma Belasri (2016)  Own computation Appendix 1 Figure S2 
Crop management factor (C0.0293–0.075a 
where
 
Almagro et al. (2019)  Own computation Appendix 1 Figures S7–S10 
conservation support practice factor (P1.00     Panagos et al. (2015); Abdulkareem et al. (2019)    

Pk: Monthly rainfall (mm); P: Annual rainfall (mm); k: Number of months; K: Soil erodibility factor (t∙ha∙MJ−1 ∙ mm−1); M: Particles percentage (% of very fine sand + % of silt) * (100 - % clay content); OM: Organic matter content (% orgC * 1.724), where orgC: % organic carbon content; b: soil structure code; c: soil permeability code; FlowAccum: Flow accumulation which describes how cell drain water to the downslope cell in the output raster; LS: Combination of slope length in meters and slope steepness; Cell size: Size of the grid cell; ϕ: Land slope in degree; m: Constant dependent on the slope gradient of the land: (0.5 if the slope angle is greater than 5%, 0.4 on slopes of 3% to 5%, 0.3 on slopes of 1% to 3%, and 0.2 on slopes less than 1%.) NIR: Surface spectral reflectance in the near-infrared band; RED: Surface spectral reflectance in the red band; NDV: Normalized Difference Vegetation Index.

aThe values of the RUSLE factor are variable in each year; therefore, the actual value within the study period is as shown in Table 1 and Figure 1 to 6 of Appendix 1.

Figure 2

Approach used for quantification of soil loss, sediment delivery, and sediment retention.

Figure 2

Approach used for quantification of soil loss, sediment delivery, and sediment retention.

Close modal

Revised universal soil loss equation (RUSLE)

The revised universal soil loss equation (RUSLE) model has been globally used for the prediction of potential areas vulnerable to soil loss in the watershed (Shikangalah et al. 2017). The key factors of the RUSLE model are; the rainfall erosivity factor (R), slope length and steepness factor (LS), soil erodibility factor (K), the support practice factor (P), and Land cover management factor (C) (Zhou et al. 2008; Khosrokhani & Pradhan 2014; Karamage et al. 2017; Abdulkareem et al. 2019; Marondedze & Schütt 2020). These factors are in the form of raster files, and in certain situations, data is available in constant values. Renard et al. (1997) stipulate that the calculation of annual soil loss per pixel, measured in tons per hectare per year, is achieved through the application of the Revised Universal Soil Loss Equation, as depicted in Equation (1). The essential factors of the RUSLE model alongside the methodologies employed to acquire each of them are summarized in Table 2.
formula
(1)
where A refers to annual average of soil erosion rate factor (t ha−1 yr−1), R refers to rainfall erosivity factor (MJ mm−1ha−1 h−1 yr−1), K refers to soil erodibility factor (t−1 h−1MJ−1 mm−1), LS refers to dimensionless slope length and steepness factor, C refers to dimensionless crop management factor (ranging between 0 and 1), and P refers to dimensionless conservation support practice factor (ranging between 0 and 1).

Approaches employed for long-term analysis of the spatial variation of soil loss, sediment export, and sediment deposition

The method employed for quantifying soil loss, sediment export and deposition within the Bontanga watershed is the same as that outlined by the InVEST-SDR model. The model uses the RUSLE equation to compute soil loss, then the SDR and other mathematical equations to estimate the amount of sediment deposited and exported downstream of the watershed (Degife et al. 2021). Sediment export is the amount of sediments that can be used to compare simulated and observed sediment load at the outlet of the watershed (Sharp et al. 2016). To comprehend the long-term impact of natural land vegetation cover on soil loss, sediment export and deposition, crop management (C) and conservation support practice (P) was assigned factors equal to 1, assuming that without improved conservation practices, the entire catchment would ultimately be converted into bareland. Due to the absence of natural land cover, a considerable potential soil loss is realistic. The absence of C and P in the watershed will assist decision-makers and watershed management authorities in assessing the severity of soil loss in the long term since C and P factors that protect soil are absent. Particularly, the InVEST-SDR model solely estimates soil loss from overland processes (sheet and rill erosion), excluding gullies, riverbanks, or landslides (Borselli et al. 2008). Output from the model encompasses sediment load exported to the stream and sediments trapped and deposited on the overland as a result of the ecosystem services in soil loss control. Figure 2 illustrates the methodology employed in modeling soil loss, sediment export and deposition within the study area.

Spatial analysis and mapping of sediment load distribution using the InVEST model

To integrate essential datasets into the InVEST-SDR model, we standardized all input data by setting them to the same Universal Transverse Mercator (UTM) projection and spatial resolution. Utilizing a DEM and Landsat imagery with 30-m pixel sizes, we prepared input data in the model, including flow accumulation, watershed boundary, flow threshold, LULC map, slope length and steepness factor (LS), rainfall erosivity factor (R), soil erodibility factor (K), and crop management factor (C), using ArcGIS 10.3. Additionally, the preparation involved creating a biophysical table in an Excel spreadsheet, saved in .csv format, for estimating soil erosion, sediment export, and deposition. This table included land use codes, land use classes, crop management factors, and support practice factors corresponding to each LULC class. Subsequently, the InVEST-SDR model was executed to combine all parameters, generating a final table of results and maps encompassing soil loss, sediment export, sediment deposition, avoided export, and predicted soil loss.

Understanding the spatial distribution of these quantities within the watershed enables the watershed management authority to monitor net changes in sediment at a pixel level (increase or decrease) and implement corrective measures proactively, preventing serious land degradation (Sharp et al. 2016; Degife et al. 2021).

InVEST-SDR model

The InVEST-SDR model is mainly developed to quantify and map sheet and rill sediment generation and delivery to the river stream. The InVEST-SDR model employs RUSLE to describe the temporal and spatial distribution of soil loss, sediment deposition, and sediment delivered annually in each pixel of the watershed through a .tif map (Guo et al. 2023). InVEST is one of the few commonly used models that is still under development (Jakubínský et al. 2019). The model output includes the amount of sediment delivered to the stream per pixel per year and the amount of sediment eroded and retained by vegetation in the catchment. The digital elevation model (DEM.tiff) is the primary input parameter in the InVEST-SDR model. Based on RUSLE, the model computes the amount of annual soil loss in each pixel, followed by an estimation of the SDR as the proportion of soil loss that actually enters the stream. Once the sediment enters the stream, it is assumed that it will be transported to the catchment outlet; therefore, in this model, no further in-stream increase or decrease of sediment loads is modeled. This approach was proposed by Borselli et al. (2008).

