Introduction

The climate of Tunisia is quite erratic and ranges from arid to semi-arid. Drought is one of the most concerning effects of this variability (Henia 1987). Drought can be defined as a temporal imbalance of water availability consisting of persistent lower-than-average precipitation of uncertain frequency, duration, and unpredictable severity (Pereira et al. 2002). The need to construct hydraulic systems has always been motivated by a lack of water supply. Hydraulic action has always been intended to permanently store extra water in rainy years (in dams and aquifers) so that it can be used in dry years (ITES 2014). These hydraulic works are constructed on the main wadis at the level of the Ichkeul watershed and stop practically all of the freshwater inflow to the lake. Climate change theory increased water demand, and the building of dams on the wadis of Sejnane, Joumine, Melah, Ghezela, and Tine are all factors that have considerably worsened the overall status of the National Park of Ichkeul. All these factors have caused an ecological and hydrological imbalance in this aquatic ecosystem. By executive order n°80–1608 on December 18, 1980, the Ichkeul National Park was created. The lake and marshes of Ichkeul have long been acknowledged as one of the four most important wetlands of the western Mediterranean basin, together with Donana in Spain, the Camargue in France, and El Kala in Algeria. It is also one of the few locations in the world to be included in three international agreements. It is the last large freshwater lake in North Africa. In 1979, the Ichkeul National Park was included in the UNESCO list of World Natural Heritage sites. After that, the indigenous population of the park and the human initiatives related to conservation had already prompted it to be included in the list of UNESCO biosphere reserves in 1977. Finally, the convention of RAMSAR in 1980, gave the lake and the marshes of Ichkeul the status of the wetland of international importance as a place of wintering for thousands of migratory water birds, among which some species are threatened (ANPE: National Agency for Environmental Protection 1994). The Ichkeul National Park is characterized by a very particular hydrological system based on the seasonal alternation of water levels and salinity. During the winter, the Ichkeul lake is fed by freshwater from its catchment area and flows into the Bizerte Lake via the Tinja wadi, but during the summer, the current reverses and the lake receives marine flows. Ichkeul's aquatic ecosystem (lake and marshes) attracts hundreds of thousands of migrating birds like ducks, coots, greylag geese… Due to a shortage of fresh water over the winter, the grass fields are naturally and swiftly harmed by the salinization of the lake waters. As a result of this dire condition, the number of birds wintering in Ichkeul has plummeted, owing to a scarcity of food and fresh water for this avifauna. Detailed characterization and modeling of the study area can provide insight into the effects of climate change and hydraulic facilities on the hydrological environment of the Ichkeul. Modeling hydrological behavior is essential to address problems related to water resource management, land use planning, or any of the various aspects of hydrological risk. Several hydrological models of watersheds are available in the literature, each with its unique set of characteristics and field of use. The Hydrologic Simulation Package Fortran (HSPF) (Holtan and Lopez 1971), the European Hydrological System (EHS) (Abbott et al. 1986), the Soil and Water Assessment Tool (SWAT) (Arnold et al. 1998), and the Hydrologic Engineering Centre Hydrologic Modeling System (HEC-HMS; HEC 2000) are the most widely used watershed-scale models.

Because runoff is a resolution feature for hydrologic models, a model that can simulate runoff realistically is needed to estimate the flow value. According to Pereira et al. (2016), SWAT, one of these models, is the most effective at reproducing watershed hydrologic functioning on many occasions. Numerous studies show the successful application of the SWAT model around the world (Abbaspour 2007; Grusson 2015; Sisay et al. 2017; Kumar et al. 2018; Hosseini and Khaleghi 2020; Muthee et al. 2022; Tarekegn et al. 2022). The most common use of the SWAT model is to perform water balances. The model has been successfully used to simulate outflows from watersheds with sizes ranging from 122 to 246 Km2 (Arnold and Allen 1999). But also, according to Arnold and Allen (1999), conducting the overall water balance on watersheds ranging in size from 2,253 to 304,620 Km2 gives very satisfactory results. Even at the continental scale, this model has proven performance (Schuol et al. 2008; Abbaspour et al. 2015). Once the model has been chosen, its ability to represent reality must be evaluated. This is most often done by comparing the simulated results with observations. In this study, using a daily time step, the capacity of the SWAT model to reproduce the flow observed at the outlet of the uncontrolled part of the Ichkeul watershed will be tested to have an idea of the water resources of the area of interest, to determine the need for the freshwater of the Ichkeul lake, and then later, to model its salinity and quantify the sediments to establish guarantees for its survival.

