Introduction

All life on earth depends on a thin layer of soil. It is a very weakly renewable resource where its degradation can be rapid, and it takes several thousand years to form and regenerate. It is very vulnerable by degradation and erosion. Erosion is the set of chemical, physical, and organic actions which lead to the alteration and weathering of the rocks and the gradual leveling of the soil. It is essentially the superficial wear of the earth’s crust. Water and wind are the main agents (Gréco 1966).

Water erosion develops when rainwater, which can no longer infiltrate the soil, trickles down onto the plot, sweeping away soil particles. This refusal of the soil to absorb excess water appears either when the intensity of the rains is greater than the infiltrability of the soil surface (Hortonian runoff) or when the ground is waterlogged (Le Bissonnais et al. 2002).

Water erosion causes stripping of soil profiles, sediment transfer but also loss of water, and carbon and nutrients for crops and accelerates the degradation of the fertility of cultivated soils. The soil torn materials are deposited in the hydrological networks leading to pollution and siltation which cause the reduction of the volumes of the water reserves. Soil degradation by water is highly responsive in semiarid and dry subhumid areas, while wind erosion dominates in arid areas. In the past 70 years, Africa has lost about 5 million square kilometers of its fertile land, affecting 65% of the agricultural land in the region. Land degradation affects some 300 million hectares in Latin America. In Europe, 157 million hectares suffer from water and wind erosion. For Mexico, the USA, and Canada, the area of land affected by degradation is close to 95 million hectares (Ged and Ifad 2002). The Mediterranean region is among the regions most affected by the degradation of agricultural land; according to FAO studies (FAO 1990) in the southern shore of the Mediterranean, precisely in the North African countries, between 40 and 45% of the land are affected by water erosion as in the case of Morocco and Tunisia, and in the northern Mediterranean as in Greece and Turkey, lands affected by water erosion are 35 and 50%, respectively.

Much research has focused on this phenomenon of water erosion around the world. In Tunisia, soil scientists are beginning to study the distribution of soils and erosion events in Matmata watersheds (Mathieu 1949; Fournet 1969) which will be extended to other basins by many study (Delhumeau 1981; M'timet and Escadafal 1982: Delhoume 1987), while agronomists link farming patterns and erosion (Prost 1954; Cormary and Masson 1964; Poncet 1962, 1970) and hydrologists and geographers analyze the effectiveness of jessour, especially during the exceptional rains of 1969 (Tixeron 1961; Bonvallot 1979; Chabahni 1984).

This phenomenon is very well known in Tunisia. Indeed, degradation of the country’s soils is favored by the soil typology, the topography, the contrasting Mediterranean climate, and the old use of agricultural land. Consequently, the problem of water erosion and its consequences remains posed. Indeed, about 3 million hectares of land are really affected by this phenomenon, half of which are seriously affected by medium to high erosion (Achouri 1995).

The North West of Tunisia is characterized by the aggressivity and the diversity of erosive processes. In order to determine the level of erosion, several methods are adopted, of which the most used are the universal soil loss equation (Wischmeier and Smith 1978) and its improved version (Renard et al. 1997). The choice of the model depends on the variability of basin area, availability of point data, and precipitation data (Bonn 1998).

In recent years, methods for calculating soil loss have been associated with GIS and remote sensing. It is true that with these modern methods, we are gradually discovering the interest of the spatialization of soil loss calculation methods, mapping to represent the results of soil loss over large areas and the creation of scenarios for human interventions on the environment (Bonn 1998). Modeling can be a tool for decision support and conservation of soil resources, by establishing erosion loss quantification maps, risk identification sensitive areas, and the presentation of priority intervention zones.

In this study, the objective is to determine the amount of water erosion in the Ayda river watershed (Kisra-Siliana); the adopted methodology was based on the revised universal soil loss equation (RUSLE) model (Renard et al. 1997).

This RUSLE model is one of the best models in this field, characterized by easy use and satisfactory efficiency. The choice of this model depends on the variability of environmental and climatic conditions, the size of the catchment area, and the degree of inclination of the slopes (Bonn 1998).

