1 Introduction

Land degradation is causing serious social, economic, and environmental problems worldwide, especially in the more vulnerable areas with an arid or semiarid climate. Jones and Montanarella (2003) reported that land degradation is the reduction or loss in arid, semi-arid, and dry sub-humid areas of the biological or economic productivity and complexity of rain-fed cropland, irrigated cropland, rangeland, pastures, forests, and woodlands, including processes arising from human activities and habitation patterns. Therefore, land degradation directly affects livelihoods and food security, particularly in dry areas with unfavorable climate, by reducing the productivity of land resources and adversely affecting the stability, functions, and services derived from natural systems (UNCCD 2015). The types and causes of land degradation vary from one site to another, even within a short distance. Essentially, land degradation is the result of the misuse/overexploitation of the natural resources base, particularly through inappropriate and unsuitable agricultural practices, overgrazing, deforestation, and forest degradation (Dengiz et al. 2015). Moreover, in recent decades, the degradation of previously naturally vegetated or productive agricultural lands, leading in many cases to a barren, desertified landscape, has dramatically accelerated in many regions of the world (Pla 2008).

The important functions of natural resources, especially soils, are threatened by many severe degradation processes. One of the major degrading factors is soil erosion caused by wind and/or water. RIVM (2000) stated that water erosion is one of the most important land degradation processes for EU countries. It was also reported that southern EU countries are at greater risk of water erosion, with high water erosion risk rates of 58, 66, 66, and 85 % in France, Italy, Spain, and Greece, respectively. Turkey is a hilly to mountainous country. The average elevation is approximately 1250 m, and 62.5 % of the total land has more than 15 % slope. Due to its topographic limitations (generally steep or sloping terrain) and to climate and soil conditions, soil erosion is Turkey’s biggest land degradation problem; some 58.7 % of the land is exposed to severe or very severe soil erosion (Ministry of Agriculture, Forestry and Villages 1987). In addition, there is active erosion in 59 % of agricultural lands, 54 % of forest lands, and 64 % of rangelands of Turkey.

The soil organic carbon content is a key component of any terrestrial ecosystem, and any change in its abundance and composition has substantial effects on many of the processes in the system. Closely linked to the process of soil erosion is the widespread and substantial decline in soil organic carbon content in arid and semi-arid areas. Soil organic carbon, which is extremely important in all soil processes, is essentially derived from plant and animal residues, and is produced by microbes and decomposed at a rate determined by temperature, moisture, and ambient soil conditions. There are two groups of factors that influence soil organic carbon level; natural factors (climate, soil parent material, land cover and/or vegetation type, and topography) and human-induced factors (land use, management, and degradation).

Simulation models are the most effective means of predicting soil erosion processes and their effects through the use of geographic information system (GIS) and remote sensing (RS) for large areas. Recent advances in space and computer technologies have provided us with the opportunity to process large amounts of data (multi-source), not only spectral data but data such as elevation, slope, aspect and relief (Bayramin 1998). Therefore, models have the potential to make major contributions toward developing better conservation practices and improving the management of land resources (Meyer 1980; Edwards et al. 2008; Csafordi et al. 2012).

Empirical models, such as the universal soil loss equation (USLE) and its revised version RUSLE (Renard et al. 1991), have been used in many regions for large scale catchment risk mapping. Burrough (1986) introduced the principles of GIS tools for collecting, storing, manipulating, and displaying spatial data, and Eedy (1995) reported the advantages of using GIS in environmental assessment. The estimation of soil erosion risk and its spatial distribution can be performed at reasonable cost and with improved accuracy across larger areas using RS and GIS techniques (Millward and Mersey 1999; Wang et al. 2003). Ouyang and Bartholic (2001) developed an interactive web-based approach to using RUSLE and GIS to predict soil erosion and Martin et al. (2003) used a GIS/RUSLE model to estimate sheet erosion in a watershed. They reported the ease with which GIS could be integrated with the RUSLE to identify discrete locations with relatively precise spatial boundaries that have a high sheet erosion potential together with the areas where management practices might be implemented to prevent soil erosion. Martin et al. (2003) also stated that GIS/RUSLE modeling represented a quick and inexpensive tool for estimating sheet erosion within watersheds when using publicly available information. Furthermore, Lu et al. (2004) applied RUSLE, remote sensing and GIS to the mapping of soil erosion risk in the Brazilian Amazonia. To contribute to improved land management practices and to establishing sustainable natural resources and environmental assessment in Turkey, the main objective of the current research was to integrate GIS and RUSLE to model soil erosion risk and carbon loss in a catchment in the Central Black Sea region.