InVEST-SDR input/output key parameters

LULC Classification
This study utilized Landsat 5 (TM), Landsat 7 (ETM+), Landsat 8 (OLI), and Landsat 9 (OLI-2) satellite imagery to produce the LULC change analysis. Confusion matrix was used to access spatial accuracy and resolution quality of the satellite imagery based on producer's accuracy, user's accuracy, and Kappa coefficient (K) (Mekuriaw 2019; Mfwango et al. 2022). LULC changes were classified into seven distinct categories, i.e., dense forest, agricultural land, grassland, built-up, mixed forest and shrubland, water bodies, and mixed shrub and grassland, each corresponding to the years 1997, 2002, 2013 and 2022. Kappa coefficient values range from 0 to 1. Kappa coefficient values below 0 signify a lack of agreement or complete randomness, while the range of 0–0.2 denotes slight agreement, 0.21–0.4 categorizes as poor, 0.41–0.6 characterizes moderate agreement, 0.61–0.8 indicates significant agreement, and 0.81–1.0 signifies almost perfect agreement (Mekuriaw 2019; Mfwango et al. 2022). The computation of the Kappa coefficient was performed using Equation (2), as suggested by Mutayoba et al. (2018); Mekuriaw (2019); and Mewded et al. (2021).
formula
(2)
The average area (km2) and the percentage (%) of the average overall area covered by each LULC change were estimated using the following formula.
formula
(3)
formula
(4)
formula
(5)
formula
(6)
where n is the total number of observed areas in each land cover class, i represents number of years, ai illustrates estimated area of each LULC change classes, Ai is calculated percentage area covered by each LULC change, Areayear i is area of land cover i at the first year, Areayear i+1 is area of land cover i at the second year, and is total land cover area for i = 1 to n years.
Sediment Delivery Ratio
The SDR was applied to the soil loss estimated from the RUSLE model to compute the net sediment load transported downstream of the watershed (Woznicki et al. 2020). The RUSLE predicts only sheet and rill erosion and not off-site erosion. Vigiak et al. (2012) outlined that the SDR ratio for a pixel i is computed from the connectivity index (IC) as shown in Equation (7). The connectivity index describes the hydrological connection between sources of sediment (landscape) and sinks (streams). Higher values of IC indicate that sediment load is more likely to enter into the sink (more connected). Lower values of IC (lower connectivity) correspond with areas with good land cover and gentler slopes.
formula
(7)
where SDRmax is the maximum SDR, set as an average value of 0.8, ICo and K are model calibration parameters. Studies conducted by Borselli et al. (2008) and Vigiak et al. (2012) found IC0 equal to 0.5 more suitable for different study areas, and a K value of 2.0 was used according to Vigiak et al. (2012).
Sediment export
The sediment export from each designated pixel i, measured in tons per pixel per year, was determined by multiplying the average annual soil loss, computed through the RUSLE model, with the corresponding SDR. Agricultural land, steep slope areas, bare land, and high rainfall areas are expected to have more soil erosion from the pixel that actually reaches a stream. When source sediment is from steep slope, low land cover soil, and is in close proximity to the streams, high connectivity index (IC) occurs, resulting in high SDR (Borselli et al. 2008; Woznicki et al. 2020). The estimation of sediment export within the study area was conducted using Equation (8).
formula
(8)
where Ei refers to the sediment export from a given pixel i (ton. ha−1. year−1), USLEi refers to the average annual soil loss (ton. ha−1. year−1), and SDRi refers to the Sediment delivery ratio.
Total amount of the sediment load exported from the catchment (ton. year−1) is given by;
formula
(9)
where E refers to the Value of sediment export used for calibration or validation of data, in combination with other sediment sources, if data are available.
Sediment deposition
In a well-conserved area, the presence of abundant vegetation cover serves to decelerate and absorb rainwater, consequently reducing the volume and speed of surface runoff. When runoff is reduced, there is less force to transport soil particles, leading to lower sediment export (Morrow & Smolen 1975). In order to understand the potential of natural land vegetation cover in soil loss control, the InVEST-SDR model was assigned with the existing actual parameters of the crop management factor (C) and conservation support practice factor (P). In this study, sediment deposition refers to the amount of overland sediment erosion exported from a pixel and trapped or retained by vegetation cover along the flowpath downslope without reaching the stream. Vegetation cover, especially trees and grasses, has extensive root systems that bind soil particles together. This root network helps stabilize the soil and prevents erosion caused by water or wind. As a result, the soil is less likely to be detached and transported as sediment. The amount of sediment load that must be deposited somewhere on the landscape along the flowpath to the stream was estimated using Equation (10).
formula
(10)
where E′ refers to the sediment deposition (ton. ha−1. year−1), USLEi refers to the average annual soil loss (ton. ha−1. year−1), and SDRi refers to the Sediment delivery ratio.
Predicting soil loss, sediment delivery, and sediment deposition over the long term within the watershed

The current soil loss, sediment delivery and deposition in the watershed were assessed by assigning the InVEST-SDR model with the crop management factor (C) and conservation support practice factor (P) of the existing natural land vegetation cover of the year 2022, while predictions for the future were made by extracting the crop management factor (C) and conservation support practice factor (P) from the model by assigning them with C-factor = 1 and P-factor = 1 (Rizeei et al. 2016; Abdulkareem et al. 2019).

Sensitivity analysis of present and future soil loss, sediment delivery, and deposition

One of the biggest challenges facing soil erosion modeling is the difficulty of validating the estimates produced, especially in areas where no reliable data exist for comparing estimates with actual soil losses (European Commission 2002; Abdulkareem et al. 2019). The Bontanga watershed is ungauged, and no records of soil loss downstream of the watershed are currently available. Therefore, in this study, due to the absence of soil loss data in the watershed, the LULC change relationship was used to validate the existing soil loss, sediment delivery and sediment deposition estimated under the good crop management factor (C) and conservation support practice factor (P) and the long-term predicted soil loss, sediment delivery, and sediment deposition was computed in the absence of the C and P factors representing cleared land. Within the study area, the spectrum of C factors spans from 0.0293 to 0.075, whereas the value of C = 1 designates land that has been cleared or developed for built-up purposes. The Pearson correlation method suggested by Akoglu (2018) was employed to analyze the linear relationship between variations in LULC change classes and the corresponding simulated and predictive values of soil loss, sediment export, and sediment deposition. The strength of the correlation coefficient (R) ranges from −1 to 1. The negative sign indicates a negative correlation between the variables, and the positive sign indicates a positive correlation between the variables. The strength values spanning from 0.00 to 0.10 indicate no agreement, 0.1–0.39 indicate weak agreement, 0.40–0.69 indicate moderate agreement, 0.70–0.89 indicate strong agreement, and 0.90–1.0 indicate almost perfect agreement (Schober & Schwarte 2018). Equation (11) was used to establish a linear relationship between the two variables.
formula
(11)
where R refers to the Coefficient of correlation, X and Y refer to the simulated and predicted sediment export, refer to the mean of simulated and predicted sediment export, n refers to the total number of observed data.

Model performance evaluation

We employed four statistical criteria to measure the performance and accuracy of the model. The reliability of the InVEST-SDR model in forecasting soil loss, sediment export and deposition were assessed using mean absolute error (MAE), ratio of RMSE and the standard deviation of the simulated data (RSR), coefficient of determination (), and mean absolute percentage error (MAPE) as shown in Equations (12)–(15).
formula
(12)
formula
(13)
formula
(14)
formula
(15)
where and illustrate simulated and predicted data values, N indicates the total number of observed land use classes, and shows mean of simulated data.

The instances of soil loss, sediment export, and deposition characterized by the lowest RSR, MAE, and MAPE values will be considered as the most accurately predicted by the InVEST-SDR model (Clement 2014; Ilie et al. 2020; Rezaiy & Shabri 2023).