The results of the model are an estimate of losses in eroded soils and a map of the risk of erosion, based on key factors, namely, soil erodibility, slope, rainfall erosivity, vegetation, and soil conservation practices. All the data entered into the geographic information system (GIS) enabled us to obtain the site erosion risk map.

In most studies, the RUSLE model has yielded reliable watershed-scale results around the world, particularly in African countries (Bonn 1998; Anys et al. 1994; Smith 1999; Mati and Veihe 2001; Angima et al. 2003; Chen et al. 2008; EL-Garouani et al. 2008; Bouhadeb et al. 2018).

Materials and methods

Study site

Our Ayda river watershed is located at the Kisra delegation in the Governorate of Siliana in northwestern Tunisia (Figs. 1 and 2). The study site is a slope forming part of 6 hill lakes programmed by the Regional Commissariat of Agricultural Development (CRDA) of the Governorate of Siliana. This project is part of the Tunisian strategies for land protection and the development of the agricultural sector in mountainous areas. The topography of the basin is relatively rugged, characterized by slopes generally medium to strong and a significant altitudinal variation, ranging from 844 m to the outlet at 1182 m on the high reliefs of the southwest end of the basin respectively in Djebel Barbrou and Kef Guedouara (Fig. 4). The climate of the area belongs to the semiarid bioclimate lower than fresh winter.

Fig. 1
figure 1

Map of geographic location and drainage network of the Ayda river watershed

Full size image
Fig. 2
figure 2

(a) Ayda river watershed and (b) left bank of Ayda river (clay series with intercalations of alluvia limestone blocks) (Photo Arbi Chafai 2018)

Full size image

The study area is delimited by the isohyets 340 and 360 mm. In the absence of a rainfall post within the watershed of Ayda river, we used information from the Skhira Kef Labeid post which is close to less than 9 km from the catchment area.

The years 1989–1990, 1995–1996, and 2004–2005 are the wettest with rainfall values greater than 470 mm per year, while the driest year is 1996–1997, or rainfall has not exceeded 180 mm (Table 2). The average annual temperatures are between 16 and 18 °C, the minimum averages vary between 17.5 and 17.7 °C, while the maximum averages vary from 34.2 to 34.4 °C.

From a hydrology point of view, the surface waters of the basin come from the mainstream represented by Ayda river, in which small tributaries such as river Melah, river Arar, and river Saad emerge.

It occupies about 80% of watershed area. The highlights of mountains (above 1050 m), such as Djebel Barbrou and Kef Guedouara, cover a small area not exceeding 9%. The hypsometric curve represents the distribution of the surface of the watershed according to its altitude. This curve provides a synthetic view of the slope and therefore gives an idea of the relief (Fig. 3).

Fig. 3
figure 3

Hypsometric curve of the Ayda river watershed

Full size image
Fig. 4
figure 4

Digital elevation model (DEM) of study area

Full size image
Fig. 5
figure 5

Geological map of study area

Full size image
Fig. 6
figure 6

Geological section (B-A) in the downstream of Ayda river watershed

Full size image
Fig. 7
figure 7

Geological section (D-C) in the upstream of Ayda river watershed

Full size image
Fig. 8
figure 8

Map of the distribution of soils in the watershed

Full size image
Fig. 9
figure 9

Map of land use of study area

Full size image
Fig. 10
figure 10

Maps of the slopes of the study area

Full size image
Fig. 11
figure 11

Organization chart for the erosion map of Ayda river watershed

Full size image
Fig. 12
figure 12

K factor map of Ayda river watershed

Full size image
Fig. 13
figure 13

LS factor map of Ayda river watershed

Full size image
Fig. 14
figure 14

C factor map of Ayda river watershed

Full size image
Fig. 15
figure 15

P factor map of Ayda river watershed

Full size image
Fig. 16
figure 16

Map of soil loss (t/ha/year) in Ayda river watershed

Full size image

Morphological indices are used to characterize the physical environment of a watershed. They have a direct influence on the response time of the basin, which is to say on the water travel time through the drainage network. The determination of these indices is based on knowledge of the hypsometric distribution over the entire watershed. These morphometric factors, illustrated in Table 1, are estimated using ArcGIS software.