Material and methods

Description of the study area

The Lake Ichkeul watershed is an exoreic basin located in the extreme north of Tunisia (9°30'N, 37°00'S) (Fig. 1). It has an area of about 2232 Km2 with an overall slope of SW-NE with altitudes ranging from 1 to 714 m. Administratively, the basin extends over three governorates of northern Tunisia that are Bizerte, Beja, and Manouba, and reveals the delegations of Bizerte, Tinja, Mateur, Ghezela, Hechachna, Sejnane, Joumine, Tebourba, El Battane, Beja, Medjez El Bab, and Nefza. The watershed of Ichkeul belongs to the hydrological unit of northern Tunisia Nefza-Ichkeul, according to the subdivision of the Tunisian territory into hydrological basins. The Nefza-Ichkeul complex covers an area of 4865 Km2. Its average annual liquid contribution is estimated at 860 Mm3. The abundant surface waters of the watershed and their chemical qualities make this region a true water tower of the Tunisian territory (Ben Mammou 2006). This region has a fairly uniform Mediterranean climate; it is a warm temperate zone between the polar front and the trade wind front. Winter is frequently the wettest time of year. It is globally included between the isohyets 600 and 900 mm and, consequently, it belongs to the humid to the subhumid bioclimatic stage with a typically Mediterranean climate. The main rivers that cross the watershed of Ichkeul are Sejnane, Joumine, Tine, Melah, and Ghezela. As such, five dams have been installed on these wadis that are; the dams of Joumine (1983), Ghezela (1984), Sejnane (1994), Melah (2014), and Tine (2015). These dams form a sort of belt that subdivides the Ichkeul watershed into two almost equal parts in terms of area. The first part is controlled by the dams and the second part is not controlled. The downstream part shelters the national park of Ichkeul, placed in the plain of Mateur, 25 km southwest of Bizerte.

Fig. 1
figure 1

Location map of the study area

Full size image

SWAT model

The SWAT (Soil and Water Assessment Tool) model is a continuous-time, semi-distributed, process-based river basin model (Arnold et al. 2012). SWAT was developed to predict the impact of land management practices on water, sediment, and agricultural chemical yields in large complex watersheds with varying soils, land use, and management conditions over long periods (Neitsch et al. 2002). SWAT is a model whose process representation is complex and based on physical representation. Differentiating hydro system components is possible in a variety of ways. It is a semi-distributed model allowing a discretization of space based on the topographic reality. Despite its complex representation of hydrological processes, the computational resources required to run it are moderate, allowing its use on large temporal and spatial scales (Grusson 2015). SWAT analyzes the entire watershed, subdividing it first into sub-watersheds and then into several sections called hydrologic response units (HRUs). Each HRU has specific land use, soil, and slope characteristics. HRUs have been used to describe spatial heterogeneity in land use, soil type, and slope class in a watershed. The hydrological cycle simulated by SWAT is based on the water balance equation Eq. 1.

$${\mathbf{S}\mathbf{W}}_{{\varvec{t}}}={\mathbf{S}\mathbf{W}}_{0}+{\sum }_{{\varvec{i}}=1}^{{\varvec{t}}}({{\varvec{R}}}_{\mathbf{d}\mathbf{a}\mathbf{y}}-{{\varvec{Q}}}_{\mathbf{s}\mathbf{u}\mathbf{r}\mathbf{f}}-{{\varvec{E}}}_{{\varvec{a}}}-{{\varvec{W}}}_{\mathbf{s}\mathbf{w}\mathbf{e}\mathbf{e}\mathbf{p}}-{{\varvec{W}}}_{\mathbf{g}\mathbf{w}})$$
(1)

where SWt is the final water content in (mm), t is time in a day, Rday is the precipitation amount on specific days i (mm), Qsurf is the runoff amount on specific days i (mm), Ea is evapotranspiration amount on a day i (mm), Wsweep is the amount of water percolated into the vadose zones on a day i (mm), and Wgw is the return amount of flow on a day i (mm).Two methods can be employed to separate precipitation into runoff and infiltrated water in SWAT: the SCS curves (USDA 1972) or the Green and Ampt curves (Green and Ampt 1911). The first technique determines runoff water amounts based on soil moisture content. Based on matric potential and hydraulic conductivity, the second technique calculates the amount of infiltrated water, but it requires sub-daily precipitation data, which is less common. Therefore, the SCS curve number method was used in this work to calculate the amount of water available for runoff The number of the SCS curve is described using Eqs. 2 and 3:

$${{\varvec{Q}}}_{\mathbf{s}\mathbf{u}\mathbf{r}\mathbf{f}}=\frac{{{({\varvec{R}}}_{\mathbf{d}\mathbf{a}\mathbf{y}}-0.2{\varvec{S}})}^{2}}{{({\varvec{R}}}_{\mathbf{d}\mathbf{a}\mathbf{y}}-0.8{\varvec{S}})}$$
(2)
$${\varvec{S}}=25.4\left(\frac{1000}{{\varvec{C}}{\varvec{N}}}-10\right)$$
(3)

where Qsurf is the depth of runoff in (mm), Rday precipitation (mm), S is maximum potential retention and, CN is the curve number.

The curve number depends on the soil, the soil use and the slope, and the soil moisture content. For each pair of soil/soil use, three NCs are determined: NC1 in dry conditions, NC2 in medium moisture conditions, and NC3 in wet conditions.

Regarding evapotranspiration, the SWAT model uses potential evapotranspiration to calculate evapotranspiration. Potential evapotranspiration is defined as the maximum evaporation potential for reference vegetation. Three methods that are included in the model are used to determine it: Penman–Monteith (Monteith 1965), Priestley-Taylor (Priestley and Taylor 1972), and Hargreaves (Hargreaves et al. 1985). The Penman–Monteith method requires solar radiation, air temperature, wind speed, and relative humidity. The Priestley-Taylor method requires radiation, air temperature, and relative humidity, whereas the Hargreaves method requires only air temperature. Therefore, given the availability of climate data for the study area, the Hargreaves method was used. The potential evapotranspiration was calculated by Eq. 4.

$${{\varvec{E}}}_{0}=\frac{0.0023{{\varvec{H}}}_{0}{\left({{\varvec{T}}}_{\mathbf{m}\mathbf{a}\mathbf{x}}-{{\varvec{T}}}_{\mathbf{m}\mathbf{i}\mathbf{n}}\right)}^{0.5}({{\varvec{T}}}_{{\varvec{a}}{\varvec{v}}}+17.8)}{{{\varvec{L}}}_{{\varvec{v}}}}$$
(4)

where E0 is potential evapotranspiration (mm.d−1), Lv is the latent heat of vaporization of water (MJ.kg−1), H0 is the extra-terrestrial radiation (MJ.m−2. d−1), and Tmax, Tmin, and Tav are, respectively, the maximum, minimum, and average temperatures of the day (°C).

Model input

In this study, the meteorological data used for the SWAT model are precipitation, minimum and maximum air temperature, relative humidity, wind speed, and solar radiation. In addition, the spatial data required are the digital elevation model (DEM), land cover, and soil data (Fig. 2a, b).

Fig. 2
figure 2

Land use map a and Soil map b of the uncontrolled part of the Ichkeul watershed

Full size image

Digital elevation model (DEM)

The definition of sub-watersheds in a SWAT project is derived from the definition of the river network, which is based on the topography. The resolution of a digital terrain model, which is raster data consisting of a network of cells or pixels containing elevation values, affects the definition of the hydrographic network and the delimitation of the watershed. For this work, a DEM with a resolution of 12.5 × 12.5 m was downloaded from https://search.asf.alaska.edu/.

Land use/land cover

The Sentinel-2A satellite image, which has a spatial resolution of 20 m, was processed to produce the land cover data. The year of observation is 2016. To identify the land use, an unsupervised classification is then performed. In the attributing table, each land cover in the SWAT model is represented by a four-letter code. Seven land cover classes were identified and reclassified to match the SWAT database. The dominant classes are cropland 69.4% and forest 12.7%. Lake Ichkeul covers an area of about 13% of the area of interest, while urbanized areas represent only 0.008% of the uncontrolled part of the watershed (Fig. 2a).