In this section, we examine in detail soil loss, sediment export, and deposition, crucial components in understanding the dynamics of landscape erosion. Analyzing these factors provides an insightful perspective on the environmental processes affecting our study area. Our goal is to understand the complex patterns and the connections existing between LULC change, soil erosion, sediment export, and deposition by utilizing the InVEST-SDR model integrated with Revised Universal Soil Loss Equation (RUSLE) alongside careful data analysis. This research not only propels our knowledge of the local ecosystem but also lays the foundation for discussions on sustainable land management and conservation strategies. The outcomes derived from this study are potential to provide valuable insights for decision-makers and water users regarding the intricate linkage between LULC change, soil loss dynamics and sediment movement and their implications on land degradation, reservoir sedimentation, water quality pollution, and disruption of aquatic habitats.

LULC change classification

Confusion matrix was used to access spatial accuracy and resolution quality of the four (4) Landsat satellite imagery based on producer's accuracy, user's accuracy, overall accuracy, and Kappa coefficient (K) as shown in Supplementary material, Appendix 3, Tables S1–S4. The Kappa coefficient of the classified imagery was 0.90 in 1997, 0.90 in 2002, 0.94 in 2013 and 0.82 in 2022. This means that the classified satellite images have shown very strong agreement with the actual ground truth data, and the features of interest in the images were successfully and accurately identified (Mekuriaw 2019; Mfwango et al. 2022).

LULC change detection and analysis

Landsat 5 (TM), Landsat 7 (ETM+), Landsat 8 (OLI), and Landsat 9 (OLI-2) were used for LULC change detection and analysis for each study period as shown in Table 3. Seven (7) LULC change classes were detected in the study area, i.e., agricultural land, mixed shrub and grassland, dense forest, built-up area, water bodies, grassland, and mixed forest and shrub land as shown in Supplementary material, Appendix 2, Figures S1–S4. Agricultural land (34.93%), mixed shrub and grassland (21.89%) and dense forest (19.86%) are the LULC change classes dominating the study area. Deforestation has been observed in all watershed areas, particularly on the lowland slopes. During analysis, it was observed that agricultural land increased by 20.81%, built-up area by 27.20%, and water bodies by 1.95%. Meanwhile, dense forest decreased by 20.31%, grassland by 29.97%, mixed forest and shrub land by 22.51%, and mixed shrub and grassland by 25.58% in the consecutive 25 years from 1997 to 2022. The increase in agricultural land and built-up area could have been due to population growth demanding more land for crop production and settlement (Rizeei et al. 2016). The decrease in dense forest, grassland, mixed forest and shrub land, and mixed shrub and grassland could be due to deforestation driven by natural resource extraction such as timber, logging, charcoal, firewood, and natural calamities like wildfires and drought, or agricultural and urban development (Mfwango et al. 2022).

Table 3

LULC change detection and analysis

LULC1997
2002
2013
2022
Average areaAverage overall area% Aver. Change
km2%km2%km2%km2%km2%
Dense forest 45.87 27.63 62.39 37.58 13.12 7.90 10.53 6.34 32.98 19.86 −20.31 
Agricultural land 25.35 15.27 47.88 28.84 17.70 10.66 141.00 84.93 57.98 34.93 20.81 
Grassland 21.68 13.06 23.73 14.29 26.40 15.90 0.03 0.02 17.96 10.82 −29.97 
Mixed forest and shrub land 31.20 18.79 21.26 12.80 2.21 1.33 4.35 2.62 14.75 8.89 −22.51 
Mixed shrub and grassland 36.09 21.74 5.93 3.57 100.11 60.30 3.21 1.93 36.33 21.89 −25.58 
Water bodies 5.79 3.49 4.45 2.68 6.10 3.67 6.52 3.93 5.72 3.44 1.95 
Built-up 0.04 0.02 0.38 0.23 0.38 0.23 0.38 0.23 0.29 0.18 27.20 
LULC1997
2002
2013
2022
Average areaAverage overall area% Aver. Change
km2%km2%km2%km2%km2%
Dense forest 45.87 27.63 62.39 37.58 13.12 7.90 10.53 6.34 32.98 19.86 −20.31 
Agricultural land 25.35 15.27 47.88 28.84 17.70 10.66 141.00 84.93 57.98 34.93 20.81 
Grassland 21.68 13.06 23.73 14.29 26.40 15.90 0.03 0.02 17.96 10.82 −29.97 
Mixed forest and shrub land 31.20 18.79 21.26 12.80 2.21 1.33 4.35 2.62 14.75 8.89 −22.51 
Mixed shrub and grassland 36.09 21.74 5.93 3.57 100.11 60.30 3.21 1.93 36.33 21.89 −25.58 
Water bodies 5.79 3.49 4.45 2.68 6.10 3.67 6.52 3.93 5.72 3.44 1.95 
Built-up 0.04 0.02 0.38 0.23 0.38 0.23 0.38 0.23 0.29 0.18 27.20 

Negative sign indicates the decrease of LULC change class. Positive sign indicates the increase in LULC change class.

Overland soil loss

The InVEST-SDR model was employed within the Bontanga watershed to assess the extent of overland soil loss due to land use and land cover changes over a 25-year period. The model calculates the soil loss output by averaging the product of the number of upstream pixels susceptible to erosion and the estimated soil loss for each pixel, measured in tons per pixel per year. The actual soil loss results, as estimated by the InVEST-SDR model, are displayed in Supplementary material, Appendix 4, Table S1 and Figure 3, with the map showing the spatial distribution of soil loss within the watershed presented in Figure 4.
Figure 3

Actual soil loss from 1997 to 2022.

Figure 3

Actual soil loss from 1997 to 2022.

Close modal
Figure 4

Spatial distribution of soil loss within the watershed between 1997 and 2022.

Figure 4

Spatial distribution of soil loss within the watershed between 1997 and 2022.

Close modal

Soil loss transformation accelerated by water is complex and can be grouped into two categories: On-site soil loss that occurs at the site and principally has a potential impact on the reduction of soil fertility, the capacity of soil to retain water, and shallowing depth of top soil. The second category is off-site soil loss, which involves the transportation and deposition of eroded soil material to other locations, has an adverse impact on natural water pollution from agricultural chemicals, inundation and burial of top soil with sediments, and reservoir sedimentation (Lal 2014). It can be observed in Supplementary material, Appendix 4, Table S1 that soil loss accelerated by anthropogenic activities in the watershed is progressively intensifying over the years. Agricultural land dominates with 34.93% of the total watershed area, which produced 11,365.39 tons/year of sediments in 1997, and increased to 17,476.85 tons/year in 2002. In 2013, agricultural land experienced a subsequent decrease in sediment load of 5,391.98 tons /year, which was ultimately followed by an increase of 91,274.53 tons/year in 2022. An increase in soil loss in the watershed could be due to agricultural land expansion, changes in cultivation methods, or variations in rainfall intensity. However, a decrease in soil loss could be due to the implementation of better land management practices or land use conservation measures taken to mitigate soil erosion. Well-managed ecosystem services can provide food, fresh water, wood, fuel, and fiber. They can also regulate soil loss, floods, climate change, and water-borne diseases (Lhomme et al. 2020). The efficient conservation and management of ecosystem services aimed at mitigating water-accelerated soil loss is an essential requirement for fostering sustainable development and enhancing environmental protection.

