Water Erosion Monitoring and Prediction in Response to the Effects of Climate Change Using RUSLE and SWAT Equations: Case of R’Dom Watershed in Morocco

Abstract

Soil erosion is an increasingly issue worldwide, due to several factors including climate variations and humans’ activities, especially in Mediterranean ecosystems. Therefore, the aim of this paper is: (i) to quantify and to predict soil erosion rate for the baseline period (2000–2013) and a future period (2014–2027), using the Revised Universal Soil Loss Equation (RUSLE) and the Soil and Water Assessment Tool (SWAT) model in the R’Dom watershed in Morocco, based on the opportunities of Remote Sensing (RS) techniques and Geographical Information System (GIS) geospatial tools. (ii) we based on classical statistical downscaling model (SDSM) for rainfall prediction. Due to the lack of field data, the model results are validated by expert knowledge. As a result of this study, it is found that both agricultural lands and bare lands are most affected by soil erosion. Moreover, it is showed that soil erosion in the watershed was dominated by very low and low erosion. Although the area of very low erosion and low erosion continued to decrease. Hence, we hereby envisage that our contribution will provide a more complete understanding of the soil degradation in this study area and the results of this research could be a crucial reference in soil erosion studies and also may serve as a valuable guidance for watershed management strategies.

1. Introduction

Soil degradation is a form of soil loss that involves the detachment, transport, sedimentation and deposition of soil from one area to another by the forces of dynamics and the actions of erosive agents (water, ice, snow, air, plants, animals and humans [1]).

Several factors including human activities and climatic variations such as topography, land cover and management, soil properties, and climate can contribute directly or indirectly in increasing the amount of degraded soil at the catchment scale. It is noted that the effects of climate change due to increased precipitation and temperature, and land use change due to human activities can lead to an activation of erosion processes [2]. This phenomenon is a worldwide environmental problem, that seriously threatens natural resources such as soil fertility, drainage and siltation [3]. It also contributes to the reduction of reservoir capacity and negatively affects aquatic habitats, hydrological systems and downstream water quality because of sediments that are generally charged with nutrients, toxic chemicals and metals [4]. In overall, soil erosion has severe effects on soil and water resources leading to the earth ecosystems disequilibrium.

An estimation of 33% of the world’s land area is classified as moderately to highly degraded because of several factors including erosion, salinization, compaction, acidification [5]. The costs of land degradation show an uncertainty in the global economic impact (ranged from 40 to 490 billion US$) that varies across countries [6]. The United Nations’ (UN) stated that “…soils constitute the foundation for agricultural development, essential ecosystem functions and food security and hence are key to sustaining life on Earth” [7]. Moreover, the protection of the soil has received considerable interest among the 17 UN Sustainable Development Goals (SDGs) [8].

Soil erosion is an increasingly issue worldwide, more particularly in the Mediterranean regions, known as “hot-spot” of climate change because of water scarcity [9]. Morocco, as one of the Mediterranean countries, is highly experienced to severe episodic drought and characterized by extreme natural inter-annual variability, spatial and temporal heterogeneity of precipitation [10], seasonality of water resources and decreasing river flows [11].

The assessment of eroded areas based on field missions monitoring is a challenging and time-consuming task because it is difficult to be applied to broad extensions [12]. To overcome this limitation, many different models are used and developed to quantify soil erosion. These approaches are divided into empirical and physically-based models [13]. Empirical models are easy to be used in areas where the availability of data and parameters is limited, these include Universal Soil Loss Equation (USLE) [14] and its revised version Revised Universal Soil Loss Equation (RUSLE) [15], Modified Universal Loss Equation (USLE Modified) [16], Erosion Potential Method (EPM) [17], and SEdiment DElivery Distributed (SEDD) Model [18]. Physically-based models, require a good knowledge of the physics of the hydrological processes. These include, Soil and Water Assessment Tool (SWAT) [19], Watershed Erosion Prediction Project (WEPP) [19], Agricultural Non-Point Source Pollution (AGNPS) model [20], Areal Non-point Source Watershed Environment Response Simulation (ANSWERS) model [21], Limburg Soil Erosion Model (LISEM) [22], European Soil Erosion Model (EUROSEM) [23], Soil Erosion Model for Mediterranean Regions (SEMMED) [24], Simulator for Water Resources in Rural Basins (SWRRB) [25], MMF (Morgan–Morgan–Finney) model [26], and the Hydrologic Simulation Program-FORTRAN (HSPF) [27].

Due to its severity, gully erosion considered as the major type of water erosion, has become a growing interest of many researchers Therefore, a large number of methods have been used in the mapping of gully erosion susceptibility based on the study of the relationship between a set of independents variables (conditioning factors) and target variable (i.e., presence/absence of gully erosion), including for example, maximum entropy model (MaxEnt) [28], Analytical Hierarchy Process (AHP) model [29], Evidential Belief Function (EBF) [30], Frequency Ratio (FR) [31], and Classification and Regression Trees (CART) [32].