Soil data is a crucial component of any hydrological model. The region's soil map was taken from the SOTWIS-Tunisia map (version 1.0), produced by ISRIC, FAO, and UNEP, under the direction of the International Union of Soil Sciences (IUSS). Twelve soil units were identified in the study area and used in the model. These units are presented in Fig. 2b.

Weather data/daily flow

Daily climate data have been provided by the National Institute of Meteorology (INM), the Regional Commissioner for Agricultural Development of Bizerte (CRDAB), and the Directorate General of Dams and Large Hydraulic Works (DG-BGTH). The meteorological data used in this study are daily precipitation, minimum and maximum temperatures, relative humidity, wind speed, and solar radiation for the period between 2007 and 2019. These data are generated in the following stations: Sidi Ahmed, Tinja, Mateur, Battan, Sejnane, Ghezela, Melah, and Joumine (Fig. 3). The data on outflow comes from the station Ecluse-Sidi Hassoun, which is situated downstream of the Ichkeul watershed. The National Environmental Protection Agency (ANPE) records this data three times a day. Dam outflow data, also provided by DG-BGTH, is used to make the necessary corrections to the outflows recorded by the above station.

Fig. 3
figure 3

Location of the meteorological stations and the flow-out measurement station

Full size image

The lock of Tinja, built in the late 1980s on the wadi Tinja, controls the exchange of water between Lake Ichkeul and the lagoon of Bizerte. Since 2015, Lock management is completely absent to avoid inaccurate outflow measurements.

Model setup

The SWAT software is a single executable file (.exe) that runs the model equations and generates output files. Arc SWAT is an ESRI-ArcGIS software module that allows users to import and create input files for the SWAT model and facilitates model configuration by leading users through a step-by-step process. This procedure starts with the import of georeferenced layers like topography, watershed boundaries, land use, soil characteristics, and so on, and ends with the calibration and validation step. The overall workflow of the SWAT model and SWAT-CUP is shown in Fig. 4. The required spatial data were projected into the same projection system, WGS 1984 UTM zone 32 N, using ArcGIS version 10.3.

Fig. 4
figure 4

Detailed methodological framework

Full size image

Manual pre-definition of the stream network and sub-watersheds

Drainage networks are a fundamental condition for watershed delineation and an indispensable component in hydrological modeling. Therefore, reasonable and accurate watershed delineation is a precondition for runoff, sediment transport, water quality simulation, and basin resource management (Billen et al. 2005). The stream network process usually starts by allocating flow directions for each DEM cell, then analyzes the flow accumulation, and finally selects those cells that have a total higher threshold of flow accumulation than a defined value (Meisels et al. 1995). Many factors, including the DEM spatial resolution, the calculated algorithm, and the physical characteristics of the basin, can directly affect the accuracy of drainage networks derived from DEM data (Li and Wong 2010; Ariza-Villaverde et al. 2015). The automatic method of delineating the hydrographic network and sub-watersheds at the plain level is imprecise. Following the discovery of numerous errors following the execution of the process of automatic generation of the hydrographic network as well as the automatic delimitation of the sub-watershed areas, the decision was made to predefine manually the hydrographic network and the sub-catchment areas of the study zone. This error was primarily due to the weak variation of the slope at the level of the plain of Mateur, so the decision was made to predefine manually the hydrographic network and the sub-catchment areas of the study zone. The “Pre-defined streams and watersheds” function of the SWAT model allow users to import pre-defined watershed boundaries and streams (Luo et al. 2011). Each defined river has to fill in the following eight fields (Table 1): "FID," "shape," "ARCID," "GRID_CODE," "FROM_NODE," "TO_NODE," "Subbasin," and "SubbasinR," while the defined watersheds have to fill in just four fields, which are respectively (Table 2): "FID," "shape," "GRID_CODE," and "subbasin". The streams and sub-basins datasets were imported using the "Predefined streams and watersheds" function after they were defined. The uncontrolled part of the Ichkeul watershed was divided into 13 sub-basins (Fig. 5 and Tables 1 and 2).