2 Materials and methods

2.1 Field description of the study area

This study was carried out in the Alaca catchment which is located in Çorum and Yozgat Provinces in the Central Black Sea region of Turkey (Fig. 1).

Fig. 1
figure 1

Location map of the study area

Full size image

The total study area is 1656.4 km2. Geological formations in the catchment are generally dominated by Mesozoic–Tertiary ophiolitic series and limestones. In addition, the low lying lands of the study area are occupied by Quaternary alluvial and colluvial deposits, consisting of mixed gravel materials. The study area consists of various topographic features and is characterized by mountains (especially intermediate and low relief mountain ranges), hills and plains, and the south-western and south-eastern areas are characterized by steep slopes. Average elevation of the catchment is 1275-m above sea level. The highest elevation is at a mountain named Toprakdede Tepesi (1765 m) that is located in the eastern part of the catchment. A continental climate prevails in the basin, with very cold and rainy/snowy days in winter and very hot and dry weather in summer. Annual mean precipitation and mean temperature of the study area, which is different from the general Black Sea climate, are 364.8 mm and 9.5 °C, respectively. Koçhisar and Alaca reservoirs are two water storage structures in the catchment.

The soils of the basin are mainly brown soil (65.2 %) and brown forest soil (9.3 %). The distribution of the other soil types in the study area is as follows: alluvial soil (6.9 %), chestnut soil (6.4 %), non-calcaric brown soil (6.3 %), and colluvial soil (5.9 %). In addition, 5 % of the area is composed of water surface, settlement areas, and bare rocks. The areas and ratios of the soil types were generated with GIS from 1:25,000 soil maps produced under the former United States of America classification system [United States Department of Agriculture (USDA) soil classification system 1938] and provided by the General Directorate of Rural Services of Turkey (KHGM). (Note: these soil maps have not yet been updated using the modern soil taxonomy classification system.)

2.2 Methods

2.2.1 The RUSLE model

The USLE model consists of a set of calculations to estimate soil erosion from a plot of land with homogeneous characteristics (Wischmeier and Smith 1978). The RUSLE model retains the general framework of the USLE but refines the calculations for each of the five erosion factors through greater temporal and spatial precision (Renard et al. 1997). With the RUSLE function, the average amount of soil loss is expressed as a function of five factors (Wischmeier and Smith 1965, 1978; Renard et al. 1997) as follows:

$$ A = R\, \times \,K\, \times \,{\text{LS}}\, \times \,C\, \times \,P, $$

where A is the computed average amount of soil loss in Mg ha−1 year−1, R is the rainfall-runoff erosivity factor (MJ mm ha−1 h−1 year−1), K is the soil erodibility factor (Mg h MJ−1 mm−1), LS is a combination of the slope length and steepness factors, C is the cover and management factor, and P is the support practices factor (LS, C, and P are dimensionless).

2.2.1.1 Rainfall erosivity factor (R)

According to Wischmeier and Smith (1978), the rainfall erosivity (R) factor represents the effect of rainfall intensity on soil erosion and requires continuous, detailed data. To calculate the R factor using monthly and annual rainfall data, Arnoldus (1980) reported a modified Fournier index (MFI):

\( {\text{MFI}} = \frac{{\mathop \sum \nolimits_{i = 1}^{12} (pi)^{2} }}{P}, \)

where pi is the mean rainfall amount (mm) for month i, and P is the mean annual rainfall amount (mm).

According to Arnoldus (1980), the MFI is a good approximation of the R factor, with which it is linearly correlated. For the current investigation, precipitation data from 1960 to 2014 from ten meteorological stations located in and around the study area were collected, and the MFI was estimated for each station. Coordinate information of the meteorological stations is listed in Table 1.

Table 1 Meteorological stations used in the study and computed MFI and R factor values for each station
Full size table

To approximate the R factor using the calculated MFI for each station, the following R–MFI relationship, as suggested by Irvem et al. (2007) for a climatologically similar area, in terms of rainy days, amount and range of precipitation distributed over the four seasons, was used as follows:

$$ R\, = \,0. 1 2 1 5 \times {\text{MFI}}^{ 2. 2 4 2 1} . $$

The MFI and R values for each meteorological station were calculated, and the R factor map of the current study area was generated by interpolation in GIS.