Numerous studies in the literature have used RUSLE equation and SWAT model with successful contributions (see for e.g., [33,34,35]). Also, in Morocco, Boufala, M’hamed, et al. [36] used the SWAT model in the Upper Sebou watershed to monitor and quantify the soil losses.Their results showed that the annual soil loss can reach a maximum value (more than 12.11 t ha⁻¹) at the upstream of the basin and a minimum value (less than 4 t ha⁻¹) and the siltation rate of the retention dam was estimated at 2.12 mm/year. The same authors reported that the soil degradation is due to the physical basin properties. In another research, Boufala M’hamed, et al. [37] applied both RUSLE equation and SWAT model for the assessment of soil erosion in the M’dez watershed in Morocco, reporting that the results obtained are similar for both the approaches, with an estimation of soil degradation at 3.95 t ha⁻¹ when considering SWAT model, and an estimation of soil degradation at 2.94 t ha⁻¹ based on RUSLE equation. Khalid [38] conducted a study in the Sebou watershed (Morocco) to quantify soil degradation due to erosion using RUSLE equation. The results showed that the erosion in this basin is influenced by topography, rainfall, soil properties, crop management conditions, and conservation practices. In the same study, the results showed that 78.83% of the study area has a low risk of erosion, 17.36% a medium risk, 3.04% a high risk and 0.77% a very high risk.

Currently, there is no “best model” for erosion risk mapping as claimed by Abdelwahab, et al. [12]. Moreover, the same authors highlighted that, model performance depends on the background of the model, the required data, the equations used to formalize the processes and the results provided.

Although RUSLE model was developed to estimate water erosion in mild climate, it is easier to adjust for tropical climate than other models. It is prepared to apply at the runoff plot or single hillslope scales, allows the estimation of average annual rate of soil erosion for a specific area for identifying targeting management interventions or practices, cropping systems and erosion control practices. The RUSLE model computes the expected average annual soil loss on hillslopes by multiplying five factors together namely [39]: rainfall erosivity (R), cover management (C), slope length and slope steepness (LS), soil erodibility (K), and support practice (P).

The SWAT model is a semi-distributed, time-continuous hydrologic model generally applied at the watershed scale to simulate surface water and groundwater quality and quantity [40]. This model works with hydrological response units (HRUs) which are areas with unique characteristics identified by land use, soil type, and slope, and then the calculations at the HRU level will be channeled through stream connections to the basin outlet. The estimation of soil loss in the SWAT model is based on the Modified Universal Soil Loss Equation (MUSLE) [41].

When it comes to handling and processing different and large spatial data, it is necessary to use RS and GIS tools because of their great opportunities provided [42]. These opportunities include (and not restrict to specific operations); processing, editing, and analysis of the required input factors for soil erosion mapping. Therefore, RS data and GIS tools play an interesting role in soil erosion monitoring through the exploitation of different satellite images and geospatial analyses ([43]. Therefore, for the present study, the RUSLE and SWAT models using RS and GIS tools were applied.

To the best of authors’ knowledge, no previous research has been carried out that explores the combination of these two approaches in this study area. Indeed, the novelty of this paper is the exploitation of the advantages of these two models. Therefore, the objective of this work is, (i) to map and to monitor the soil erosion degradation in the R’Dom watershed using SWAT and RUSLE models; and (ii) to predict the annual variation of soil losses due to the effects of climate change variations through the application of statistical downscaling method. We hereby envisage that our contribution will provide a more complete understanding of the soil degradation in this study area and could be a good reference for decision planning strategies.

2. Materials and Methods

2.1. Study Area

The R’Dom river watershed is located in the northwestern part of Morocco between longitudes 5.29° and 5.75° W and latitudes 33.47° and 34.01° N (Figure 1) covering an area of 1970 km². The watershed has a maximum elevation of 1778 m at the southern end and minimum elevation of 29 m at the outlet. The current study area is characterized by a semi-arid climate, with an average annual precipitation of 474.4 mm, with a dry period ranging from May to mid-September [44]. 44% of the rainiest seasons are winter, whereas 25% are spring, while the average annual temperature is 17.6 °C with January as the coldest month and August as the warmest month. In terms of hydrology, the basin consists of the main river, Oued R’Dom and its affluent rivers (Boufakrane, Ouislane, Chajara, Ishak, Sidi Ali, Frah and Khoumane). According to [45], the hydrogeological system of the watershed is composed of two main aquifers units, the deep aquifer and the Plio-Quaternary aquifer. From a geological point of view, the study area is essentially filled by a deep series of blue marls of Tortonian age followed by Pliocene fauvist sands and conglomerates and lacustrine limestone of Plio-Quaternary era. The dominant crops in the study area are rainfed cereals, sunflower, onion and olive trees (45% of total surface of the watershed) [46], in addition to forest and pasture lands. The main economic activity in this study area is agriculture, and the use of water resources has increased, leading to an unsustainable agriculture, hence contribute to land degradation in this watershed. Therefore, it is urgently needed to develop adaptation practices and strategies to cope with soil degradation.