Sediment export

The amount of the sediment exported from each pixel and reaches the stream in the study area was estimated using the InVEST-SDR model. Exported sediment can be used to validate simulated and observed sediment at the outlet of a watershed. In order to comprehend the impact of LULC changes on sediment export downstream of the watershed, the InVEST-SDR model utilizes RUSLE parameters extracted from rainfall, soil data and Landsat images, as depicted in Supplementary material, Appendix 1. The results revealed that expansion of agricultural land both on upland and lowland fields, resource extraction from the forest, and bush clearing are more vulnerable areas to sediment export downstream of the watershed. Agricultural land exported about 56.16% of sediment load downstream of the watershed, followed by dense forest at 13.39% and grassland at 13.30% as shown in Supplementary material, Appendix 4, Table S2 and Figure 5. The spatial distribution of sediment load exported to the river stream is shown in Figure 6. Agricultural practices within the Bontanga watershed encompass both upland and lowland areas, which could potentially contribute to the observed rise in sediment export from the watershed. Typically, lowland areas exhibit a low slope and steepness factor (LS), resulting in minimal sediment load generation (Woznicki et al. 2020). Resource extraction from forest areas and expansion of agricultural land could be the source of an increase in sediment export from dense forest and grassland. For a well-conserved watershed, forest and grassland should have played a more important role in keeping water clean and ensuring minimal sediment export to the stream compared to any other land cover class (de Mello et al. 2018).
Figure 5

Sediment exported downstream of the Bontanga watershed.

Figure 5

Sediment exported downstream of the Bontanga watershed.

Close modal
Figure 6

Spatial distribution of sediment exported downstream of the watershed between 1997 and 2022.

Figure 6

Spatial distribution of sediment exported downstream of the watershed between 1997 and 2022.

Close modal

Sediment deposition within the watershed

The spatial distribution of the quantities of the sediment deposition on the pixel from upslope sources that was trapped by different natural land vegetation covers was estimated using the InVEST-SDR model. This quantity assists researchers in monitoring the net changes in sediment on a pixel level, providing valuable insights into land degradation indices. The findings presented in Supplementary material, Appendix 4, Table S3 show that dense forest can retain 41.54% of the sediment load, followed by mixed shrub and grassland at 30.49% and grassland at 10.85%, when compared to other land use classes (Figure 7). Agricultural land can retain only 8.11% of the sediments generated in the watershed. The spatial distribution of sediment deposition in the watershed is shown in Figure 8. Areas prone to more deposition of sediment were observed more in the upland than in the lowland. This is due to the fact that the InVEST model assumes no sediment is exported to the stream during estimation of sediment deposition; hence, it assigns a ‘no data’ value to the pixels that are in the stream (lowland). Sediment deposition between dense forests, mixed shrub and grasslands, grasslands, and agricultural lands is primarily variable due to the characteristics of each land use type and the way they interact with water and sediment trapping capacity. Dense forests, and mixed shrub and grassland have a significant amount of vegetation and trees with dense networks of root systems that bind the soil particles together, preventing erosion and sediment movement (Gashaw et al. 2018). Also, the canopy of these plants intercepts rainwater, reducing the impact velocity of direct raindrops on the soil surface, which further decreases erosion. In areas where agricultural activities are intensively practiced, there is little or no natural vegetation cover. The absence of soil vegetation cover indicates no root system to trap and bind soil particles together, making it vulnerable to erosion (Degife et al. 2021).
Figure 7

Sediment deposition within the watershed.

Figure 7

Sediment deposition within the watershed.

Close modal
Figure 8

Spatial distribution of sediment deposition of the Bontanga watershed between 1997 and 2022.

Figure 8

Spatial distribution of sediment deposition of the Bontanga watershed between 1997 and 2022.

Close modal

Predicting soil loss, sediment delivery, and sediment deposition over the long term within the watershed

The prevailing soil loss, sediment delivery and retention of the study year 2022 in the watershed were assessed by assigning the InVEST-SDR model with the crop management factor (C) and conservation support practice factor (P) of the existing natural land vegetation. Meanwhile, predictions for the future soil loss were made by removing the crop management factor (C) and conservation support practice factor (P) from the model by assigning them with C-factor = 1 and P-factor = 1. The findings indicated that within the spectrum of five (5) land use and land cover change classes, agricultural land areas emerged as the predominant source of soil erosion, with a significant annual quantity of 91,274.53 tons per year, leading to substantial downstream sediment export of 4,659.80 tons per year. These figures surpass all other LULC change classes. In contrast, dense forest areas exhibited an impressive capacity for soil deposition, accounting for 45,593.45 tons per year, a capacity that surpasses all other land use classes within the study area. In the absence of substantial conservation measures within the watershed over the long term as shown in Supplementary material, Appendix 5, Figure S1 projections indicate that agricultural land will contribute to an alarming annual soil loss of 2,347,414.04 tons, resulting in a substantial downstream sediment export of 388,497.56 tons per year. Additionally, sediment deposition amounting to 1,048,258.78 tons per year is anticipated to be deposited within the agricultural field. The results of the analysis are displayed in Supplementary material, Appendix 5, Table S1.

Expansion of agricultural land without proper conservation measures can accelerate soil erosion and sediment export due to the disruption of natural vegetation cover. These natural covers play a crucial role in stabilizing soil and preventing erosion (Ali & Osman 2008). When vegetation is removed through plowing and tilling, the protective layer that helps bind soil particles together is lost, making the soil more susceptible to erosion. Without natural vegetation cover, raindrops can directly impact the soil surface, causing soil particles to detach and be transported by runoff water (Astuti & Suryatmojo 2019). An increased soil loss from agricultural land could lead to a significant reduction in soil nutrients necessary for crop growth and decreased crop yields (Zhang et al. 2021). Sediment transportation downstream of the watershed could lead to reservoir sedimentation, a reduction in water quality, and disruption of aquatic habitats. Sediments brought from the upper catchment and deposited on agricultural land may have a different nutrient composition as compared to the original soil, potentially leading to imbalanced nutrient levels and low crop yields. Thick layers of deposited sediments in the agricultural field can compact the soil, reducing its porosity and water-holding capacity and making it difficult for plant roots to access water, nutrients, and oxygen (Basiri Jahromi et al. 2020). Economically, soil loss and sediment deposition can lead to reductions in crop yields, arable land area, and land degradation. Sediment transportation downstream of the watershed can lead to higher costs for reservoir dredging by 15–100 times more than original dam construction (van der Knijff & Smith 2008). To mitigate these impacts, proper soil management and conservation practices in the watershed are essential. These practices may include the implementation of soil erosion control measures by adopting appropriate tillage techniques, contour farming, constructing sediment basins, and promoting vegetation buffers along watercourses (Bagarello 2017). Also, sustainable land use planning that considers erosion and sedimentation control that can minimize the negative effects of sediment deposition on agricultural land.