Table 1 Geometric characteristics of Ayda river watershed
Full size table

Geologically, based on the geological map of Djebel Barbrou (Rouhia) 1/50000, the watershed of Ayda river is constituted (Fig. 5) in large part by lands of Lutetian age superior-Priabonian. These soils are formed by greenish gray clays with lumachellic calcareous alternations. The rest of the basin is covered by Oligocene sands which occupy the high reliefs of the southwestern end of the watershed at Djebel Barbrou and Kef Guedouara, respectively. The Oligocene series are intersected by a single SW-NE fault at the northern flank of Djebel Barbrou.

The two sections of Figs. 6 and 7 summarize, in a simplified way, the geological structure and lithology of the basin area of the Ayda river watershed. Indeed, the two banks are formed by greenish brown clay series, with lumachelles and intercalations of limestone levels with nodular appearance, which constitute the lithological continuity of the mounds of Rass El Fedja and Koudiet Znida (Fig. 2).

The main soil types that occupy the Ayda watershed are raw mineral soils (lithosols, regosols) (69.9%), poorly developed soils (18.6%), and complex soils (11.5%) (Fig. 8). The raw mineral soils are sandy-sandy loam soils with a depth of 20 cm. The complex soils formed mainly of the mineral soils of erosion associated with poorly developed soils and some calcimagnetic soils are distributed on the left bank of the watershed. They are medium-textured soils (clay loam to clay-sand), also with a profile depth of 20 cm.

These two types of soil units are sensitive to water erosion by their texture and their location on slopes steep slope. The poor alluvial soils are located close to the outlet, and they are fine-textured soils with a high clay accumulation compared with other types of soils.

Distribution of the land use at the Ayda river watershed indicates several areas of occupation; the agricultural land (cereals, olive, and arboricultures) totals an area of 274 ha or 37.5% of the watershed. Forest (clear scrubland) covers a rate of 14.6% of the total area and located in the northeastern part of the catchment area. Pastures dominate this basin, and they occupy 47.5% of the area, the equivalent of 347 ha. The remainder of the watershed area is occupied by bare ground (Fig. 9).

This watershed is characterized by a relatively low vegetative cover because of the harsh climate and the steepness of the slopes which accelerates the flow velocity; therefore, the water erosion is manifested at the beginning by a superficial soil stripping.

When the runoff is thus concentrated and the speed increases, the erosion gullies soils. The slope map is based on the digital elevation model (DEM) of the study area (Fig. 10).

From a slope perspective, six classes have been identified. Table 2 shows that about 42% of the catchment area has a low to very low slope class (less than 10%). When 32% of the basin area is marked by fairly steep to very steep slopes (greater than 20%), it is located in the high reliefs of the southwest end of the watershed respectively at Djebel Barbrou and Kef Guedouara and at level of watercourses. The dominant class is that between 5 and 10% is the weak class; it occupies almost 30% of the total area.

Table 2 Classification of slopes according to areas
Full size table

Presentation of RUSLE

The work of Renard et al. (1997) resulted in the revised universal ground loss equation (RUSLE); this model is more advanced than the first model (USLE) of Wischmeier and Smith (1978). RUSLE allows the estimation of the amount of soil uprooted by rainwater but also obtains a geographical map of the distribution of the erosion risk.

It estimates land losses for long-term predictions (> 20 years) and for sheet and rill erosion of agricultural parcels. RUSLE is empirical since it is based on observations and statistics. This model is based on a linear function involving the spatial combination of the different factors contributing to soil erosion: C, P, LS, R, and K.

The loss in soil (A) is determined by the product of the six factors according to the following equation:

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

  • A: Loss in soil per unit area and year (t/ha/year)

  • R: Rain runoff (MJ mm ha−1 h−1 year−1)

  • K: Erodibility of the soil (t ha h/ha MJ mm)

  • LS: Slope length/slope steepness

  • C: Cover management

  • P: Support practice

Database

The evaluation of each of these factors requires the realization of a set of analysis and processing operations to obtain the basic data, indispensable for the development of the factors of water erosion maps of the watershed of the river Ayda.