2.2.1.2 Soil erodibility factor (K)

K is defined as the rate of soil loss per unit of R as measured on a unit plot, and it accounts for the influence of soil properties on soil loss during storm events. The study area was divided into grid squares (Fig. 2).

Fig. 2
figure 2

Soil sampling design on the study area

Full size image

A total of 348 soil samples were collected from the surface (0–20 cm) of the study area, according to the grid squares. The samples were transported to the laboratory where root material was removed, while the soil sample was being gently crumbled. These samples were used to determine physicochemical soil properties, including: sand %, silt %, clay %, organic matter %, structure, and permeability classes. The K values were computed from these soil properties with the following equation (Wischmeier and Smith 1978):

$$ K = \frac{1}{100}\left\{ {2.1 \times 10^{ - 4} \times \left( {12 - {\text{OM}}} \right) \times \left[ {{\text{SI}} \times \left( {{\text{SA}} + {\text{SI}}} \right)} \right]^{1.14} + 2.5 \times \left( {{\text{PE}} - 3} \right) + 3.25 \times \left( {{\text{ST }} - 2} \right)} \right\}, $$

where K is expressed in units of Mg h MJ−1 mm−1, and OM, SI, SA, PE, and ST are percentages of soil organic matter content, silt content, sand content, permeability class, and structure code, respectively. If soil organic matter content was equal to or greater than 4 %, OM was assumed constant at 4 % (Renard et al. 1997).

Soil samples were analyzed, and particle size distribution (Gee and Bauder 1986), hydraulic conductivity (Klute and Dirksen 1986), aggregate stability (Kemper and Rosenau 1986), and organic matter content (Jackson 1958) were determined in the laboratory. The K factor map of the study area was generated with the interpolation models of GIS.

2.2.1.3 Slope length and steepness factor (LS)

Slope is an important factor influencing overland flow generation and soil erosion, particularly beyond the critical slope degree. The slope length and steepness (LS) factor accounts for the effect of topography on soil erosion (Renard et al. 1997). Slope length is defined as the horizontal distance from the point of origin of the overland flow to the point where either the slope gradient decreases sufficiently for deposition to begin or runoff is concentrated in a defined channel (Wischmeier and Smith 1978). An increase in the LS factor indicates higher overland flow velocities and correspondingly greater erosion (Onori et al. 2006).

Hickey (2000), Boggs et al. (2001), Kinnel (2001), Gertner et al. (2002), Wang et al. (2003), and van Remortel et al. (2004) used digital elevation model (DEM) to estimate the LS factor. In the present study, the grid-based DEM was generated from contour vector data which were digitized using a 1:25.000 scale topographic map.

2.2.1.4 Land cover and management factor (C)

The land cover and management factor (C) is defined as the ratio of soil loss from land with specific vegetation cover to the corresponding soil loss from continuous fallow (Wischmeier and Smith 1978). In the present study, remotely sensed data were used to estimate the C distribution based on land use and land cover (LULC) classification results (Millward and Mersey 1999; Reusing et al. 2000), assuming that areas with the same amount of land cover have the same C values.

The supervised classification method (maximum likelihood) was used to produce the LULC classes, as described by Lillesand and Kiefer (2000), for determining the C factor. The LULC classes are listed in Table 2. The C factors used in this study were adopted from Renard et al. (1997), Yang et al. (2003) and İrvem and Tülücü (2004).

Table 2 LULC classes and C factors
Full size table

After creating the LULC map of the study basin, the C factors for the land classes were entered as attributes, and a C factor map of the study area was generated using the reclassification method in the GIS.

2.2.1.5 Support practice factor (P)

Renard et al. (1997) indicated that the support practice factor (P) is the ratio of soil loss using a specific support practice to the corresponding loss with upslope and downslope tillage. As in most agricultural lands in Turkey, agricultural practices in the study area consist of upslope and downslope tillage without any conservation support practices, such as contouring or terracing. This situation is particularly evident during ground truthing observations. To remove the P factor from the soil erosion estimates, P was set at one, as suggested by Wischmeier and Smith (1978).