Sensitivity analysis of present and future soil loss, sediment delivery, and deposition as influenced by LULC change

Pearson's correlation coefficient was used to measure the strength and direction of the linear relationship between variations in LULC change classes and the corresponding simulated and predictive values. Schober & Schwarte (2018) suggested that a negative coefficient of correlation indicates a negative relationship between the variables, wherein an increase in one variable corresponds to a decrease in the other, and vice versa. On the other hand, a positive coefficient indicates a direct correlation between the variables. This implies that as one variable increases, the other tends to increase as well, and as one variable decreases, the other tends to decrease. The results of the analysis revealed that there is an almost perfect linear relationship between the variations in LULC change classes and the corresponding simulated and predictive values of soil loss, and sediment export (r = 0.999). However, sediment deposition exhibited a weak negative correlation of −0.218 between simulated and predicted values within the watershed. An observed correlation coefficient of 0.999 indicated that as one variable (LULC changes) changes, the other variables (soil loss, and sediment export) also change in a highly predictable and consistent manner.

Model performance evaluation

Forecasting accuracy of the InVEST-SDR model was achieved through utilization of statistical measurements, namely: mean absolute error (MAE), Ratio of RMSE and the standard deviation of the simulated data (RSR), coefficient of determination (R2), and mean absolute percentage error (MAPE). The statistical evaluation of the model is summarized in Supplementary material, Appendix 5, Table S2. During soil loss evaluation, MAPE demonstrates a lower value of 2,485.768 and higher R2 of 0.995, indicating the model's ability in accurately predicting soil loss. Furthermore, the MAE statistic reveals a lower value of 86,154.25 and higher R2 of 0.995, highlighting the robustness of the InVEST-SDR model in accurately predicting sediment export. On the other hand, sediment deposition yields a lower RSR value of 26.362 and R2 of 0.065, a lowest R2 signifying that the model has a limited ability to predict the sediment deposition data within the watershed. In accordance with the studies conducted by Clement (2014), Ilie et al. (2020), and Rezaiy & Shabri (2023), it can be concluded that the InVEST-SDR model is a suitable tool for predicting soil loss, and sediment export in the scarcely gauged watershed.

Limitations for model applications

The InVEST-SDR model is designed to estimate only sheet and rill soil erosion, limiting its applicability to the assessment of soil losses in gullies and landslide areas (Sharp et al. 2016). Therefore, it is imperative to conduct additional studies within Bontanga watershed utilizing a different model (e.g., the Soil and Water Assessment Tool (SWAT) and the HEC-RAS) (Marques et al. 2021) capable of identifying and quantifying the actual amount of soil losses, sediment export, and deposition in each pixel without converting the LULC change classes as bare soil. This approach is essential for enhancing the accuracy of planning, monitoring, and managing soil loss in the watershed. The absence of measured soil erosion records within the watershed or neighboring watersheds with similar characteristics hinders the validation of the results output derived from the InVEST-SDR model. Therefore, this study strongly recommends the installation of a gauging station downstream of the watershed. This station would facilitate the collection of hydrometeorological data, enabling the straightforward validation of research findings and enhancing the reliability of the InVEST-SDR model in the study area.

The InVEST-SDR model was employed to quantify and analyze soil loss, sediment export, and sediment deposition within the Bontanga watershed for the years 1997, 2002, 2013, and 2022. A confusion matrix was employed to evaluate the spatial accuracy and resolution quality of the four (4) Landsat images based on producer's accuracy, user's accuracy, overall accuracy, and Kappa coefficient (K). The Kappa coefficient of the classified imagery was 0.90 in 1997, 0.90 in 2002, 0.94 in 2013, and 0.82 in 2022. This analysis leads to the conclusion that the classified images exhibit a high level of agreement with the actual ground truth data, and the features of interest in the images were effectively and accurately identified. Pearson's correlation coefficient method was used to measure the strength and direction of the linear relationship between the variables. The observed correlation between variations in LULC change classes and the corresponding simulated and predictive values of soil loss, and sediment export indicated a strong positive correlation coefficient of 0.999. Conversely, for sediment deposition, a weak negative correlation of −0.218 was observed between simulated and predicted values within the watershed. Similarly, forecasting accuracy of the InVEST-SDR model was evaluated using four (4) statistical measurements namely; mean absolute error (MAE), Ratio of RMSE and the standard deviation of the simulated data (RSR), coefficient of determination (R2), and mean absolute percentage error (MAPE).

Based on the results, the following conclusions can be drawn:

  • Deforestation was observed in all the watershed areas, particularly on the lowland slopes within the study area, and the dominant land use and land cover change classes was agricultural land (34.93%), mixed shrub and grassland (21.89%) and dense forest (19.86%).

  • Soil loss for agricultural land was 11,365.39 tons/year which increased to 17,476.85 tons/year in 2002. In 2013, agricultural land experienced a soil loss of 5,391.98 tons /year, which ultimately increased to 91,274.53 tons/year in 2022.

  • Agricultural land, forest, and grassland are prominent land use and land cover change classes within the watershed that are more vulnerable to sediment export downstream of the watershed. Agricultural land exported about 56.16% of sediment load downstream of the watershed, followed by dense forest 13.39% and grassland 13.30%.

  • Dense forest deposited 41.54% of the sediment load, followed by mixed shrub and grassland 30.49% and grassland 10.85%, when compared to other land use classes. These phenomena may be attributed to the robust root systems of the plants in the soil, which effectively bind soil particles together, thus preventing erosion and sediment movement.

  • In the absence of substantial conservation measures within the watershed over a long period, predictive models foresee agricultural land contributing to an annual soil loss of 2,347,414.04 tons, resulting in a substantial downstream sediment export of 388,497.56 tons per year. Furthermore, an approximate annual sediment deposition of 1,048,258.78 tons is anticipated to be deposited within the agricultural field.

  • The lower MAPE value of 2,485.768 and higher R2 of 0.995, indicating the model's ability in accurately predicting soil loss.

  • The MAE statistic reveals a lower value of 86,154.25 and higher R2 of 0.995, which underscored the robustness of the model in accurately predicting sediment export.

  • Sediment deposition yields a lower RSR value of 26.362 and R2 of 0.065, a lowest R2 signifying that the model has a limited ability to predict the sediment deposition data within the watershed.

The author would like to thank Savanna Agricultural Research Institute (SARI) for providing updated meteorological data for this research, United States Geological Survey (USGS), for providing remote sensing data and digital elevation model (DEM), University for Development studies – Nyankpala Campus, and WACWISA (West Africa Center for Water, Irrigation and Sustainable Agriculture).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.