2.2. Data Used and Methodology

In order to carry out the current research reported in this manuscript, we used several data given in Table 1. It should be mentioned that all data used are resampled to the Sentinel 2A spatial resolution (10 m) based on the nearest neighbor algorithm.

More details are given below.

Table 1. Data used in this research.

Data	Spatial Resolution	Coordinate System	Source
Digital Elevation Model (DEM)	30 m	UTM/WGS84 1, Zone 30 N	[47]
Land cover Map	10 m	UTM/WGS84 1, Zone 30 N	[47]
Soil data	30 m	UTM/WGS84 1, Zone 30 N	Field collection/reports
Climate data	30 m	UTM/WGS84 1, Zone 30 N	Sebou Hydraulic Basin Agency (SHBA)

¹ World Geodetic System (WGS84) datum and the Universal Transverse Mercator (UTM) projection system.

Table 1. Data used in this research.

Data	Spatial Resolution	Coordinate System	Source
Digital Elevation Model (DEM)	30 m	UTM/WGS84 1, Zone 30 N	[47]
Land cover Map	10 m	UTM/WGS84 1, Zone 30 N	[47]
Soil data	30 m	UTM/WGS84 1, Zone 30 N	Field collection/reports
Climate data	30 m	UTM/WGS84 1, Zone 30 N	Sebou Hydraulic Basin Agency (SHBA)

¹ World Geodetic System (WGS84) datum and the Universal Transverse Mercator (UTM) projection system.

2.2.1. Sentinel-2 Images

In this study, one sentinel-2A image acquired on 24 June 2016, composed of 13 spectral bands in the VNIR and SWIR with spatial resolutions ranging from 10 to 60 was used for producing Land use/Land cove map. The image was downloaded free of charge from website [47]. The image was acquired during the dry season due to minimal cloud coverage (less than 10%).

The image was preprocessed in QGIS software version 3.22 using the developed plugin Semi-Automatic Classification Plugin (SCP), developed by Luca Congedo [48]. The processing was done following various steps including: the conversion of the digital number (DN) to top-of-atmosphere reflectance (TOA) and the subsequent atmospheric correction by the dark object subtraction (DOS) algorithm [49]. Sentinel-2A/B images have been successfully used for several environmental applications including land use/land cover mapping [50].

2.2.2. Digital Elevation Model (DEM)

Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Model (GDEM) was downloaded from the U.S. Geological Survey website [51]. This digital elevation model was used to extract slope length (L) and slope steepness (S).

2.2.3. Soil Data

In order to determine the physical and chemical characteristics of the soil such as texture, apparent density, permeability, structure and organic matter, soil sampling was carried out in the field, and the physical and chemical analyses were performed in the laboratory. In this study area, the most dominant soils are calcimagnesic soils with a percentage of 24%, followed by raw mineral soils with 20%, vertisoils with 18%, poorly developed soils with 14%, fersiallitic soils 13% and isohumic soils 11%.

2.2.4. Climatic Data

Daily, monthly and annual rainfall data for the period from 1 September 2000 to 31 August 2013 for six stations located in the study area were used in this work. Then, the Statistical DownScaling Model (SDSM) has been used to predict future climate related to various Representative Concentration Pathway (RCP) scenario RCP 4.5.This method allows the downscaling of the most important climate variables, such as temperature, precipitation, evaporation, in order to assess hydrological responses under climate change scenarios [52].

2.3. Methodology Adopted

The estimation of soil erosion was performed using two approaches, the first is based on the application of the RUSLE equation and the second is based on the application of the SWAT model which is based on the MUSLE model for estimating sediment yields [53]. The climatic data were used to calculate the rainfall erosivity factor (R). The available geological data and the sampled soil properties were used to estimate the erodibility factor (K) and the map of soil classes was used in SWAT model. The two factors of slope length (L) and slope steepness (S) were estimated from the DEM, and calculated by the LS factor tool (LS-TOOL) developed by [54].The sentinel satellite image was used for producing land cover/land use and to estimate Normalized Difference Vegetation Index (NDVI) used for the elaboration of the cover and management map served as input for both RUSLE equation and SWAT model. All the used factors in this study were calculated using ArcGIS software with a pixel size of 10 m and a coordinate system (WGS84 Zone 30N). The methodology developed in this research for RUSLE equation and SWAT model is presented in Figure 2. To generate the soil loss maps for both the studies periods, using both RUSLE equation and SWAT model, natural break classification technique, available in ArcGIS 10.5, was used to classify the soil losses into 5 classes, named, very low, low, moderate, high and very high. All layers were developed in GIS environment with a spatial resolution of 10 m.

