Modeling the sediment yield and estimating the best management practices in the Seybouse basin, Northeastern Algeria

Algeria is a country situated in the Mediterranean region, mostly in semi-arid and arid zones. The annual precipitation ranges from 400 to 670 mm, along the coastal area of the Tell (Mrad et al. 2020), the highest amount of rainfall, exceeding 1,000 mm per year, is observed in the eastern part of the country. Unfortunately, due to uncontrolled urbanization, the population is increasingly at risk of flooding. Recent disasters such as the Bab El Oued flood in 2001 and the heat wave of 2003 and 2017 have highlighted the rise of extreme weather events in Algeria. Such incidents are likely to trigger erosion and sediment transport (Tesema & Leta 2020; Hassen et al. 2022; Leta et al. 2023). According to Markhi et al. (2019), the physical process of sediment transport is responsible for the development of river platforms, such as meandering and river braiding. Several factors can affect sediment transport, including climatology, drainage area, site geology, slope, and the intensity and duration of rainfall (Roy et al. 2020). Managing water and soil resources is difficult because of the intricate relationships between the many physical, social, and ecological aspects that are included in the watershed planning process (Pal et al. 2020; Arjomandi et al. 2021; Erkan & Nihat 2022).

Several studies have been conducted in Algeria to investigate erosion in rivers and estimate sediment transport over time and space. Achaite & Ouillon (2007) analyzed sediment washdown over the Wadi Abd basin, which covers an area of 2,480 km², and estimated it to be 136 t/km² year for the period between 1973 and 1995. In the same region, El Ahcene et al. (2013) found that the annual average of specific degradation in the Wadi Bellah watershed was about 610 t/km² year. Bouzeria et al. (2023) found that erosion rates depend not only on rainfall and runoff but also on soil type, vegetation, and the topography of the area. Over the past decade, the National Agency of Dams in Algeria has experienced significant growth due to efforts to address reservoir siltation. The Algerian government has invested heavily in this process to ensure that downstream reservoirs in the study watersheds can meet the increasing demand for drinking water and irrigation. As a result, these reservoirs are being mobilized or will be mobilized in the future to fulfill these needs. A number of models, including the Revised Universal Soil Loss Equation (RUSLE), the Water Erosion Prediction Project (WEPP), and the Soil and Water Assessment Tool (SWAT), have been developed in various regions to quantify soil erosion and sediment production. Among these models, RUSLE is widely used for studying soil erosion by water due to its simplicity, despite its extensive input data requirements. However, researchers around Algeria have faced challenges in assessing and predicting soil erosion accurately using these models (Bouguerra et al. 2017; Sahli et al. 2019; Salhi et al. 2022; Sakhraoui & Hasbaia 2023). For the purpose of simulating and describing water and sediment transport in catchments of varying sizes and locations across the world, the SWAT is currently being utilized frequently and effectively (Markhi et al. 2019; Erkan & Nihat 2022; Daba et al. 2023), to assess agricultural management practices (Epelde et al. 2015), in order to assess the consequences of land use and land cover (LULC) (Alawi & Özkul 2023), to assess the spatial and temporal sediment transfer in many watersheds (Zettam et al. 2017; Koycegiz et al. 2021; Naseri et al. 2021). The SWAT has been utilized by numerous authors in North Africa (Markhi et al. 2019; Echogdali et al. 2022) including Algeria (Hallouz et al. 2018a; Kateb et al. 2020; Maref et al. 2022; Salhi et al. 2022). According to these studies, land degradation and soil erosion have an impact on Algeria western, eastern, and northeastern highlands. Furthermore, the negative effects of soil erosion and degradation change the potential of the country water resources, which increases the frequency of drought episodes. The great majority of studies published in the field were conducted to investigate the regional and temporal distribution of sediment output. To our knowledge, there are no studies in the Seybouse basin that apply best management approaches and decision-making to minimize sediment production. As a result, it was hoped that the selection of the Seybouse basin, with its Mediterranean climate and extensive agricultural activity, would serve as a successful model for filling this area. Due to its regional water resources, the originality of the paper appears to hold out a quantitative assessment of the best land management practices to reduce and limit soil losses in the sediment of the Seybouse watershed in the future. The most significant advantage of the model is the assessment of the effects of land management techniques on long-term water and sediment loads within a watershed (Mosbahi & Benabdallah 2020). However, before execution, it is necessary to build models that simulate various management situations to make suitable and future sustainability strategic decisions.