AbdelRahman
M. A. E.
2023
An overview of land degradation, desertification and sustainable land management using GIS and remote sensing applications
. In:
Rendiconti Lincei
, Vol.
34
, Issue
3
.
Springer International Publishing
.
https://doi.org/10.1007/s12210-023-01155-3
.
Abdulkareem
J. H.
,
Pradhan
B.
,
Sulaiman
W. N. A.
&
Jamil
N. R.
2019
Prediction of spatial soil loss impacted by long-term land-use/land-cover change in a tropical watershed
.
Geoscience Frontiers
10
(
2
),
389
403
.
https://doi.org/10.1016/j.gsf.2017.10.010
.
Abebe
Z. D.
&
Sewnet
M. A.
2014
Adoption of soil conservation practices in North Achefer District, Northwest Ethiopia
.
Chinese Journal of Population Resources and Environment
12
(
3
),
261
268
.
https://doi.org/10.1080/10042857.2014.934953
.
Adongo
T. A.
,
Kyei-Baffour
N.
,
Abagale
F. K.
&
Agyare
W. A.
2020
Assessment of reservoir sedimentation of irrigation dams in northern Ghana
.
Lake and Reservoir Management
36
(
1
),
87
105
.
https://doi.org/10.1080/10402381.2019.1659461
.
Akoglu
H.
2018
User's guide to correlation coefficients
.
Turkish Journal of Emergency Medicine
18
(
3
),
91
93
.
https://doi.org/10.1016/j.tjem.2018.08.001
.
Alhassan
E. H.
,
Ezekiel
A.
&
Akongyuure
D. N.
2014
Frame survey and fish catch assessment of the bontanga reservoir in Northern Ghana. June
.
Ali
F. H.
&
Osman
N.
2008
Shear strength of a soil containing vegetation roots
.
Soils and Foundations
48
(
4
),
587
596
.
https://doi.org/10.3208/sandf.48.587
.
Almagro
A.
,
Thomé
T. C.
,
Colman
C. B.
,
Pereira
R. B.
,
Marcato Junior
J.
,
Rodrigues
D. B. B.
&
Oliveira
P. T. S.
2019
Improving cover and management factor (C-factor) estimation using remote sensing approaches for tropical regions
.
International Soil and Water Conservation Research
7
(
4
),
325
334
.
https://doi.org/10.1016/j.iswcr.2019.08.005
.
Aneseyee
A. B.
,
Elias
E.
,
Soromessa
T.
&
Feyisa
G. L.
2020
Land use/land cover change effect on soil erosion and sediment delivery in the Winike watershed, Omo Gibe Basin, Ethiopia
.
Science of the Total Environment
728
,
138776
.
https://doi.org/10.1016/j.scitotenv.2020.138776
.
Arnold
J. G.
,
Kiniry
J. R.
,
Srinivasan
R.
,
Williams
J. R.
,
Haney
E. B.
&
Neitsch
S. L.
2012
Chapter 22 SWAT Input Data: .SOL. SWAT Input/Output File Documentation, Version 2012, 301–316. Available from: http://swat.tamu.edu/media/69365/ch22_input_sol.pdf.
Asma Belasri
A. L.
2016
Estimation of Soil Erosion Risk Using the Universal Soil Loss Equation (USLE) and Geo-Information Technology in Oued El. February, 98–107
.
Astuti
H. P.
&
Suryatmojo
H.
2019
Water in the forest: Rain-vegetation interaction to estimate canopy interception in a tropical borneo rainforest
.
IOP Conference Series: Earth and Environmental Science
361
(
1
).
https://doi.org/10.1088/1755-1315/361/1/012035
.
Bagarello
V.
2017
Effective Practices in Mitigating Soil Erosion from Fields. In Oxford Research Encyclopedia of Environmental Science (Issue August). https://doi.org/10.1093/acrefore/9780199389414.013.242
.
Bagherzadeh
A.
2014
Estimation of soil losses by USLE model using GIS at Mashhad plain, Northeast of Iran
.
Arabian Journal of Geosciences
7
(
1
),
211
220
.
https://doi.org/10.1007/s12517-012-0730-3
.
Basiri Jahromi
N.
,
Lee
J.
,
Fulcher
A.
,
Walker
F.
,
Jagadamma
S.
&
Arelli
P.
2020
Effect of biochar application on quality of flooded sandy soils and corn growth under greenhouse conditions
.
Agrosystems, Geosciences and Environment
3
(
1
),
1
8
.
https://doi.org/10.1002/agg2.20028
.
Belay
T.
&
Mengistu
D. A.
2021
Impacts of land use/land cover and climate changes on soil erosion in Muga watershed, Upper Blue Nile basin (Abay), Ethiopia
.
Ecological Processes
10
(
1
).
https://doi.org/10.1186/s13717-021-00339-9
.
Bendito
B. P. C.
,
Chaves
H. M. L.
&
Scariot
A.
2023
Erosion and sedimentation processes in a semi-arid basin of the Brazilian savanna under different land use, climate change, and conservation scenarios
.
Water (Switzerland)
15
(
3
).
https://doi.org/10.3390/w15030563
.
Boardman
J.
2006
Soil erosion science: Reflections on the limitations of current approaches
.
Catena
68
(
2–3
),
73
86
.
https://doi.org/10.1016/j.catena.2006.03.007
.
Bocco
G.
1991
Gully erosion: Processes and models
.
Progress in Physical Geography
15
(
4
),
392
406
.
https://doi.org/10.1177/030913339101500403
.
Borselli
L.
,
Cassi
P.
&
Torri
D.
2008
Prolegomena to sediment and flow connectivity in the landscape: A GIS and field numerical assessment
.
Catena
75
(
3
),
268
277
.
https://doi.org/10.1016/j.catena.2008.07.006
.
Clement
E. P.
2014
Using normalized Bayesian Information Criterion (Bic) to Improve Box – Jenkins Model Building
.
American Journal of Mathematics and Statistics
4
(
5
),
214
221
.
https://doi.org/10.5923/j.ajms.20140405.02
.
Degife
A.
,
Worku
H.
&
Gizaw
S.
2021
Environmental implications of soil erosion and sediment yield in Lake Hawassa watershed, south-central Ethiopia
.
Environmental Systems Research
10
(
1
).
https://doi.org/10.1186/s40068-021-00232-6
.
de Mello
K.
,
Valente
R. A.
,
Randhir
T. O.
,
dos Santos
A. C. A.
&
Vettorazzi
C. A.
2018
Effects of land use and land cover on water quality of low-order streams in Southeastern Brazil: Watershed versus riparian zone
.
Catena
167
(
December 2017
),
130
138
.
https://doi.org/10.1016/j.catena.2018.04.027
.
De Vente
J.
,
Poesen
J.
,
Verstraeten
G.
,
Govers
G.
,
Vanmaercke
M.
,
Van Rompaey
A.
,
Arabkhedri
M.
&
Boix-Fayos
C.
2013
Predicting soil erosion and sediment yield at regional scales: Where do we stand?
Earth-Science Reviews
127
,
16
29
.
https://doi.org/10.1016/j.earscirev.2013.08.014
.
European Commission
2002
Soil Erosion Risk in Europe. Cereal Research Communications, 36(January 2002), 939–942. http://eusoils.jrc.ec.europa.eu/esdb_archive/pesera/pesera_cd/pdf/P07ESF_ErosR7.pdf
