1 Introduction

Natural hazards threaten permanently the people throughout the world. The quantization of brutal events such as earthquakes, mass wasting, and floods requires the estimation of the hazard/susceptibility level of the specific risk and the consideration of the vulnerability degrees, (Hadji et al. 2013, 2014a2016, 2017; Mekki et al. 2017; Dahoua et al. 2017a, b, 2018; Manchar et al. 2018; Zeqiri et al. 2019; Anis et al. 2019; Fredj et al. 2020). However, for long-term perils, such as shrinkage-swelling and soil erosion, the concept is quite different (Hadji et al. 2014b). The environmental issues related to water have been dealt with in countless works, dealing with a lot of aspects (Rouabhia et al. 2012; Hamed et al. 2014, 2021; Mokadem et al. 2016; Besser et al. 2018, 2021; Ncibi et al. 2020a, b, 2021; Benmarce et al. 2021; Bensoltane et al. 2021; Hamad et al. 2021a, b).

Water-borne erosion is the worst form of land degradation in the MENA countries. It causes severe socio-economic and environmental problems (Fig. 1), and leads to soil loss, water-quality degradation, and dam siltation. The rate of soil erosion is controlled by different parameters of the basin such as geology, geomorphology, climate as well as hydrological dynamics. Global projects as the United Nations Environment Program (UNEP) or the United Nation Convention Desertification Combat (UNCDC) have dealt with this problem. According to the Global Assessment of Human-induced Soil Degradation project (GLASOD), about 15% of the land surface has already degraded (Oldemanet et al. 1991). The hardest rates of soil erosion were reported in Europe (25%), Asia (18%), Africa (16%); and less in North America (5%). Algeria is one of most affected countries by the phenomenon in North Africa (Hamed et al. 2018), the western part is particularly eroded, with 47% of losses in fertile lands; followed by the middle (27%) then the East (26%). The specific erosion reaches 4000 t/ km2 year, and the siltation rate exceed 15% (Bougerra et al. 2017).

Fig. 1
figure 1

Photos of some disasters caused by the effect of water erosion

Full size image

The evaluation models of soil erosion make available the quantification of the process under various conditions (Aiello et al. 2015). By consulting the geo-scientific literature; we noticed different studies that rigorously discussed the spatio-temporal interactions between soil erosion and controlling factors. They vary from empirical, physical to conceptual models (Viney and Sivalapan 1999). Empirical-based models are the simplest and the widely applied (Merritt et al. 2003). The Universal Soil Loss Equation (USLE) method was initially developed by Wischmeier and Smith (1965), Modified (MUSLE) by Williams and Berndt in (1977), then Revised (RUSLE) by Renard (1997). The latest RUSLE version has proven a great ability to estimate the soil loss (Alexakis, et al. 2013; Xiao et al. 2015).

Because of their high cost, long time-consuming and large study areas; the conventional methods to assess water-born erosion no longer fit with the current natural hazard mapping needs. The advent of RS data and the GIS environments gave the necessary support to these methods (Ganasri and Ramesh 2016; Nezhadafzali et al. 2019; Mohammed et al. 2020). The RS/GIS geospatial algorithms allow the prediction of the erosion potential in large expanses on a cell-by-cell basis (Tamene et al. 2017; Thapa, 2020). They have the ability to integrate complex factors such as soil, climate, vegetation, terrain, and human practice (Teng et al. 2018). The Meskiana catchment is characterized by an acute orography, a semi-arid climate, rainfall-stormy precipitation, torrential/intermittent runoff and silty-clay surface soil textures. These factors favoured the intense susceptibility to erosion. We have chosen to apply RUSLE model in our study because the method proved in many close contexts to be a simple and practical in the investigation of the soil erosion process.

The main purpose of the present article consists in the implementation of an environmental info-system based on the RUSLE model that considers all the causative factors of water-born erosion to assess of the soil-loss susceptibility in the Meskiana catchment. The method passes through: (i) The geo-processing of soil loss controlling parameters in the study area; (ii) The implementation of an approach that combines RS data and GIS to estimate the spatial distribution of water born-erosion and soil loss through RUSLE model; (iii) The study of geochemical and mineralogical terrain features; and (iv) the identification of the most affected area.

The outcomes of the study could help in the establishment of effective management scenarios and provide skills to decision-makers for mitigate the soil erosion threats.