For that reason, the study will involve simulating stream flow and sediment yield for the Seybouse basin and analyzing the spatial sediment yield. The main purpose behind employing the SWAT model was to assess the impact of conservation measures implemented in soil erosion-prone regions as part of a comprehensive watershed management strategy. It will be determined whether three management techniques, such as terracing, grade stabilization structures (GSSs), and vegetated filter strips (VFSs), are effective at lowering sediment yield. The results of this study will provide insights into the most effective land management practices for reducing sediment yield in the Seybouse watershed, which could have important implications for water resources and drought mitigation in the region.

The subsequent sections of this paper are structured into five parts, as follows: the second section “Materials and methods” provides a geographical assessment of the study area, the methodology employed in constructing the SWAT model, and data input; the third section presents the results of the performance model and evaluation of the best management practices (BMPs); the fourth section “Discussion” examines major findings resulting from the analysis in this study; and finally, the fifth section “Conclusions” summarizes the work and provides prospective possibilities for further inquiry.

The SWAT model primarily requires the following contextual data: topography, land use, soil, and climate. Topography is important for sediment susceptibility growth because it affects the stream network's density, the flow's erosive power, and its velocity (Arabameri et al. 2019). In this study, elevation is extracted from the United States Geological Survey (USGS) Global with a spatial resolution of 30 m; this map is projected in a WGS 1984 UTM zone 32 system (Figure 1).

Due to the lack of available data on relative humidity, wind speed, and solar radiation for the study area, we used weather data available at the Global Weather Data for SWAT (GWDS) database. The GWDS database uses the Climate Forecast System Reanalysis version 3 (CFSR.v3) dataset, which is based on the National Centers for Environmental Prediction (NCEP) for a period of 36 years (1979–2014) (available at: https://swat.tamu.edu/data/cfsr).

Precipitation data are a critical input for the SWAT. The data were collected from the National Agency for Resources Hydraulic (ANRH) in Algeria for a period of 42 years (1970–2012), as shown in Table 1. For monthly stream flow, we used data from three gauge stations obtained from the ANRH in Algeria for a period of 40 years (1970–2010), as seen in Figure 1 and Table 1.

To meet the different objectives, several evaluation criteria have been developed, which are either graphic criteria or analytical criteria. Since the objective functions are mutually constrained and each objective function places different emphasis on different systematic or dynamic behavioral errors, it is difficult for a hydrologist to clearly evaluate the advantages and disadvantages of the solutions of parameter optimization (Huo & Liu 2020). We used criterion items from the literature to assess the efficacy of the stream flow and the sediment prediction model produced from the SWAT. The calibrated model performance was evaluated using parameters such as the determination coefficient (R²) and the Nash–Sutcliffe efficiency (NSE) coefficient (Nash & Sutcliffe 1970). For monthly flow simulations, the model is considered adequate if R² is greater than 0.60 and NSE is greater than 0.50. Comparably, monthly sediment performance that is deemed good is attained when R² is higher than 0.40 and NSE is higher than 0.45 (Leta et al. 2023).

To optimize land management and improve decision-making, we selected three BMPs: terracing, GSSs, and VFS. In order to implement BMPs at the watershed scale, each sub-basin's average yearly sediment output is examined (Nepal & Parajuli 2022). The parameter values in the SWAT models were changed to simulate these different BMP situations, and the impact on sediment yield in each scenario was examined at sub-watershed levels.

VFSs are vegetated areas located between surface water bodies, cropland, grazing land, and forest land that can offer erosion control (Waidler et al. 2011). Improved vegetation practices can be achieved through range planting and seeding on highly erodible sites. The implementation of VFS in the SWAT model requires generating parameters in the scheduled management operation file for each HRU, including Manning's N for channel, flag for simulation, ratio of field area to filter strip area, fraction of HRU draining to the most concentrated 10% of the VFS area, and fraction of flow within the most concentrated 10% of the fully channelized VFS (Cibin et al. 2018; Xue et al. 2019).