.
Gashaw
T.
,
Tulu
T.
,
Argaw
M.
,
Worqlul
A. W.
,
Tolessa
T.
&
Kindu
M.
2018
Estimating the impacts of land use/land cover changes on Ecosystem Service Values: The case of the Andassa watershed in the Upper Blue Nile basin of Ethiopia
.
Ecosystem Services
31
,
219
228
.
https://doi.org/10.1016/j.ecoser.2018.05.001
.
Ghana Irrigation Development Authority
2010
National Irrigation Policy, Strategies and Regulatory Measures (Issue May)
.
Guo
Z.
,
Yan
Z.
,
PaErHaTi
M. E. Z. T.
,
He
R.
,
Yang
H.
,
Wang
R.
&
Ci
H.
2023
Assessment of soil erosion and its driving factors in the Huaihe region using the InVEST-SDR model
.
Geocarto International
38
(
1
).
https://doi.org/10.1080/10106049.2023.2213208
.
Huang
C.
,
Zhou
Z.
,
Teng
M.
,
Wu
C.
&
Wang
P.
2020
Effects of climate, land use and land cover changes on soil loss in the Three Gorges Reservoir area, China
.
Geography and Sustainability
1
(
3
),
200
208
.
https://doi.org/10.1016/j.geosus.2020.08.001
.
Ilie
O. D.
,
Ciobica
A.
&
Doroftei
B.
2020
Testing the accuracy of the arima models in forecasting the spreading of COVID-19 and the associated mortality rate
.
Medicina (Lithuania)
56
(
11
),
1
23
.
https://doi.org/10.3390/medicina56110566
.
Jakubínský
J.
,
Pechanec
V.
,
Procházka
J.
&
Cudlín
P.
2019
Modelling of soil erosion and accumulation in an agricultural landscape-A comparison of selected approaches applied at the small stream basin level in the Czech Republic
.
Water (Switzerland)
11
(
3
).
https://doi.org/10.3390/w11030404
.
Kanito
D.
,
Bedadi
B.
&
Feyissa
S.
2023
Sediment yield estimation in GIS environment using RUSLE and SDR model in Southern Ethiopia
.
Geomatics, Natural Hazards and Risk
14
(
1
).
https://doi.org/10.1080/19475705.2023.2167614
.
Karamage
F.
,
Zhang
C.
,
Liu
T.
,
Maganda
A.
&
Isabwe
A.
2017
Soil erosion risk assessment in Uganda
.
Forests
8
(
2
).
https://doi.org/10.3390/f8020052
.
Keshtkar
M.
,
Mokhtari
Z.
&
Sayahnia
R.
2022
Impact of land use/land cover change on soil retention service: A case of agricultural-urbanized Landscape in Northern Iran
.
Journal of Landscape Ecology(Czech Republic)
15
(
2
),
34
58
.
https://doi.org/10.2478/jlecol-2022-0010
.
Khodadadi
M.
,
Alewell
C.
,
Mirzaei
M.
,
Ehssan-Malahat
E.
,
Asadzadeh
F.
,
Strauss
P.
&
Meusburger
K.
2023
Understanding deforestation impacts on soil erosion rates using 137Cs, 239 + 240Pu, and 210Pbex and soil physicochemical properties in western Iran
.
Journal of Environmental Radioactivity
257
(
June 2022
).
https://doi.org/10.1016/j.jenvrad.2022.107078
.
Khosrokhani
M.
&
Pradhan
B.
2014
Spatio-temporal assessment of soil erosion at Kuala Lumpur metropolitan city using remote sensing data and GIS
.
Geomatics, Natural Hazards and Risk
5
(
3
),
252
270
.
https://doi.org/10.1080/19475705.2013.794164
.
Kosmas
C.
,
Detsis
V.
,
Karamesouti
M.
,
Kounalaki
K.
,
Vassiliou
P.
&
Salvati
L.
2015
Exploring long-term impact of grazing management on land degradation in the socio-ecological system of Asteroussia Mountains, Greece
.
Land
4
(
3
),
541
559
.
https://doi.org/10.3390/land4030541
.
Lal
R.
2014
Soil conservation and ecosystem services
.
International Soil and Water Conservation Research
2
(
3
),
36
47
.
https://doi.org/10.1016/S2095-6339(15)30021-6
.
Lhomme
P.
,
Michez
D.
,
Christmann
S.
,
Scheuchl
E.
,
El Abdouni
I.
,
Hamroud
L.
,
Ihsane
O.
,
Sentil
A.
,
Smaili
M. C.
,
Schwarz
M.
,
Dathe
H. H.
,
Straka
J.
,
Pauly
A.
,
Schmid-Egger
C.
,
Patiny
S.
,
Terzo
M.
,
Müller
A.
,
Praz
C.
,
Risch
S.
,
Kasparak
M.
,
Kuhlmann
M.
,
Wood
T. J.
,
Bogusch
P.
,
Ascher
J. S.
&
Rasmont
P.
2020
The wild bees (Hymenoptera: Apoidea) of Morocco
.
Zootaxa
4892
(
1
).
https://doi.org/10.11646/zootaxa.4892.1.1
.
Majoro
F.
,
Wali
U. G.
,
Munyaneza
O.
,
Naramabuye
F.-X.
&
Mukamwambali
C.
2020
On-site and off-site effects of soil erosion: Causal analysis and remedial measures in agricultural land – a review
.
Rwanda Journal of Engineering, Science, Technology and Environment
3
(
2
),
1
19
.
https://doi.org/10.4314/rjeste.v3i2.1
.
Marondedze
A. K.
&
Schütt
B.
2020
Assessment of soil erosion using the rusle model for the Epworth district of the harare metropolitan province, Zimbabwe
.
Sustainability (Switzerland)
12
(
20
),
1
24
.
https://doi.org/10.3390/su12208531
.
Marques
S. M.
,
Campos
F. S.
,
David
J.
&
Cabral
P.
2021
Modelling sediment retention services and soil erosion changes in Portugal: A spatio-temporal approach
.
ISPRS International Journal of Geo-Information
10
(
4
).
https://doi.org/10.3390/ijgi10040262
.
Mekuriaw
T.
2019
Evaluating impact of land-use/land-cover change on surface runoff using Arc SWAT model in sore and Geba Watershed, Ethiopia
.
Journal of Environment and Earth Science
.
January. https://doi.org/10.7176/jees/9-10-02
.
Mewded
M.
,
Abebe
A.
,
Tilahun
S.
&
Agide
Z.
2021
Impact of land use and land cover change on the magnitude of surface runoff in the endorheic Hayk Lake basin, Ethiopia
.
SN Applied Sciences
3
(
8
).
https://doi.org/10.1007/s42452-021-04725-y
.
Mfwango
L. H.
,
Kisiki
C. P.
,
Ayenew
T.
&
Mahoo
H. F.
2022
The impact of land use/cover change on surface runoff at Kibungo sub-catchment of Upper Ruvu catchment in Tanzania
.
Environmental Challenges
7
(
January
),
100466
.
https://doi.org/10.1016/j.envc.2022.100466
.
Morrow
S.
&
Smolen
M.
1975
Using Vegetation for Erosion Control on Construction Sites. Oklahoma Cooperative Extension Service, Stillwater, OK. https:/extension.okstate.edu/fact-sheets/print-publications/bae/using-vegetation-for-erosion-control-on-construction-sites-bae-1514.pdf
.
Mutayoba
E.
,
Kashaigili
J. J.
,
Kahimba
F. C.
,
Mbungu
W.
&
Chilagane
N. A.
2018
Assessing the impacts of land use and land cover changes on hydrology of the Mbarali River Sub-Catchment. The case of Upper Great Ruaha Sub-Basin, Tanzania