2.2.2 Soil organic carbon density estimation

Soil organic carbon density (SOCD) for each soil sample was estimated with the following equation:

$$ {\text{SOCD}}_{\text{D}} = \frac{{\left( {1 - \delta_{i} \,\% } \right) \times \rho_{i} \times C_{i} \times T_{i} }}{100}, $$

where SOCDD represents the SOC density of a soil sample with a depth (20 cm); δ i % represents the volumetric percentage of the fraction >2 mm (rock fragments), ρ i is the bulk density (g cm−3), C i is the SOC content (g kg−1), and T i represents the thickness (cm) of the layer i.

2.2.3 Interpolation analysis

In this study, different interpolation methods (inverse distance weighing—IDW, radial basis function—RBF and kriging) were applied for predicting the spatial distribution of R, K and SOCD. Kriging is a geostatistical technique similar to IDW in that it uses a linear combination of weights at known points to estimate the value at an unknown point. Kriging uses a semivariogram, a measure of spatial correlation between two points, so that weightings change according to the spatial arrangement of the samples. In contrast to other estimation procedures, kriging provides a measure of the error or uncertainty of the estimated surface. Several forms of kriging interpolation exist, including ordinary kriging (OK), simple kriging (SK), and universal kriging (UK).

In the present study, root mean square error (RMSE) was used to evaluate the interpolation techniques, with the lowest RMSE indicating the most accurate prediction. Estimates were determined with the following formula:

$$ {\text{RMSE = }}\sqrt {\frac{{\sum {(z_{i*} - z_{i} )^{2} } }}{n}} , $$

where z i is the predicted value, z i* is the observed value, and n is the number of observations.

3 Results and discussion

3.1 Distribution of R, K, and SOCD was determined with geostatistical analysis

Interpolation analyses for R, K, and SOCD were used to identify the best predictive model from among 14 different semivariogram models, namely, inverse distance weighting—IDW with weightings of 1, 2, and 3 powers and radial basis function—RBF with thin plate spline (TPS), ordinary kriging (OK), simple kriging (SK), and universal kriging (UK) with spherical, exponential, and Gaussian variograms. Following that, the variogram or function of each interpolation method yielding the best results was determined. Comparison of interpolation methods for R, K, and SOCD is provided in Table 3. Finally, according to the interpolation analyses, simple kriging with the spherical model was used to estimate or predict K and SOCD, while the exponential model was found to determine the R value at unsampled locations. The spatial distribution maps of R, K, and SOCD are shown in Figs. 3, 4 and 5.

Table 3 Cross validation according to interpolation methods
Full size table
Fig. 3
figure 3

The R factor map of the study area (MJ mm ha−1 h−1 year−1)

Full size image
Fig. 4
figure 4

The K factor map of the study area (Mg h MJ−1 mm−1)

Full size image
Fig. 5
figure 5

Spatial distribution of SOCD (kg m−2)

Full size image

Figure 4 shows the R factor map of the study area. The average annual R factor ranged from 362.51 to 966.27 MJ mm ha−1 h−1 year−1. The computed MFI and R factors for each meteorological station are listed in Table 1. There was more rainfall erosivity in the south-western part than in the north and north-eastern parts, because rainfall erosivity is closely related to precipitation, which increases from north to south in the catchment (Fig. 3).

Başkan and Dengiz (2008) compared soil erodibility (K) maps prepared by traditional and geostatistical methods for Sogulca Basin soils and reported that a kriged K map displayed significantly better results than a K map developed with traditional methods. Figure 4 shows the map for the K factor, generated with the geostatistical approach, which varied from 0.01 to 0.14 Mg h MJ−1 mm−1. The K factors were low in the southwest and southeast of the study area, whereas high K values were determined in northern parts of the catchment where the river has formed and streamed throughout an alluvial land located almost in the central part of the catchment at the discharge point into Koçhisar Dam. The K values were also typically high in the eastern parts of the catchment on steep slopes where mostly medium and sandy textured soils occur.

In this study, for the SOCD in surface soil (0–20 cm), the isotropic spherical model provided the best fit for the computed semi-variance points (the distribution map of SOCD of the surface soil is shown in Fig. 5). In addition, SOCD values were between 0.18 and 4.92 kg m−2 in the catchment (Fig. 5). Guggenberger et al. (1995) stated that the type of land use system is an important determinant of SOC levels, particularly in top soils. Furthermore, changes in land use and management practices influence the amount and rate of SOCD gain or loss.