  • For the R factor, the rainfall data are collected from the nearest rainfall station of the Ayda river watershed (Skhira Kef Labeid) because of the absence of a rainfall post within the basin (Table 3).

  • We determined the K factor using the two maps of agricultural vocation and pedology of the Governorate of Siliana, produced by the Hydraulic and Rural Development Company (HRDC) in 2003 on a working scale of 1/100,000. These maps were digitized under ArcGIS and then updated and verified using recent satellite images obtained from Google Earth 2018.

  • The values of the erodibility factor are determined from a literature review Zante and Collinet (2001) at El Hnach site (Maktar, Tunisia) and Collinet et al. (2001) at the Zanfour site (Kef, Tunisia).

  • The LS factor was calculated from an SRTM (Shuttle Radar Topography Mission) image; its resolution can reach 30 m obtained on the site: http://opentopo.sdsc.edu/datasets. These data refer to topographic vector and matrix files provided by two US agencies: NASA (National Aeronautics and Space Administration) and NGA (National Geospatial Intelligence Agency); radar interferometry data was collected by the space shuttle Endeavor in February 2000 over a height of 233 km.

  • The land use is extracted from the data of the agricultural map of Siliana, the land use map carried out by the Hydraulic and Rural Development Company in 2003 on a working scale of 1/100,000, and recent data digitized from Google Earth on satellite images of 2018.

  • The values for the anti-erosion practice factor are determined based on the studies done by FAO (1977) in Tunisia in 1977.

Table 3 Information sources used for the creation of the database river Ayda watershed data
Full size table

As regards the operation of RUSLE model, the diagram in Fig. 11 summarizes the different steps.

The digital maps are summarized in Table 3. These maps will be used by the model applied in our work, in order to obtain thematic maps of synthesis.

Results and discussion

Rainfall erosivity factor (R)

Wind and water are the two key variables of erosion. Torrential rains promote runoff more than infiltration, which causes the particles to be torn off and transported with the water (Le Bissonnais et al. 2002). Thus, the role of the factor is to characterize the erosive force of precipitation on the ground.

For factor R, we used the formula most adapted for the Tunisian region that is developed by Renard and Freimund (1993).

This regression establishes a relation between the average annual precipitation and the erosivity of the rains; the estimated error for the calculation of the factor R is 10 to 15%.

$$ R=0.0483\ast {P}^{1.610}\kern1.25em \left(\mathrm{Renard}\ \mathrm{and}\ \mathrm{Freimund}\ 1993\right) $$
(2)

where R is the rainfall erosivity (MJ mm ha−1 h−1 year−1) and P is the average annual rainfall (mm).

Examination of the table shows that the average annual erosivity index, calculated over 23 years of observation at the Skhira Kef Labeid station, is of the order of 567.1 MJ.mm/ha h year which corresponds to 337.11 mm of rain (Table 4).

Table 4 Rainfall erosivity (MJ mm ha−1 h−1 year−1) for the Skhira Kef Labeid station during the period 1988–2011
Full size table

The 1995/1996 hydrological year is the most erosive with a value of about 1239.53 MJ mm ha−1 h−1 year−1, and the 1996/97 hydrological season is the least erosive with 203.37 MJ mm ha−1 h−1 year−1; the corresponding rainfall is 547.9 mm and 178.3 mm, respectively.

Soil erodibility factor (K)

The factor K tells us about the resistance of soil particles to detachment and transport by water. This is already controlled by soil texture and structure, and the vegetation grows on the same soil (Roose and Sarrailh 1990). It is determined experimentally, for a given soil sample, from percentage of the organic matter. In our case study, there is not enough grain size data to define erodibility according to the commonly used equation.

Indeed, the K factor values for the RUSLE model are determined from the soil map (Table 5). Each pedological unit is assigned a K index, from Zante and Collinet et al. (2001) at El Hnach site (Maktar, Tunisia) and Collinet et al. (2001) at the Zanfour site (Kef, Tunisia).