.
Engineering
10
(
09
),
616
635
.
https://doi.org/10.4236/eng.2018.109045
.
Ozsahin
E.
,
Duru
U.
&
Eroglu
I.
2018
Land use and land cover changes (LULCC), a key to understand soil erosion intensities in the Maritsa Basin
.
Water (Switzerland)
10
(
3
).
https://doi.org/10.3390/w10030335
.
Panagos
P.
,
Borrelli
P.
,
Meusburger
K.
,
van der Zanden
E. H.
,
Poesen
J.
&
Alewell
C.
2015
Modelling the effect of support practices (P-factor) on the reduction of soil erosion by water at European scale
.
Environmental Science and Policy
51
,
23
34
.
https://doi.org/10.1016/j.envsci.2015.03.012
.
Qiao
X.
,
Li
Z.
,
Lin
J.
,
Wang
H.
,
Zheng
S.
&
Yang
S.
2023
Assessing current and future soil erosion under changing land use based on InVEST and FLUS models in the Yihe River Basin, North China. International Soil and Water Conservation Research, xxxx. https://doi.org/10.1016/j.iswcr.2023.07.001
.
Renard
K.
,
Foster
G.
,
Weesies
G.
,
McCool
D.
&
Yoder
D.
1997
Predicting Soil Erosion by Water: A Guide to Conservation Planning With the Revised Universal Soil Loss Equation (RUSLE). (U.S. Depar). Agriculture Handbook No. 703
.
Rezaiy
R.
&
Shabri
A.
2023
Drought forecasting using W-ARIMA model with standardized precipitation index
.
Journal of Water and Climate Change
14
(
9
),
3345
3367
.
https://doi.org/10.2166/wcc.2023.431
.
Rizeei
H. M.
,
Saharkhiz
M. A.
,
Pradhan
B.
&
Ahmad
N.
2016
Soil erosion prediction based on land cover dynamics at the Semenyih watershed in Malaysia using LTM and USLE models
.
Geocarto International
31
(
10
),
1158
1177
.
https://doi.org/10.1080/10106049.2015.1120354
.
Sardari
M. R. A.
,
Bazrafshan
O.
,
Panagopoulos
T.
&
Sardooi
E. R.
2019
Modeling the impact of climate change and land use change scenarios on soil erosion at the minab dam watershed
.
Sustainability (Switzerland)
11
(
12
).
https://doi.org/10.3390/su10023353
.
Schober
P.
&
Schwarte
L. A.
2018
Correlation coefficients: Appropriate use and interpretation
.
Anesthesia and Analgesia
126
(
5
),
1763
1768
.
https://doi.org/10.1213/ANE.0000000000002864
.
Sharp, R., HT Tallis, T Ricketts, Anne D Guerry, Stephen A Wood, Rebecca Chaplin-Kramer, E Nelson, D Ennaanay, S Wolny, N Olwero, K Vigerstol, D Pennington, G Mendoza, J Aukema, J Foster, J Forrest, D Cameron, K Arkema, E Lonsdorf, C Kennedy, G Verute, J. D. 2016 InVEST Sediment Delivery Ratio (SDR) model. Stanford University, University of Minnesota, the Nature Conservancy, and World Wildlife Fund. Available from: http://releases.naturalcapitalproject.org/invest-userguide/latest/en/sdr.html.
Shikangalah
R.
,
Paton
E.
,
Jetlsch
F.
&
Blaum
N.
2017
Quantification of areal extent of soil erosion in dryland urban areas: An example from Windhoek, Namibia
.
Cities and the Enviroment(CATE)
10
(
1
),
8
.
https://digitalcommons.lmu.edu/cate/vol10/iss1/8/
.
Tadesse
L.
,
Suryabhagavan
K. V.
,
Sridhar
G.
&
Legesse
G.
2017
Land use and land cover changes and soil erosion in Yezat Watershed, North Western Ethiopia
.
International Soil and Water Conservation Research
5
(
2
),
85
94
.
https://doi.org/10.1016/j.iswcr.2017.05.004
.
Teng
M.
,
Huang
C.
,
Wang
P.
,
Zeng
L.
,
Zhou
Z.
,
Xiao
W.
,
Huang
Z.
&
Liu
C.
2019
Impacts of forest restoration on soil erosion in the Three Gorges Reservoir area, China
.
Science of the Total Environment
697
(
1
),
134164
.
https://doi.org/10.1016/j.scitotenv.2019.134164
.
Tu
T. A.
&
Mitani
Y.
2011
Human Impacts on Erosion and Deposition in Onga River Basin, Kyushu. May
.
van der Knijff
J. M.
&
Smith
C.
2008
Sedimentation in Our Reservoirs: Causes and Solutions. 71–101
.
Vanmaercke
M.
,
Poesen
J.
,
Broeckx
J.
&
Nyssen
J.
2014
Sediment yield in Africa
.
Earth-Science Reviews
136
,
350
368
.
https://doi.org/10.1016/j.earscirev.2014.06.004
.
Vigiak
O.
,
Borselli
L.
,
Newham
L. T. H.
,
Mcinnes
J.
&
Roberts
A. M.
2012
Comparison of conceptual landscape metrics to define hillslope-scale sediment delivery ratio
.
Geomorphology
138
(
1
),
74
88
.
https://doi.org/10.1016/j.geomorph.2011.08.026
.
Wang
D.
,
Yuan
Z.
,
Jing
D.
&
Zhao
C.
2021
Response of grassland slope erosion to vegetation characteristics
.
IOP Conference Series: Earth and Environmental Science
668
(
1
).
https://doi.org/10.1088/1755-1315/668/1/012013
.
Woznicki
S. A.
,
Cada
P.
,
Wickham
J.
,
Schmidt
M.
,
Baynes
J.
,
Mehaffey
M.
&
Neale
A.
2020
Sediment retention by natural landscapes in the conterminous United States
.
Science of the Total Environment
745
,
140972
.
https://doi.org/10.1016/j.scitotenv.2020.140972
.
Yaekob
T.
,
Tamene
L.
,
Gebrehiwot
S. G.
,
Demissie
S. S.
,
Adimassu
Z.
,
Woldearegay
K.
,
Mekonnen
K.
,
Amede
T.
,
Abera
W.
,
Recha
J. W.
,
Solomon
D.
&
Thorne
P.
2022
Assessing the impacts of different land uses and soil and water conservation interventions on runoff and sediment yield at different scales in the central highlands of Ethiopia
.
Renewable Agriculture and Food Systems
37
,
S73
S87
.
https://doi.org/10.1017/S1742170520000010
.
Zakaria
H.
,
Abujaja
A. M.
,
Adam
H.
&
Nabila
A. Y.
2014
Factors affecting Farmers willingness to pay for improved irrigation service : A case study of Bontanga Irrigation Scheme in Northern Ghana
.
International Journal of Agricultural Economics and Extension
2
(
1
),
68
76
.
Zhang
L.
,
Huang
Y.
,
Rong
L.
,
Duan
X.
,
Zhang
R.
,
Li
Y.
&
Guan
J.
2021
Effect of soil erosion depth on crop yield based on topsoil removal method: A meta-analysis
.
Agronomy for Sustainable Development
41
(
5
),
1
13
.
https://doi.org/10.1007/s13593-021-00718-8
.
Zhou
P.
,
Luukkanen
O.
,
Tokola
T.
&
Nieminen
J.
2008
Effect of vegetation cover on soil erosion in a mountainous watershed
.
Catena
75
(
3
),
319
325
.
https://doi.org/10.1016/j.catena.2008.07.010
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).

Supplementary data