Terracing is a soil conservation technique that is implemented where excessive slope length and soil erosion caused by water are problems and where there is a need to conserve water. In the SWAT model, the implementation of terracing involves generating parameters in file management operation (.ops) for each HRU. These parameters include the initial soil conservation service (SCS) runoff curve number (CN_2), the Universal Soil Loss Equation (USLE) practice factor (USLE_P), and the average slope length (SLSUBBSN).

GSS is a physical structure composed of concrete, rock, or steel that helps to stabilize the grade and reduce erosion (Palmate & Pandey 2021). To simulate the GSS in the SWAT model, the parameters that require adjustment are the channel erodibility factor (CH_ERODMO) and the average slope of the main channel throughout the channel length (CH_S2).

Model calibration is performed by carefully selecting the values of the model input parameters and comparing the model predictions for a given set of assumed conditions with the observed data for the same conditions (Arnold et al. 2012; Ricci et al. 2020). In this research, 16 of the most sensitive parameters are listed in Table 2 for calibration stream flows that are linked to the stations under consideration. The same procedures and period of calibration and validation were used for the Ain Berda. Seven parameters used in sediment calibration are presented in Table 3. These stream flow and sediment calibration parameters were selected from a variety of sources of literature depending on how frequently they were used (Ricci et al. 2020; Khorn et al. 2022). As indicated in Tables 2 and 3, the range of most parameter sensitivity is calibrated for stream flow and sediment yield, respectively.

The spatial sediment yield is an important factor to consider when assessing the impact of land management practices on soil erosion and sediment production in a watershed (Arabameri et al. 2019; Daba et al. 2023). Figure 7 illustrates the spatial distribution of the average annual simulated sediment yield over the 26 sub-basins of the Seybouse basin. Forty-three percent of the basin area falls into the high erosion class (10–15 t/ha year), with the corresponding sub-basins being 4, 5, 13, 15, and 16. These areas were characterized by steep slopes, sparse vegetation cover, and intensive land use practices such as cultivation and high rainfall. The SWAT model reproduction revealed a difference in sediment yield due to geographic variance in major landscape information, soil types, and land use. Land cover, in addition to rainfall, is a critical component in determining whether sections within sub-basins are at high risk of erosion (Bouzeria et al. 2023). A total of 24% of the basin's surface is susceptible to moderate soil erosion (7–10 t/ha year), while 7% is prone to slight erosion (<7 t/ha year). According to the model findings, there is a significant risk of soil erosion in the catchment area in the central and southeast regions. The severity of the sediment yield (Table 4) differs from region to region.

In recent years, the SWAT model has gained widespread recognition as a useful instrument for evaluating the production of sediment in sub-basins. The purpose of this study was to capitalize on the SWAT model to simulate sediment yield and create BMPs in the Seybouse watershed in northern Algeria. During calibration, the R² values for the monthly stream flow phase (1970–1997) and sediment yield at Ain Berda station (1998–2010) were 0.90 and 0.74, respectively, and the NSE values were 0.84 and 0.74, respectively. The analysis shows that the SWAT model adequately captures the hydrologic characteristics of the Seybouse basin. This model can be effectively used for land use management. With an average annual sediment production of 14.43 t/ha year, 71% of the entire soil loss in the Seybouse basin is estimated to have occurred. Furthermore, the sediment yields in 12 sub-basins were found to be below the acceptable rate of soil loss, but high to severe soil erosion was found in 14 significant areas (4, 5, 13, 15, 16, 2, 6, 9, 10, 14, 7, 8, 11, and 12). In order to limit sediment in the basin, three soil and conservation measures – VFS, GSS, and terracing – were assessed in the SWAT as BMPs. The application of this management to the sub-basins leads to significant reductions in average annual sediment yield by 37.3, 20.40, and 42.30%, respectively. Among these strategies, terracing results in the greatest reduction, followed by VFS. With regard to the successful use of BMPs in a specific catchment or other comparable situations, these findings are of great importance to policymakers and water resource engineers.

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

The authors declare there is no conflict.