Table 5 Erodibility factor of its K soils adopted for Ayda river watershed
Full size table

The values of the K index in the watershed range from 0.036 t ha h/ha MJ mm for soils resistant to 0.08 t ha h/ha MJ mm for the most erodible soils (Fig. 12).

Indeed, the synthetic map shows that the average erodibility is 0.036 covering 69.9% of the total area of the basin, materialized by the class of mineral soils of the lithosol and regosol type. The rest of the basin has high values with predominantly poorly developed (18.5%) and complex (11.5%) soils which show strong erosion where K is between 0.054 and 0.08 (Table 5).

Topographic factor (LS)

This factor LS is obtained from two other sub-factors L and S of the slope which are its length and its gradient, respectively. According to Roose (1994), the morphology of the slope and its extension are the two key variables of water erosion. They are calculated from the digital elevation model (DEM). From the formula of Wischmeier and Smith (1978), we calculated the LS factor based on the inclination of the slopes.

$$ \mathrm{LS}={\left(\mathrm{Flow}.\mathrm{acc}\ast \mathrm{Resolution}/22.13\right)}^{\mathrm{m}}\times \left(0.065+0.045\times S+0.065\times {S}^2\right)\ \left(\mathrm{Wischmeier}\ \mathrm{and}\ \mathrm{Smith}\ 1978\right) $$
(3)

LS factor can reach 32.4 as the maximum value and zero as the minimum value (Fig. 13). Table 6 shows that more than 95% of the values are between 0 and 2.67 and correspond to areas with slight slopes and with very low LS values (therefore not very sensitive to the erosive phenomenon). At the level of the left bank of the basin, the steep slopes, having a high LS factor and the values between 2.67 and 32.4 which correspond only 5% of the total surface of the Watershed.

Table 6 Class apportionment of watershed topographic factor Ayda river
Full size table

Cover management factor (C)

The vegetation covering the soil decreases the erosive effect of raindrops, slows runoff, and increases infiltration. This amortization by the plant cover varies according to its density, type of vegetation (tree or shrub or grass), and the density of the root system (Souter et al. 2007). Values of C attributed to each land use class are determined from the work done in Tunisia by Cormary and Masson (1964), Masson (1971), and Zante and Collinet (2001) relating to the El Hnach site (Maktar, Tunisia), Zante et al. (2003) on the site of Abdessadok (Maktar, Tunisia), and Collinet et al. (2001) in the Zanfour (Kef, Tunisia). The values of C vary between 0.2 for clear scrubland and 1 for bare grounds (Table 7). In the Ayda river watershed, there are six types of land use (Fig. 14).

Table 7 Factor C values by soil type
Full size table

Conservation support practice factor (P)

The P factor is concerned with all that is soil conservation whether it exists or not in order to oppose water erosion (Bouguerra and Bouanani 2016). The values of the P factor do not exceed 1, and the value of 1 is assigned on undisturbed natural terrain. Besides that, P also considers the slope. In the Ayda river watershed, the lowest and average P factor values are assigned to steep to moderate slopes. The P factor varies in our case between 0.55 and 0.6. However, for areas with a slight slope, P oscillates between 0.08 and 1 for steep slopes (Fig. 15).

Evaluation of soil losses

Intersection of the main factors of the RUSLE model, including the aggressivity of the R rain, the erodibility of the K soils, and the combined effect of the degree and extent of the LS angle, the plant cover C, and the anti-erosion practices, allows to develop an erosion risk map informing about erosion potential in t/ha/year at any point in the Ayda river watershed. The only variable factor from one year to another is the erosivity of the R rainfall. For this factor, an annual average was used for the period 1988–1989 to 2010–2011 in the order of R = 567.1 for an annual precipitation P = 337.11. The results obtained show a spatial distribution of erosion classes, while the minimum and maximum losses are 0 and 330 t/ha/year, respectively (Fig. 16). Erosion risk rates in the Ayda river watershed are grouped into 5 classes: very low (0 and 9 t/ha/year), low (9 and 38 t/ha/year), medium (38 and 98 t/ha/year), elevated (98 and 189 t/ha/year), and very high (greater than 189 t/ha/year).