2.3.1. RUSLE Equation

The application of the RUSLE equation requires the preparation of several maps of the different factors involved: rainfall erosivity (R), cover and management (C), slope length and slope steepness (LS), soil erodibility (K), and support practice (P) according to the Equation (1):

$$A=R\times C\times L\times S\times K\times P$$

(1)

where:

• A is the computed average soil loss over a period selected for R, usually on yearly basis (t ha⁻¹ y⁻¹);
• R is the rainfall-runoff erosivity factor (MJ mm ha⁻¹ h ⁻¹ y⁻¹);
• K is the soil erodibility factor (t ha h ha⁻¹ MJ⁻¹ mm⁻¹);
• LS is the slope length (L) and slope gradient (S) factor (dimensionless);
• C is the cropping management factor (dimensionless, ranging between 0 and 1);
• P is the supporting conservation practice factor (dimensionless, ranging between 0 and 1).

A.

Rainfall and Runoff erosivity factor (R)

The R-factor represents the erosive force of a specific rainfall [55]. It is calculated from the rainfall energy E and the maximum rainfall intensity during a 30-min period (I30) using the Equation (2) developed by [56]:

R = E × I30 and E = 210 + 89 log10 × I

(2)

where:

• E: kinetic energy of rains (MJ/ha)
• I30: maximum rainfall intensity in 30 min mm/h
• I: rain intensity

The equation proposed by [56] is difficult to calculate, especially in areas where rainfall data are not regular or where detailed climatic data are not available. To overcome this limitation, many other empirical formulas are used by several authors especially for Mediterranean countries and in semiarid areas, where large uncertainties remain due to a lack of a high-resolution rainfall [57]. In their paper, Petroselli et al. [58] present 12 formulas, widely applied in Mediterranean areas. Therefore, in our study, this factor has been calculated using the Renard and Freimund equation which relates R (MJ mm ha⁻¹ h⁻¹ y⁻¹) to the average annual rainfall (P) (Equation (3)) [59].

$$\mathrm{R}=0.0483\times {\mathrm{P}}^{1.610} \mathrm{If} \mathrm{P} <850 \mathrm{mm}$$

(3)

where:

• R is the rainfall-runoff erosivity factor (MJ mm ha⁻¹ h ⁻¹ y⁻¹)
• P is the average annual rainfall (mm)

Table 2. Monitoring and prediction annual average rainfall and R Factor of R’Dom watershed.

			2000/2013		2014/2027
Station Name	Coordinate (m)		Average Rainfall	R	Average Rainfall	R
	X	Y
P3401	237,745	3,772,098	395.30	732.82	497.69	1061.82
P6170	234,726	3,726,750	488.72	1031.18	543.69	1224.24
P6405	301,106	3,741,223	529.02	1171.49	695.13	1818.33
P6769	283,324	3,775,316	467.92	961.44	501.92	1076.38
P6924	285,738	3,705,901	615.66	1495.50	792.65	2246.3
P7188	242,214	3,802,127	373.58	669.09	390.67	719.06

shows the R factor values for the different stations used in this study.

In this work, we used the inverse distance weighting (IDW) spatial interpolation method available in ArcGIS software version 10.4 for presenting the spatial distribution for the rainfall data used. Then, the Statistical DownScaling Model (SDSM) was used to predict future climate for the period 2014–2027 related to various Representative Concentration Pathway (RCP) scenario RCP 4.5. Rainfall data were used to calculate the R-factor based on both Equations (2) and (3).

B.

Soil erodibility factor (K)

The soil erodibility factor (K) is a factor that has an important influence in the RUSLE equation. It is directly related to soil and/or geological characteristics, such as parent material, texture, structure, organic matter content and porosity. This parameter was calculated using Equation (4) [56]:

$$K=\left[27.66 m^{1.14}\times {10}^{-8}\times \left(12- \mathrm{OM}\right)\right]+\left[0.0043\times \left(\mathrm{S}-2\right)\right]+\left[0.0033\times \left(\mathrm{P}-3\right)\right] \times 0.1317$$

(4)

where:

K is the soil erodibility factor ton ha hr ha⁻¹ MJ⁻¹ mm⁻¹,

m is the (silt % + sand %) × (100 − clay %),

OM is the % organic matter,

S is the structure code: (1) very structured or particulate, (2) fairly structured, (3) slightly structured, and (4) solid,

P is the profile permeability code: (1) rapid, (2) moderate to rapid, (3) moderate, (4) moderate to slow, (5) slow and (6) very slow.