2 General Settings

The Meskiana catchment is a large watershed situated between the 07°11ʹ 48ʺ to 07° 54ʹ 12ʺ longitude, 35° 09ʹ 35ʺ to 35° 48ʹ 06ʺ latitude and covers about 1680 km2 of land near the Algerian-Tunisian international border (Fig. 2a), (Tamani et al. 2019). The study region have an semi-arid climate characterized by a hot dry summer and a temperate cold winter with an average annual rainfall of 400 mm and a temperature of 16.3 °C (Hamad et al. 2021a, b). The average annual evaporation fluctuates between 1300 and 1500 mm (Rais et al. 2017; Hamad et al. 2018a, b). The land use of the catchment shows barren lands, cereals, rangelands forests, and olive groves. The Halfa steppes are scarce in the basin, indicating poor protection against water erosion. The morphology of the basin is characterized by hills alternating with valleys and steep mountains (Boulemia et al. 2021). Geologically, the formations are made up of Triasic facies, emerging in diapirs; Cretaceous stacking of Aptian and Maastrichtian age formed by marl and limestone, and of Miocene sandstone limestone formation; covered in the middle of the basin by a Mio-Plio-Quaternary continental filling (Mouici et al. 2017). The outcrops represent 56% of marl, 26.5% of alluvium, 3% of limestone, and 3% of Triassic (Fig. 2b, c) (Nekkoub et al. 2020; Brahmi et al. 2021). The soil of the watershed exhibits a swelling character with sensitivity to erosion (Table 1). This soil is a clay loam (Table 2). The diffractometric analysis and fluorescence spectra of clay minerals identify the types of clay minerals in samples collected from the main outcrops of the study area (Fig. 3). The fluorescence showed a consequent rate of Alumina. This geochemical class favors the development of the phenomenon at the microscopic scale (Table 3). The fluctuation of the piezometric level depends closely on the contributions of precipitation (Demdoum et al. 2015). The watershed is drained by the wadi Meskiana; the main collector of runoff water. On the right bank, spills wadi Rbâa, wadi Hmadjra, wadi Smar el Houd and Wadi El Mallah. Whereas, on the left bank, spills wadi Jdida and wadi Ain Sedjra. The study area shows a hilly terrain with high to moderate slope ranging from 49.56% in the mountains ridges to sub-flat in the river basins, (Fig. 2d).

Fig. 2
figure 2

a Geographic location of the study area, b Simplified lithofacies map of the study area; c geologic cross-section in Meskiana catchment, d = slope map of the study area. [Legend of b: 1: Quaternary, 2: Plio-quaternary, 3: Middle Miocene; 4: Lower Miocene, 5: Ypresian and Lutecian, 6: Thanetian, 7: Upper Maestrichtian, 8: Campanian, 9: Santonian, 10: Coniacian, 11: Upper Turonian, 12: Lower Turonian, 13: Upper Cenomanian, 14: Aptian, 15: Cretaceous, 16: Triassic

Full size image
Table 1 Argillosity and organic matter content of the analysed soil samples
Full size table
Table 2 Soil goniometric size (LCPC standards)
Full size table
Fig. 3
figure 3

XRD of the samples from the main outcrops of the study area

Full size image
Table 3 Results of XRF
Full size table

3 Materials and Methods

In order to rank the most susceptible areas to soil erosion, various thematic layers were rasterized and integrated into GIS software. The modelling involved a weighted overlay of the five factor maps. The higher weight is given to the feature with the highest influential factor and the lower to the least influential one. The RUSLE model combines five factors: rainfall (R), topography (LS), soil erodibility (K), cover management (C), and support practice (P) (Wischmeier and Smith 1978). The model can be mathematically expressed by the Eq. (1)

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

where A is the annual rate of soil loss (t/ha/yr); R is the rainfall erosivity (MJ mm*m−2 h−1); K is the soil erodibility factor (t*ha J−1*mm −1); LS are slope length and slope steepness factor (dimensionless); C is the cover and management factor (dimensionless in %); P is the support practice factor (dimensionless in %).

The flow chart of the present study can be used as a practical method for the estimation of the soil loss (Fig. 4).