3.2 Slope length and steepness (LS)

Figure 6 shows the map of the LS distribution, which ranged from 0.01 in the flat areas in the north-central part of the basin to 28.85 in the highlands (elevation approximately 1700 m) in the Southwestern, and some northern parts of the basin which have the steepest slopes, the greatest variability in elevation and high LS values. The LS values were highest in the mountainous areas and deep valleys. These areas were mostly located in the upper part of the basin.

Fig. 6
figure 6

The LS factor map of the study area

Full size image

Figure 7 shows the map of the C generated from satellite imagery and forest inventory for the LULC map of the study area. The study area was composed of 57.6 % rain-fed agricultural lands, 32.2 % mixed rain-fed agriculture and pasture, 9.2 % irrigated agriculture 0.9 % sparse coniferous forest and pasture, and 0.1 % bare rocks, artificial area, and water bodies. The C values ranged from 0 to 1.0 (Table 2) and were especially low in southwestern parts of the basin, indicating that forests (broad leaf and coniferous forests) and some pasture land have low erosion risk, but most of the cultivated and fallow agricultural land and degraded pasture areas have high or very high erosion potential.

Fig. 7
figure 7

The C factor map of the study area

Full size image

The P factor should be increased in agricultural areas to decrease soil erosion to an acceptable level. However, in flat or almost flat areas, where agriculture is the main land use type, the erosion risk was found to be low (note: no supporting practice was observed in the study area). Therefore, the P value of 1 was not used in calculations (Wischmeier and Smith 1978; Çanga 1985).

The RUSLE function estimates only local erosion amounts and cannot be used to estimate the sediment yield for an entire watershed (Renard et al. 1997). The map of the potential soil losses predicted by the RUSLE as a product of R, K, LS, and C is shown in Fig. 8. Annual soil losses in tonnes per hectare per year with respect to the different soil erosion classes of Bergsma et al. (1996) are provided in Table 4. Table 4 shows that more than the half of the study area is in the very low and low erosion risk classes which range from 0 and 12 t ha−1 year−1, whereas 4.4 % of the total area (723.6 ha) is under high and very high erosion risk of 35 to 150 t ha−1 year−1.

Fig. 8
figure 8

Distribution the different soil erosion classes for the study area

Full size image
Table 4 Annual soil loss predicted for the different soil erosion classes of the study area
Full size table

4 Conclusions

The RUSLE model was applied to estimate soil loss with RS and GIS in the Alaca catchment located in the Central Black Sea region of Turkey. Detailed pluviograph data for calculating the R factor were not available. Therefore, the R factor was estimated with MFI data based on mean annual and monthly rainfall data from the meteorological stations located in and around the study area. Existing soil data were inadequate for the determination of the K factor. A new K map was produced by collecting soil samples from the field and using a geostatistical method. An SODC map was also produced. The LS values were estimated with a GIS program automated hydrologic procedure to calculate the slope length and steepness using DEM. The C values were determined from the LULC map of the study area, which was derived from satellite images and from forest inventories. The P factor was assumed to be 1, meaning that soil conservation support practices were not used in the study area, as confirmed by field observations. Soils and topographical properties of the watershed had a greater influence on the magnitude of potential soil losses than land use/land-cover type, since the R factor did not change significantly in the study area. In addition, the distribution of SOC losses almost matched the distribution of soil erosion risk classes in the study area. Particularly, carbon levels on south west and northwest parts of the study area associated with high soil erosion levels.

The RUSLE erosion model is very useful for the assessment of erosion risk status, because the conventional methods impose high time and labor costs for data collection, including the measurement of soil erosion in heterogeneous, patchy, and large areas. Bayramin et al. (2003) stated that these problems can be overcome using predictive models and new techniques. With GIS, the collected data can be easily analyzed and used to map soil erosion risk and prepare a land management plan for the sustainable use of land resources. Moreover, the model can identify the areas with erosion risk for decision-makers, so that they can develop soil and water conservation plans in general, including research in the areas of high erosion risk to mitigate the problem.

Finally, the application of the RUSLE methodology produced a consistent pattern of soil erosion mapping among different land uses, slope positions, and soil groups and reasonably predicted the potential annual soil losses, including the identification of the most erosion-prone areas due to concentrated flow. This approach can also be used to determine the area’s most sensitive to soil erosion on a regional scale, which could help facilitate comprehensive soil conservation management and sustainable land use practices.