Unfortunately, there is no detailed soil map available for the study area. In our case, we used the map generated by the Regional Institute of Agronomic Research (RIAR) to identify homogeneous units. A total number of 25 soil samples were taken from these units (0–25 cm deep), and the physical and chemical soil properties were carried out in the laboratory (NF P 94-056) to determine the main soil properties influencing the K factor such as, soil texture, organic matter, and soil structure. We computed the K factor as suggested by [56]. soil types in this study were determined according to Commission de Pédologie et de Cartographie des Sols (CPCS) [60]. The results of soil properties are summarized in

Table 3. Soil characteristics of study area.

Soil Type	Sand (%) (0.05–2 mm)	Silt (%) (0.02–0.05 mm)	Clay (%) (<0.02 mm)	Organic Matter (%)	K-Factor
Calcimagnesic Soils	7.3	37.7	55	2.16	0.031
Isohumic Soils	55.7	21.1	24	1.07	0.062
Raw Mineral Soils	8.8	34.5	56.1	2.62	0.025
Vertisols	53.8	19.3	27.1	1.62	0.052
Poorly developed Soils	63.7	20.8	15.5	2.51	0.062
Fersiallitic Soils	18.6	36.2	45.2	1.63	0.045

and Figure 3.

C.

Slope Length and Steepness factor (LS)

The LS factor is very important, affecting sediment production and transport, compaction and soil disturbance. For the generation of LS factor, we used DEM to calculate flow accumulation and slope steepness using the spatial analyst tool and ArcHydro availaible in ArcGIS 10.4.

The LS in this study was calculated using the Equation (5) developed by [61]:

$$\mathrm{LS}=((\frac{“\mathrm{FlowAcc}\_\mathrm{flow}”\times 30}{22.1}{)}^{\mathrm{NN}})\times (0.065+0.045 \times “\mathrm{Slope}” \times “\mathrm{Slope}”)$$

(5)

where NN represent dimensionless exponent that assumes the value of 0.2–0.

D.

Cover management factor (C)

The cover-management factor (C) is used to describe the effect of crop management practices on erosion rates. The C factor was determined based on the NDVI index following the Equation (6) developed by [62]:

$$C=0.431-\left(0.805\times NDVI\right)$$

(6)

The NDVI was developed for the first time by Rouse Jr. et al. [63], representing a good indicator for monitoring of vegetation activity. NDVI was calculated based on the reflectance between the infrared and red portions of the electromagnetic spectrum using the Equation (7):

$$\mathrm{NDVI}=\frac{\mathrm{NIR}-\mathrm{R}}{\mathrm{NIR}+\mathrm{R}}$$

(7)

where NIR is the near-infrared reflectance, and R is the reflectance in the red band. NDVI values varies between −1 and 1.

E.

Support practice factor (P)

According to [64], the factor P is defined as the ratio of soil loss with a specific support practice to the corresponding loss with up slope and down slope tillage. The value of P factor is determined based on the types of soil conservation measures. Thus, in this study, we determined the value of P factor based on the methodology recommended by Wischmeier and Smith [56]. For this purpose, sentinel 2A image was used to prepare the LULC map using the maximum likelihood classification algorithm. In total, eight LULC classes were established including: WATER (water), RNGE (Range-Grassess), FRST (Forest), WETN (Wetlands-non-Forsted), AGRL (agricultural Land-Generic), BARR (Barren), URHD (Residential-High Density), and UIDU (Industrial). Next, we reclassified the land cover/land use of the study area into two groups: “agricultural areas and other classes”. After that, we classified the agricultural areas into 5 classes based on the slope, and then we assigned for each class of agricultural a P value and for all other classes we assigned 1, as can be seen in

Table 4. P factor for different land use /land cover based on slope gradient [38].

Land Use Class	Slope (%)	P Factor Values
Agricultural area	0–4	0.5
	4–12	0.6
	12–20	0.7
	20–25	0.8
	>52	0.9
Other land	All	1.0

.

Table 5. Accuracy assessment for supervised classification of Sentinel image 2016.

	WATER	FRST	RNGE	WETN	AGRL	BARR	URHD	UIDU	User Accuracy
WATER	127	2	1	23	2	0	0	0	81.93
FRST	1	230	13	8	2	1	8	2	86.79
RNGE	0	35	159	3	19	2	0	8	70.35
WETN	8	0	2	56	0	0	2		82.35
AGRL	1	6	0	0	72	7	4	2	78.26
BARR	2	0	0	0	1	66	3	9	81.48
URHD	5	0	0	0	0	0	276	11	94.52
UIDU	2	7	0	3	5	18	26	54	46.95
Producer Accuracy	86.98	82.14	90.85	60.21	71.28	70.21	86.52	62.79	OA = 0.80% Kappa index = 76.8

presents the error matrix, along with the overall accuracy (OA) and the Kappa coefficient. The OA of the classification was 80% and the Kappa coefficient was 76.8%.