Fig. 4
figure 4

The adopted methodological Flow chart for soil erosion assessment

Full size image

3.1 Data Used

The study area watershed has been bounded in ArcGIS 10.3 software under the hydrology/spatial analyst tools. The prepared base map was then used for the extraction of study area from the satellite image and Shuttle Radar Topography Mission digital elevation model (SRTM-DEM). For remote sensing; we used Sentinel 2-A satellite data earned on January 1st 2018 for generating all the needed maps layers. The designation of the geology has relied on the Oum El Bouaghi, Ain Beida, El Aouinet, Graet El Tarf, Fkirina, Meskiana, Khenchela, Dalâa, Hammamet, Jebel Djehfa, and Zoui 1:50,000 scale, geologic maps in the same GIS environment. Twelve weather stations located at the within the study area or in its vicinity, and rainfall data for 20 years (2000–2020) was collected from the (ONM National Meteorological office) database. Extensive field survey data and field images were used for calibration, processing, and result validation.

3.2 Predictive Factors

3.2.1 Erosivity

Estimating the erosivity factor "R" according to the Wischmeier and Smith (1978) formula requires knowledge of the kinetic energies (ec) and the 30-min average intensity (i30) of the raindrops in each downpour. "R" is given by the Eq. (2)

$$ {\text{R}} = {\text{k}}.{\text{ec}}.i_{30} $$
(2)

With "k" is a coefficient depending on the system of measurement units.

However, the precipitation data available in the stations of our study area are only monthly or inter-annual averages.

Authors such as Arnoldus (1980) have developed alternative formulas that involve monthly and annual precipitation to determine "R" factor. The formula of Rango and Arnoldus (1987) was already successfully applied to twelve sites in the greater Maghreb; hence the formula corresponds well to our study area (Eq. 3):

$$ {\text{log}}\;{\text{R}} = 1.74.{\text{log}}\sum {\left( {{\raise0.7ex\hbox{${P_{i}^{2} }$} \!\mathord{\left/ {\vphantom {{P_{i}^{2} } P}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$P$}}} \right)} + 1.29 $$

With: "R" rainfall erosivity factor expressed in t/ha.

Where Pi is monthly precipitation and P is annual precipitation in mm; calculated using a large number of rainfall stations, subsequently the results were interpolated over the rest of the basin. The calculation of the erosivity factor r is applied over a series of twenty (20) years of precipitation in and around the watershed.

3.2.2 Erodibility

The adjustment of "K" factor for coarse elements was based on the soil map of the Meskiana watershed; several samples were analysed to identify the characteristics of each type of soil. Thus, the relation was used for the estimation of this factor calculated by the Nomogram equation, which makes a combination between the size of the particles, % of the organic matter, the structure of the soil and the class of the permeability of the profile according to the Eq. (4):

$$ {\text{K }} = \, [\left( {{2}.{1 }\left( {{1}0^{{ - {4}}} } \right)^{{}} {\text{m 1}}.{14} \left( {{12} - {\text{om}}} \right)} \right) \, + \left( {\left( {{3}.{25 }\left( {{\text{s}} - {2}} \right)} \right) \, + \left( {{2}.{5 }\left( {{\text{p}} - {3}} \right)} \right)} \right] \, /{ 1}00 $$
(4)

with: K: soil erodibility determined on an annual basis expressed in t.h/ha. Mj.mm; om: organic matter content in %; m: textural term = (silt + fine sand) % * (100—clay%); s: soil structure code (1 to 4), 1 for a very fine grained structure and 4 for a massive structure; p: permeability code (1 to 6), ranging from 1 for fast-draining soils to 6 for very slow-draining soils.

The erodability class is ranked as low when K is between 0.1 and 0.2; as moderate if K is from 0.2 to 0.30 and as high if K is between 0.30 and 0.40.

3.2.3 Topography

"LS" is a topographic factor combining the slope (s in %) and the slope length (l in m) (Wischmeier and Smith 1965). LS is calculated from the slope inclination and slope length, obtained from the geo-processing of DEM in GIS software using Moore and Burch's formula (Eq. 5):

$$ {\text{LS}} = { 1}.{6}*{\text{pow}}\left( {\left[ {{\text{flowacc}}} \right]*{\text{resolution}}/{22}.{1},^{{0.{6}}} } \right) \, *{\text{pow}}\left( {{\text{sin}}\left( {\left[ {{\text{slope}}} \right]*0.0{1745}} \right)/0.0{9},^{{{1}.{3}}} } \right)) $$
(5)

The classes of LS factor are ranked as follows: If slope = [0, 7.0%] so LS = 0.55. If s = [7, 11.3] so LS = 0.6, If s = [11.3, 17.6] so LS = 0.8; If s = [17.6, 26.8] so LS = 0.95 and if s is > 26.8 so LS = 1.