Table 1 Attribute table of the predefined hydrographic network of the uncontrolled part of the Ichkeul watershed
Full size table
Table 2 Attribute table of the predefined sub-basins of the uncontrolled part of the Ichkeul watershed
Full size table
Fig. 5
figure 5

Summary map of the delimitation steps of the studied territories (from the DEM to the definition of the sub-watersheds)

Full size image

HRUs

In SWAT, a watershed is divided into multiple sub-watersheds, which are then further subdivided into hydrologic response units (HRUs) that consist of homogeneous land use, management, topographical, and soil characteristics (Arnold et al. 2012). The objective of the HRU definition was to reduce the heterogeneities due to climate, soil types, topography, and geology that influence hydrologic response (Sisay et al. 2017). With the combinations and the reclassifications of slope class, land use, and soil data, a total of 206 HRUs were defined. The hydrologic cycle is climate driven and provides moisture and energy inputs, such as daily precipitation, maximum/minimum air temperature, solar radiation, wind speed, and relative humidity, that control the water balance (Arnold et al. 2012).

Model execution

The hydrologic cycle is climate driven and provides moisture and energy inputs, such as daily precipitation, maximum/minimum air temperature, solar radiation, wind speed, and relative humidity, that control the water balance (Arnold et al. 2012). SWAT reads those observed data directly from the input files and generates this data at runtime to start the simulation. The climate data was prepared for the period 2013–2019. The SWAT model was run to simulate the different hydrological components which are canopy storage, surface runoff, infiltration, evapotranspiration, lateral flow, tile drainage, water redistribution in the soil profile, return flow, and infiltration recharge from surface water bodies, ponds, and tributary channels.

Sensitivity analysis, calibration, and validation of the SWAT model

The first step in the calibration and validation process in SWAT is the determination of the most sensitive parameters for a given watershed or subwatershed (Arnold et al. 2012). It is necessary to identify key parameters and the parameter precision required for calibration (Ma et al. 2000). The user determines which variables to adjust based on sensitivity analysis. Sensitivity analysis is the process of determining the rate of change in model output with respect to changes in model inputs (parameters) (Arnold et al. 2012). The second step is calibration. Calibration is an effort to better parameterize a model to a given set of local conditions, thereby reducing the prediction uncertainty (Arnold et al. 2012). Model calibration is performed by carefully selecting the values of the model input parameters already identified during the sensitivity step and comparing the observed data with the model predictions (in this case the outflow). The final step is validation. Model validation is the process of demonstrating that a given site-specific model is capable of making sufficiently accurate simulations, although “sufficiently accurate” can vary based on project goals (Refsgaard 1997).

The calibration of the model can be done in two different ways. Calibration can be accomplished manually or using autocalibration tools in SWAT (Van Griensven and Bauwens 2003; Van Liew et al. 2005) or SWAT-CUP (Abbaspour 2007). The present study is based on the autocalibration method via the SWAT-CUP software.

SWAT-CUP links five algorithms, used for sensitivity analysis, calibration, and validation of the SWAT model. These algorithms are the Sequential Uncertainty Fitting Algorithm (SUFI-2), the Particle Swarm Optimization (PSO), the Markov Chain Monte Carlo (MCMC), the Generalized Likelihood Uncertainty Estimation (GLUE), and finally the Parameter Solution (Parasol). The SUFI-2 algorithm has been used to perform the calibration and validation sensitivity analysis processes.

It is recommended to evaluate the performance of a model graphically and by statistical criteria (Moriasi et al. 2007). Krause et al. (2005) recommend a combination of different effectiveness criteria for scientifically sound model calibration and validation. In the present study, the coefficient of determination (R2), the Nash and Sutcliffe model efficiency coefficient (NSE) (Nash and Sutcliffe 1970), and the percentage bias (PBAIS) are the statistical criteria that were used to evaluate the performance of the SWAT model. NSE, PBIAS, and R2 are calculated by Eqs. 5, 6, and 7, respectively:

$$\mathbf{N}\mathbf{S}\mathbf{E}=1-\frac{\sum_{{\varvec{i}}=1}^{{\varvec{n}}}{(\mathbf{O}\mathbf{i}-\mathbf{P}\mathbf{i})}^{2}}{\sum_{{\varvec{i}}=1}^{{\varvec{n}}}{(\mathbf{O}\mathbf{i}-\mathbf{O}\mathbf{a}\mathbf{v}\mathbf{r})}^{2}}$$
(5)
$$\mathbf{P}\mathbf{B}\mathbf{I}\mathbf{A}\mathbf{S}=\frac{\sum_{{\varvec{i}}=1}^{{\varvec{n}}}\left(\mathbf{O}\mathbf{i}-\mathbf{P}\mathbf{i}\right)*100}{\sum_{{\varvec{i}}=1}^{{\varvec{n}}}\mathbf{O}\mathbf{i}}$$
(7)
$${{\varvec{R}}}^{2}={(\frac{\sum_{{\varvec{i}}=1}^{{\varvec{n}}}\left(\mathbf{O}\mathbf{i}-\mathbf{O}\mathbf{a}\mathbf{v}\mathbf{r}\right)\left(\mathbf{P}\mathbf{i}-\mathbf{P}\mathbf{a}\mathbf{v}\mathbf{r}\right)}{\sum_{\mathbf{i}=1}^{{\varvec{n}}}{\left(\mathbf{O}\mathbf{i}-\mathbf{O}\mathbf{a}\mathbf{v}\mathbf{r}\right)}^{2}\sum_{{\varvec{i}}=1}^{{\varvec{n}}}{\left(\mathbf{P}\mathbf{i}-\mathbf{P}\mathbf{a}\mathbf{v}\mathbf{r}\right)}^{2}})}^{2}$$
(8)

In addition to the evaluation criteria listed above, Abbaspour (2007) suggest the use of two additional measures, known as the P-factor and R-factor. The P-factor is the percentage of the measured data bracketed by the 95PPU and the R-factor, is a measure of the quality of the calibration and indicates the thickness of the 95PPU. The combination of the P-factor and R-factor together indicates the strength of the model calibration and uncertainty assessment, as these are intimately linked (Arnold et al. 2012).

Daily outflow data from January 2013 to December 2017 were used during the calibration period and the remaining data from January 2018 to December 2019 were used to validate the model performance at the only gauging station (Ecluse-Sidi Hassoun). Data from the first 2 years (January 2013 to December 2015) were used as the "warm-up" period.

Results and discussion

The SWAT model was calibrated to a daily time scale. Data are available for the observation period (2015–2019), and they have been divided into calibration data (2015–2017) and validation data (2018–2019) (Fig.6 and 7). Uncertainty analysis is performed using the sequential uncertainty fitting (SUFI-2) algorithm (Abbaspour 2005).

Fig. 6
figure 6

Comparison of observed and simulated daily flows at the Ecluse-Sidi Hassoun station after the Calibration phase

Full size image
Fig. 7
figure 7

Comparison of observed and simulated daily flows at the Ecluse-Sidi Hassoun station after the validation phase

Full size image

Specification and parameter determination are two major steps in calibration (Sorooshain and Gupta 1995). Understanding the studied system's hydrological processes enables accurate parameterization of a model, which is a crucial step in its calibration. Correct parameterization can therefore result in model calibration that is quicker, more precise and has lower prediction uncertainty. Parameter sensitivities are determined by performing a multiple regression analysis, which regresses the parameters generated by Latin Hypercube against the objective function (Abbaspour 2007). To determine the relative significance of each parameter, a t-test and p-value are used. The most sensitive parameters in the calibration step are presented at the top of the rankings, that is, the highest value of the t-stat index module which represents the ratio of the parameter coefficient by the standard error; and the lower value of the "p-value" which is related to the rejection of the hypothesis that an addition in the value of the parameter provides a significant increase in the variable response (Abbaspour 2007). Six out of a total of nineteen parameters, in this study, were found to be sensitive. The parameters used in the sensitivity analysis are listed in Table 3, along with the sensitivity of each parameter. The sensitive parameters are Moist bulk density (SOL_BD), Saturated hydraulic conductivity (SOL_K), Initial SCS runoff curve number for moisture condition II (CN2), Base flow alpha factor (ALPHA_BF), Groundwater delay (GW_DELAY), and Effective hydraulic conductivity in the tributary channel (CH_K1).

Table 3 Result of the sensitivity analysis of the parameters chosen for the calibration phase
Full size table

Numerous methods have been used to calibrate and validate SWAT. Most published SWAT applications report both graphical and statistical hydrologic calibration results, especially for streamflow, and hydrologic validation results are also reported for a large percentage of the studies (Arnold et al. 2012). According to Coffey et al. (2004), there are close to 20 statistical tests that can be used to evaluate the predictions of SWAT. R2, NSE, and PBIAS are the statistics that are most frequently used for calibration and validation. The R2 obtained value for the calibration period (2015–2017) is 0.88, while the Nash–Sutcliffe efficiency (NSE) for the same period is 0.87. The percentage bias value (PBIAS) is − 6.7, as shown in Table 4. The outcome of the validation process (2018–2019) is an R2 of 0.94, an NSE value of 0.93, and a PBIAS of 6.8% for the uncontrolled portion of the Ichkeul watershed.