2.3.2. SWAT Model

The SWAT model is a semi-distributed, time-continuous hydrological model, generally applied at the catchment scale. It allows the simulation of surface and groundwater quality and quantity, and also allows the prediction of the effect of field management practices on sediment, water and agricultural chemicals in large complex catchments with soils [65]. In addition, this model also allows the calculation of the simulated daily water balance based on the Equation (8) [65]:

$$SWt=SW0+{{\displaystyle \sum }}_{i=1}^t{(R-Qsurf-Ea-Wrchrg-Qlat)}_i$$

(8)

where:

• SWt is the final soil water content (mm),
• SW0 is the initial soil water content (mm),
• t is the time (days),
• Rday is the amount of precipitation on day i (mm),
• Qsurf is the amount of surface runoff on day i (mm),
• Ea is the amount of evapotranspiration on day i (mm),
• Wseep is the amount of water entering the vadose zone from the soil profile on day i (mm),
• Qlat is the amount of return flow on day i (mm).

The estimation of soil erosion rate by SWAT was based on the application of MUSLE (Equation (9)). It is a modified equation of USLE of which the rainfall energy factor is replaced with a runoff factor and uses the soil characteristic to calculate the runoff erosive energy variable. For the land use factor, it was performed from the processing of the sentinel-2 image by a supervised classification approach. The DEM was use to derive the factors relating to the morphology and the soil map was ranked according to the physical characteristics such as soil texture, humidity or water availability capacity (AWC), soil depth, rock fragments, organic material (OM), hydraulic conductivity (K factor) and moist bulk density. In our case, we used Arc-SWAT to compile the SWAT input factors.

$$\mathrm{Sed}=11.8 {\left({\mathrm{Qsurf} \times \mathrm{q}}_{\mathrm{peak}}{ \times \mathrm{area}}_{\mathrm{hru}}\right)}^{0.56} \times \mathrm{K} \left(\mathrm{usle}\right) \times \mathrm{P}\left(\mathrm{usle}\right) \times \mathrm{C}\left(\mathrm{usle}\right) \times \mathrm{LS}\left(\mathrm{usle}\right) \times \mathrm{CFRG})$$

(9)

where:

• Sed is the sediment yield on a given day (tons),
• Qsurf is the surface runof volume (mm H₂O/ha),
• qpeak is the peak runof rate (m3 /s),
• Areahru is the area of the HRU (ha),
• KUSLE is the soil erodability factor
• CUSLE is the cover and management factor,
• PUSLE is the support practice factor,
• LSUSLE is the topographic factor and CFRG is the coarse fragment factor

2.3.3. Spatial Autocorrelation Analysis

The spatial autocorrelation represents the spatial dependence between numerical values of a spatially located variable. It is used to assesses spatial autocorrelation using both feature locations and feature values at the same time. The most commonly autocorrelation index used is the Global Moran’s [66]. It determines whether the pattern displayed is clustered, scattered, or random based on a collection of characteristics and an associated attribute. The Global Moran’s I range from −1 to 1. If the absolute value is more approaching 1, it means it has a stronger autocorrelation [66]. In this study, the Global Moran’s I was used to characterize the spatial relationships between soil erosion risks for both RUSLE equation and SWAT model.

3. Results and Discussion

3.1. Stream Flow

To test the applicability of the SWAT model, we based on the comparison of the monthly hydrograph and simulated flows at the Souk Elhad station. During the calibration for the period of 2002 to 2009, we used the sequential uncertainty fitting version 2 (SUFI-2) algorithm.

Our results showed that the SUFI-2 gives good results with −1.1% for the Pbais factor and the high values of R² and the NSE larger than 0.75 and 0.70 respectively. The sensitivity analysis results of the model to each sub-basin delineation and HRU in this basin are 1613 HRUs in the whole watershed. Regarding the uncertainty results, the 95% prediction uncertainty (95PPU) was well correlated with the observed flow. Figure 4 shows the comparison of observed and simulated monthly flows for the Souk Elhad station for the calibration period. it is was showed that both the simulated and observed flows are well correlated.

3.2. Factors Maps Analysis

Figure 5a, b shows the spatial distribution of R factor covering the study area during the two periods 2000–2013 and 2014–2027. The results show that this R-factor ranged from 673 to 1480.6 MJ·mm/(ha·hr·y) for the first period and from 729.2 to 2214.6 MJ·mm/(ha·hr·y) for the second period.

The results of LS factor (Figure 6a) show that the watershed is relatively flat with LS values range between 0 and 15%. The most part of the study area has low LS factor values, the relatively high values of the LS factor (>15%) are mostly located in the eastern and western and the north parts of the basin. The range of elevation in the study area is from 29 to 1778 m (Figure 7a).

The spatial distribution of soil erodibility factor is given in the Figure 6b. The soil erodibility factor was found in the range of 0.003 to 0.0081. The major soil groups in the watershed are presented in Figure 7d.