3.2.4 Cover Management

RS algorithms have the ability to generate the land use-land cover of the study area. "C" factor represents the effect of vegetation cover. Based on the normalized vegetation index NDVI derived from Sentinel 2 A imagery. The index transforms the reflectance of vegetation into a percentage (Karaburun 2010). The relationship between the NDVI and the "C" parameter is determined by the Eq. (6), (Benselama et al. 2018)

$$ {\text{C}} = \, 0.{9167 }{-}{\text{ NDVI}}*{1}.{1667} $$
(6)

3.2.5 Support Practice

The factor "P" integrates anti-erosion cultivation techniques, such as contour crops, alternating strips or terraces, reforestation in benches, and ridging. The values of the anti-erosion practice factor of human origin are extracted according to the sloping cultivation methods cited by (shin, 1999). According to Gitas et al. (2009), the P factor is calculated as the ratio of the rate and amount of soil loss due to a specific support practice to the soil loss due to row farming in upward and downward of the slope condition. The "P" values rages between 0 and 1, in which the highest value is given to the areas where there is absence of any conservation practices such as grasslands and open areas, and the minimum values are assigned to plantation area with contour cropping and built-up land.

4 Results and Discussions

4.1 Parameter Estimation

4.1.1 Assessment of Soil Erodibility

The erodibility of soils represents the susceptibility of soil to erosion reflected by the rate of soil loss under a standard rainfall condition estimated on the basis of soil textures (Cassol et al. 2018). Nomograph is used to determine the "K" factor for a soil, based on texture (m); % organic matter (om), soil structure (s), and permeability (p) according to Wischmeier and Smith (1978) model; who incorporates a corrective factor to take into account coarse diameters. The model is well suited to the soil nature of arid and semi-arid environments in the MENA region." K" factor is determined according to the Eq. (4):

The calculated values of the erodibility index "K" vary between 0.14 and 0.34 (Fig. 5a). They are divided into four classes of soils. The calculated "K" erodibility values for each type of soil show that almost all the soils (93.6%) have moderate erodibility (0.20–0.30). Vertisoils have low erodibility (0.10–0.20), they are the least represented in the region (2%). Some poorly evolved soil exhibits high erodibility rates (0.30–0.40), occupying 3.9% in the Southwest of the study area. The comparison of the pedological map with the topography shows that the little developed soils on steep slopes, bordering the valleys where soil degradation is very intense.

Fig. 5
figure 5

Spatial repartition of: a = K Factor, b = R factor, c = NDVI, d = C factor, e = P factor and f = LS factor

Full size image

4.1.2 Assessment of Rain Erosivity

The rainfall erosivity was calculated according to the modified Fournier equation (Arnoldus 1980) using the monthly rainfall of agricultural years, (September to August), over the period 2000–2020, from 12 pluviometric stations located at the within the study area and in its vicinity. We noted that the used rainfall series corresponds to a relatively dry period. The average value of the rainfall erosivity over 2000/2020 period was obtained by performing the monthly and annual average precipitation calculated for each year. The distribution of the different classes of rainfall erosivity "R" in our study basin is shown in Fig. 5b.

4.1.3 Vegetation Cover and Cultural Practices Factor

The "C" factor combines vegetation cover, the degree of plant production, and the associated cultivation technique. The factor varies between 1 for bare soil to 0.1 for dense forests. The spatial distribution of vegetation covers and cultural practices shows that areas most vulnerable to erosion are the "bare land" land-use type. This class covers almost the entire upstream of the catchment (Fig. 5c). According to the "C" factor map (Fig. 5d); the study area has been classified into five major soil classes; the lowest value (0.0007) is on land planted with agriculture and arboriculture, however, the coefficients of 0.8 to 0.9 are found in areas covered by sparsely dense and spaced forests. The highest coefficients (close to 1) correspond to bare land.

4.1.4 Anti-Erosion Practice

The "P" factor determines the ratio of soil loss to specific practices on agricultural land, especially with parallel plowing against slopes (Wischmeier and Smith 1978). Note that in our study the low and medium values of "P" are assigned to areas with low to moderate slopes. The factor varies from 0.55 to 0.6 for areas with a low slope, and from 0.8 to 1 for areas with steep slopes (Fig. 5e).