Table 4 The values of the statistical evaluators of the simulation after the two phases of calibration and validation
Full size table

The R2 statistic can range from 0 to 1, where 0 indicates no correlation and 1 represents perfect correlation, and it provides an estimate of how well the variance of observed values is replicated by the model predictions (Krause et al. 2005). A perfect fit between the simulated and observed data is indicated by an NSE value of 1 (Arnold et al. 2012). According to Gupta et al. (1999), the optimal value of PBIAS is 0. This means that smaller values of PBIAS indicate an accurate simulation of the model. This holds for the outcomes; R2, NSE, and PBIAS are very close to the optimal limits defined above by the authors. The scatter plot (Figs. 8a, b, and 9) shows the strong correlation between the variables the model predicted and observed during the calibration and validation phases. These findings demonstrate the efficacy of the outflows predicted by SWAT. Studies that have been published by Santhi et al. (2001); Moriasi et al. (2007) and Van Liew et al. (2003) confirm the findings. When the Nash-Sutcliff efficiency (NSE) and coefficients of determination (R2) are greater than 0.5 and the bias percentage is under 15%, calibration and validation are deemed to have been successful.

Fig.8
figure 8

Comparison of the simulated-observed a and calibrated-observed b daily flow at Ecluse-Sidi Hassoun station by linear regression before and after the calibration phase

Full size image
Fig. 9
figure 9

Comparison of simulated and observed daily flow at the Ecluse-Sidi Hassoun station by linear regression after the validation phase

Full size image

The R-factor and the P-factor should be measured, according to Abbaspour (2007), as they validate the accuracy of the uncertainty assessment and model calibration. A value of the P-factor close to 1 means that all observations are included in the prediction uncertainty. The R-factor should, in general, be less than 1.5 (Abbaspour 2007). After the adjustment phase, the values of the P and R factors for the calibration and validation period are (0.97, 1.17) and (0.95, 1.11), respectively (Table 4); these obtained values are within the limits of the P and R factors stated above.

Based on the statistical evaluations presented in Table 4, there appears to be a strong correlation between observed and simulated daily flows. Based on the results obtained, SWAT is capable of reproducing the observed daily outflow at the outlet of the study basin with a high level of accuracy.

Conclusion

To determine the need for the freshwater of Ichkeul lake, and then later, to model its salinity and quantify the sediments to establish guarantees for its survival, the SWAT model was applied. The model was then calibrated and validated on a daily basis by applying the SUFI-2 algorithm in SWATCUP-2012 using historical stream flow recorded at the Ecluse-Sidi Hassoun station, located downstream of the Ichkeul watershed. 7 years (2013–2019) of datasets have been arranged for model setup, first 2 years (2013–2014) of data were kept as the warm-up period for the model. A sensitivity analysis is performed via SWAT-CUP to identify the most sensitive values and ranges of various hydrological parameters for the study area. In this study, six out of a total of nineteen parameters were found to be sensitive. Saturated hydraulic conductivity (SOL K), Initial SCS runoff curve number for moisture condition II (CN2), Base flow alpha factor (ALPHA BF), Groundwater delay (GW DELAY), and Effective hydraulic conductivity in the tributary channel (CH K1) are the sensitive parameters. Five statistical evaluators, Nash-Sutcliff efficiency (NSE), coefficient of determination (R2), bias percentage (PBIAS), P-factor, and R-factor, were used to assess the performance of the models. During the calibration and validation periods, the R2, NSE, and PBIAS values were 0.88, 0.94, 0.87, 0.93, -6.7, and 6.8, respectively. The P-factor and R-factor values during the calibration and validation period are 0.97, 0.95, -6.7, and 6.8, respectively. The result shows SWAT holds promise for use in the uncontrolled part of the Ichkeul watershed. It can be concluded that the outcome of the model indicated that there had been a good agreement between the simulated and observed flow. Therefore, the model is capable of reproducing the observed daily outflow at the outlet of the study basin with a high level of accuracy.