The Figure 6c shows the spatial distribution of C factor with values ranging from 0.016 to 0.138. C factor values close to 0 indicate well-protected land cover and good conservation efforts, while values close to 1 indicate urbanized area and barren land, as well as agricultural fields that are exposed to heavy rainfall. Therefore, the results of this factor show that the southern and northern parts are most affected. In this study, Land Use/Land Cover (LULC) map produced is shown in Figure 7c.

The P factor (Figure 6d) value ranged from 0.55 to 1 where a higher value indicates there is no any support practice such that erosion is at its maximum due to the absence of any practice.

3.3. Spatial Distribution of Soil Erosion Rate

3.3.1. Soil Erosion Rate Using RUSLE Method

The spatial distribution maps of soil erosion using RUSLE method during the period 2000/2013 and 2014/2027 are presented in Figure 8. The areas of soil erosion classes are given in

Table 6. Soil erosion classes with area occupied using RUSLE model.

	2000–2013		2014–2027
Soil Erosion Severity Class	Soil Loss (t ha−1)	Area (%)	Soil Loss (t ha−1)	Area (%)
Very low	<10	71.90	<10	58.33
Low	10–20	20.20	10–20	27.71
Moderate	20–30	6.25	20–30	11.94
High	30–40	1.42	30–40	1.73
Very high	>40	0.20	>40	0.27

.

The results obtained for the 2000/2013 period showed that 71.90% of the total area of the watershed is exposed to a very low erosion, whereas only 0.20% of the area was exposed to a very high-risk erosion, and the results obtained for the period of2014/2027, showed that the study area is expected to have a very low erosion risk for a portion of 58.33% of the total area, and a very high risk for 0.27%.

It can be seen that the risk of erosion is very low in the basin, especially in the central part, which belongs to the Saiss plain characterized by down slope, and the rate of erosion is relatively high in the areas characterized by heterogeneity of relief, located especially in the north parts of the study area.

3.3.2. Soil Erosion Rate Using SWAT Model

The spatial distribution maps of the soil erosion using SWAT model during the period 2000/2013 and 2014/2027 are presented in Figure 9. The areas of soil erosion classes are given in

Table 7. Soil erosion classes with area occupied using SWAT model.

	2000–2013		2014–2027
Soil Erosion Severity Class	Soil Loss (t ha−1)	Area (%)	Soil Loss (t ha−1)	Area (%)
Very low	<10	35.57	0–10	36.97
Low	10–20	28.28	10–20	25.22
Moderate	20–30	18.45	20–30	18.35
High	30–40	8.7	30–40	10.21
Very high	>40	9.00	>40	9.24

, The results obtained for the period of 2000/2013 showed that 35.57% of the total area of the watershed is exposed to a very low erosion, whereas only 9% of the area was exposed to a very high risk erosion, and the results obtained for the period 2014/2027 showed that the study area is expected to have a very low erosion risk for a portion of 36.97% of the total area, and a very high risk for 9.24%.

It is important to highlighted here that the affected areas by soil erosion are expected to be affected in the future in the north and south parts of the watershed. This is strongly linked to the morphology of the watershed and the slope factor. The lowest total soil erosion is observed in the center parts of the watershed, characterized by down slopes.

3.4. Soil Erosion Rate in Relation to Land Use

Using the ‘Zonal Statistics’ toolset and cross-tabulate (available in ArcGIS 10.4), the relation between soil erosion rate and LULC for both RUSLE and SWAT was established. From the examination of

Table 8. Soil erosion rate in relation to land use using RUSLE equation.

	Area (%)	Very Low	Low	Moderate	High	Very High
Water	0.15	0.09	0.36	0.17	0.10	0.00
forest	8.01	7.66	5.53	8.58	10.74	7.53
Range-Grasses	0.1	0.19	0.14	0.13	0.06	0.00
Wetlands-Non-Forested	0.00	0.00	0.00	0.00	0.00	0.00
Agricultural Land-Generic	48.1	67.45	61.06	48.45	35.89	27.65
Barren	38.27	15.25	26.19	38.01	50.12	61.80
Residential-High Density	4.83	8.83	5.90	3.85	2.55	3.01
Industrial	0.51	0.48	0.79	0.78	0.51	0.00

and

Table 9. Soil erosion rate in relation to land use using SWAT model.

	Area (%)	Very Low	Low	Moderate	Very High	High
Water	0.12	0.05	0.01	0.41	0.05	0.09
Forest	7.33	4.67	14.95	3.93	5.82	7.30
Range-Grasses	0.06	0.11	0.04	0.15	0.01	0.01
Wetlands-Non-Forested	0.00	0.00	0.00	0.00	0.00	0.00
Agricultural Land-Generic	59.58	75.47	68.26	66.08	56.03	32.05
Barren	25.26	11.01	8.43	19.18	34.35	53.32
Residential-High Density	7.30	8.39	8.01	10.09	2.90	7.03
Industrial	0.33	0.27	0.27	0.12	0.82	0.17

, it was verified that erosion loss was found mainly in agricultural lands (AGRL) with an area of 48.1% and 59.58%, Followed by bare lands (BARR) with 38.27% and 25.26% in RUSLE equation and SWAT model respectively. These findings are in line with the results of previous studies [67].