The soil loss map shows that 91.5% of the catchment area has relatively low losses and belongs to class 0 to 9 t/ha/year. The low losses lie largely in the center of the catchment area in the study on low slopes. 6.5% of the area is in class 9 at 38 t/ha/year and reflects low losses, while the rest of the basin represents only 2% of average to high soil losses (more than 38 t/ha/year) (Table 8).

Table 8 The different classes of soil losses in t/ha/year
Full size table

The description and discussion of the map of erosion classes necessarily require a parallel reading of the physical parameters (slope and altitudes) and natural maps (lithology and vegetation) of the entire watershed. Referring to the maps of the various factors, the highest erosion class is located mainly in the high reliefs of the southwestern end of the watershed at Djebel Barbrou and Kef Guedouara, respectively, which are mainly occupied by rangelands and agricultural land with altitudes exceeding 1050 m and slopes of up to 70%. This zone shows the combined effect of vegetation cover and that of steep slopes. These allowed us to conclude that the losses in soils are well correlated with the topographic factor and the vegetation cover. Roose (1994) showed that an increase of 0.25% at the slope level can lead to an 11% increase in the annual average runoff rate and an increase of 3.6 t/ha/year in losses, averages in soils.

Referring to the soil map of the basin, there are friable clay-like formations, silts mainly occupied by little evolved soils of deep but very erodible alluvial input, and complex soils with a strong erodibility (K between 0.054 and 0.08). The factors of the revised global equation (topography, erodibility, climatology, and vegetation cover) of soil loss are well correlated with soil loss.

These results are close to those obtained on the Siliana dam (Tunisia) (Chabaan 2016) which has erosion (among 0 and more than 100 t/ha/year), while the low losses (< 7 t/ha/year) cover 70% of this basin. Similar results at the Jannet watershed (Maktar, Tunisia) (Ben Cheikha and Gueddari 2008) with erosion allying 0 and 435 t/ha/year under the effect of steep slopes (0 to70%) and the presence of very dense vegetation. Concordant with the latter found in the watershed of Ayda river, another study conducted by Touihri (2013) shows that soil losses in the Jannet basin vary (to each of 0 and 258 t/ha/year) in the M’Richet El Anze watershed (Bargou, Tunisia).

Conclusion

The main goal of our study is to map areas vulnerable to water erosion and estimate the rate of soil loss. Indeed, the empirical RUSLE (revised universal ground loss equation) model has been applied which considers the impact of most factors, including the erosivity of rainfall (R), the ability of soil to erosion (K), soil occupation (C), and slope length/inclination (LS).

This model is developed from field-acquired data or by remote sensing data managed with GIS. The results, in the form of an erosion risk map, show that the Ayda river watershed has soil losses (t/ha/year) ranging from 0 to 330. More than 91% of the area total watershed have values below 9 t/ha/year, while 8% show moderate to high erosion rates. Water erosion is visible throughout the Ayda river watershed. The slopes are subject to greater erosion, whereas reduced erosion is observed in the alluvial valleys with low relief. Basin geology is formed mainly by friable formations. Soil erodibility is determined in a synthetic way based on a literature review to assign to each soil type a K factor value ranging from 0.036 to 0.08. The slopes are between 0 and 70%, and the LS values are between 0 and over 32.

The losses can reach alarming values of 35%; they resulted from the interaction of several factors combined. The percentage of the total area of the basin has slopes greater than 15%, soils with high erodibility (30% of soils show a K factor among 0.054 and 0.08), and sparse/degraded plant cover (cereal zones, rangelands, clear scrubland, and bare soil).

Finally, the RUSLE model provides a lot of information about the erosion process and the spatialization of erosion-prone areas in this basin. These results only present relative values which can help assess the rate of potential erosion and the determinants of land degradation, develop evolution scenarios, and identify priority areas requiring conservation intervention and emergency actions to stop erosion.