4.1.5 Topographic Factor (LS)

The length and inclination of the slope influence the erosion process. We used a 30 m SRTM DEM for the slope, exposure, slope length, and LS calculation in Arcgis 10.3 software. The stretched values of the LS were grouped into five classes. 60% of the catchment area belongs to very high areas (Fig. 5f). It is concentrated in the upper valley and in intrusive areas of the Wadi Meskiana tributaries in the southern part of the basin. The lowest values of the LS index (which vary from 0 to 5) occupy only 7% of the surface of the basin.

4.2 Estimation of the Erosion Rate

An erosion rate of soil loss results from the combination of the the erosivity of precipitation, the erodibility of soils, the combined effect of the degree and the length of the slope, the vegetation cover, and anti-erosion practices in the RUSLE model. The five factor maps were multiplied with Raster Calculator/Map Algebra tools, to calculate the annual rate of soil loss (t/ha/yr); in the study area on a pixel-by-pixel basis scale (Fig. 6).

Fig. 6
figure 6

The spatial distribution of the potential of erosion in the study area

Full size image

The weighted average loss per unit area is 61 t/ha/year or a total of 742,674 t/year for the entire basin. The losses vary between 34.65 t/ha/year as the minimum value measured in the matorrals and 129.12 t/ha/year as the maximum value estimated at the level of the rangelands. Average losses are estimated in the plain at 46.5 t/ha/year on the most resistant soils where the forest cover is fairly dense. The land cultivated by seasonal crops and arboriculture shows high susceptibility to erosion with annual losses of 57.61 t/ha and 64.36 t/ha/year respectively. The map of average annual soil losses shows a clear fragility of the eastern and north-eastern part of the basin on marl and Miocene sandstone marls where values exceed 35 t/ha/year. These cover approximately 3250 ha or 27% of the total area of the catchment. The weak losses (< 7 t/ha/year) dominate the slopes of the right bank of the basin where a scrub of holm oak and reforestation cover the sandstone grounds.

Several studies have proved the improvements made by the adoption of the GIS-RS technology to the RUSLE method. Perhaps the most important is the estimation and mapping of "R" factor into a raster surface using the IDW Interpolation technique (Panditharathne et al. 2019). In addition, the combination of GIS/RS enabled the derivation of "LS" factor with better accuracy at the large catchment scale, and in complex landscapes (Wilson and Gallant 2000). Without the use of GIS-RS, obviously, the results will be less precise with low spatial resolution. Although, we recommend the approval of RUSLE based GIS/RS model as an effective tool for soil erosion hazard mapping at the catchment scale. The main limitation of the technique implicate the inability to validate the model with accurate tools such as (Area Under Curve (AUC) of the receiver operating characteristic (ROC) curve) and the deficiency to compare between the calculated and the standardized soil loss values (Somasiri et al. 2021).

5 Conclusions

Soil erosion has substantial impact on natural and physical environment in North-eastern of Algeria semiarid zone. The present study proved that GIS and RS techniques are powerful tools for assessing waterborne erosion and for the estimation of soil loss in the Meskiana catchment. The adaptation of the empirical models of Wischmeier and Smith in the revised versions RUSLE model made possible to quantify the specific degradation due to the phenomenon. The process resulted from the combined effect of several natural and anthropogenic factors such as rainfall, soil texture, topography, land use, and human practices. The soil erosion susceptibility map was assessed by combining layers' maps of those factors. The GIS analysis has been carried out to estimate the annual soil loss on a pixel-by-pixel basis and the spatial distribution of the waterborne erosion area in the Meskiana catchment. The results proved the critical state of the soil in the catchment and how much human practices have deteriorated the natural environment in the study zone. We assessed the average loss at 61 t/ha/year with a total weight of 742,674 t/year for the entire basin. The losses vary from a minimum in the matorrals to a maximum value in rangelands. The land planted by cereal cultivation and seasonal crops are the most susceptible. Highest rates of eroded areas were recorded in the eastern and north-eastern part of the basin on marl outcrops. Whereas, the lowest rates were recorded in the right bank of the basin where scrub of holm oak and trees covers the sandstone outcrops. These values not much different from those recorded in similar environments in NE Algeria. The comparison of mathematical methods with deterministic ones makes possible to identify the mineralogy and geochemistry of the soils most vulnerable to erosion. The adopted method can be applied in the entire Medjerda-Mellegue transboundary basin for the erosive potential estimation. It would help to take suitable erosion control measures in the severely affected areas; and could assists decision makers to implement mitigation measures in North East of Algeria.