3.5. Spatial Autocorrelation of Soil Erosion Rate

The spatial autocorrelation analysis report for both RUSLE and SWAT methods were obtained using the spatial autocorrelation analysis tool in the ArcGIS 10.4 toolbox. In addition to the global Moran’s index, the analysis results in the report also include z-scores and p-values, to determine the index’s significance. P-values are numerical estimates of the area under the curve given a known distribution, using the test statistic as a constraint. The results of the Moran’s I index are shown in

Table 10. Result of Moran’s I index.

Moran’s I	RUSLE Equation	SWAT Model
Moran’s Index	0.12	0.32
Variance	0.00	0.00
z-score	125.37	731.57
p-value	0	0

and Figure 10. According to the results, the value of Moran’s I index for the RUSEL equation and SWAT model was 0.12 and 0.32, respectively. According to these results both the models present a positive correlation at a spatial distribution. p-value is 0.000 (it is lower than 0.01), which indicates that the random distribution probability is less than 1%. Furthermore, z-score is higher than 2.58, which demonstrates that the spatial aggregation trend of the erosion loss at global distribution is significant.

The phenomenon of erosion is a serious problem that requires intervention, especially in agricultural areas, and its management is necessary for sustainable land use and overall soil conservation management. RS techniques and GIS tools are integrative and generative tools for prevention, warning, monitoring and modelling of environmental hazards. In the current study, the methodology applied showed a strong capacity in estimating soil losses in the R’Dom watershed. Our results are quantitatively and spatially quite similar between the two methods. As we found out, high risk erosion is mainly concentrated in the agricultural land and barren lands, where erosion is very high. These results are in line with previously published studies [68,69].Thus, further policy interventions are therefore very urgent.

In summary, the results of our study showed that the study area is not largely affected by soil erosion. These is also confirmed by other authors in other areas with similar background in Morocco, for example the results of a study conducted by [70] showed that the rate of soil loss was 7–20 t ha⁻¹, accounting for 75% of the basin area. Likewise, the rate of soil loss was estimated at 3.95 t ha⁻¹ at the M’dez Watershed when calculated by the RUSLE model, and the results of SWAT method showed that the soil loss is estimated at 2.94 t ha⁻¹ according to Boufala et al. [38]. Conversely, Tahiri et al. [71] indicated that the soil loss reached 47.18% in the Oued Haricha Sub-Basin, Western Rif. Thus, in terms of the efficacity of the used approaches and the outcomes obtained, both these methodologies could be useful in other areas where the soil erosion is a heavy issue.

4. Conclusions

This study proposes an approach to estimate the mean annual soil erosion in the R’ Dom watershed Morocco. In this way, for the first time, the RUSLE equation and SWAT model were used based on the use of GIS and remote sensing through the exploitations of open source and free of charge data. The main results of our study are listed below.

-

Standard calibration statistics were used to assess the performance of SWAT model. Comparison of modeled and observed monthly streamflow datasets resulted from R2, NSE, and PBIAS values of 0.75, 0.70, and −1.1 respectively, which indicated that the hydrological cycle of the R’Dom watershed could be accurately simulated using the SWAT model.

-

The erosion rate values vary from 0 to 8 t ha⁻¹ for both periods. For the 2000/2013 period, the RUSLE showed that 71.90% of the total area of the watershed is exposed to a very low erosion, whereas only 0.02% of the area was exposed to a very high erosion, while the SWAT showed that 35.57% is exposed to a very low erosion, whereas only 9% of the area was exposed to a very high erosion. For the period of 2014/2027 it is expected to have a very low erosion risk for a portion of 58.33% and 36.97% of the total area, and a very high risk for 0.27% and 9.24% for RUSLE method and SWAT respectively.

-

It was verified that erosion loss was found mainly in agricultural lands (AGRL) with an area of 48.1% and 59.58%, Followed by bare lands (BARR) with 38.27% and 25.26% in RUSLE equation and SWAT model respectively.

This study demonstrates that GIS provides a flexible environment for spatial analysis and data manipulation due to its ability to process large amounts of spatial data necessary for erosion studies, also RS imagery is the most important data resources of GIS, providing a spectrum of imaging capabilities, resolutions, temporal and spatial coverage. RS plays a pivotal role in achieving these objectives. Furthermore, the results obtained from this study can provide a valuable assistance, at very low cost, to decision-makers in simulating evolution scenarios and then targeting priority areas that require conservation and erosion control actions. Finally, the validation of the estimated soil erosion rates of this study area can be further done with field sampling data, which are the main gap of this work.