RECENT ADVANCES IN THE ASSESSMENT OF SOIL EROSION VULNERABILITY IN WATERSHEDS

: Water induced soil erosion has always been a matter of concern in watersheds as they increase the soil vulnerability towards erosion. If unchecked, the eroded material reduces the capability of the river to carry the adequate amount of water and increase the amount of sediments in the watershed area. Determining vulnerability of soil to erosion plays a key role in identifying the extent of fragility and helps in making appropriate plans for conservation. Among various methods present to assess soil erosion vulnerability, there is a need to under stand the frequently used methods so far and its advancement with time. Various models have been used in past two decades (1991-2019) and the Revised Universal Soil Loss Equation (RU -SLE) is the most used model because of its quantitative ability to estimate the average annual soil loss due to erosion in a watershed and its compatibility with the GIS interface. Different ap proaches like MCDM, SWAT etc. are being utilised to study soil erosion vulnerability of water sheds. This review showed that the frequently used MCDM method is a Compound Factor (CF) method and that RUSLE is a most used quantitative approach. The review identifies 14 differ ent methods which includes 4 methods which provide quantitative estimation while the other 10 methods are used for qualitative assessment of soil erosion vulnerability. Being the most adopted approach, various modifications of different factors of RUSLE introduced by research ers have made it more efficient with time. This review identifies the trend in advancement of various approaches and methods to study soil erosion vulnerability of watersheds around the world and also how various studies are distributed in the Himalayan and non-Himalayan region. The review also provides an understanding of the status of various current approaches to study soil erosion in a watershed and lists the improvements adopted in the frequently used approaches during 1991 and 2019.


INTRODUCTION
Watershed comprises different land uses and water bodies. Vulnerability of watersheds towards soil erosion is triggered by a combination of factors such as steepness of slope, climate, inappropriate land use & land cover patterns (e.g. sparse vegetation) and ecological disasters (e.g. forest fires) (Pa r veen et al., 2012) and depends on erosivity, erodibility and land use management practices. Soil erosion vulnerability plays a key role in identifying the extent of fragility and making appropriate plans for conservation of a watershed. Soil erosion has both on-site and off-site detrimental impacts and is one of the most critical environmental hazards as it adversely affects both the economy and environment.
To understand soil conservation and ecosystem management mechanisms in a watershed, soil erosion evaluation and mapping of soil loss susceptible area is required (G e l a gay and M i n a l e , 2016). There are several methods used to access a soil erosion susceptible region. These methods are based on various parameters like land use, soil quality and topography etc. So far, water quality being a strong indicator (G h o s h et al. 2013) of soil erosion is often neglected in conducting susceptibility studies. The relationship between soil erosion vulnerability and land use allows the identification of more susceptible areas to erosion and the need for implementation of soil management and conservation practices to reduce soil erosion vulnerability. The assessment of the annual soil erosion rate and development of a soil erosion map (Zhou et al., 2014) for a watershed in various studies have provided spatial patterns of classified soil erosion risk zones indicating areas with high, severe and low erosion risk area. Topography plays an important role in controlling soil movement in a watershed and areas mostly covered by high fraction vegetation are at a lower level of soil erosion risk (Prasannakumar et al., 2011). Unsystematic land use pattern has a certain impact on soil and water resulting in its erosion and deteriorating its stream quality. The eroded materials carried down to the lower reaches of the rivers make them incompatible to carry excess amount of water and sediment load during a monsoon period (G h o s h et al., 2013).The protective effect of land cov-er leads to demotion of vulnerability categories (Stathopoulos et al., 2017).
In watersheds, soil erosion is a matter of concern as it is likely to have an impact on water quality along with soil degradation. Study of soil erosion in various regions has increased in the past decade. A wide variety of methods have been adopted by different researchers to study and model soil erosion. The objectives of this paper are a) to understand advances in the study of soil erosion vulnerability around the globe, b) to identify the most frequently used methods to assess soil erosion vulnerability under different circumstances, in the past two decades, c) to know the proportion of studies conducted in the fragile region of Himalayas and d) to understand how water quality has been included in research related to the assessment of the soil erosion vulnerability factor.

METHODOLOGY
This paper attempts to understand the advancement in use of several models in assessing soil erosion vulnerabilities in the last three decades . Since the Himalayan region is one of the most fragile ecosystems because of its steep slope, poor soil and heavy monsoon rains, it is important to study the fragility of such an ecosystem. In this view, it becomes necessary to be familiar with the frequency of previous studies conducted in this region and the methods employed globally to assess such an ecosystem in order to plan on future strategies. For the selection of related academic papers in past two decades (1991-2019) on related subjects web search engine (Google scholar) and publishers' websites were used. To obtain papers on soil erosion vulnerability the primary key words used were: soil erosion vulnerability assessment, Himalayan region, RUSLE for the period 1991-2019 in three different sets and combinations. The papers including any qualitative or quantitative method to assess soil erosion vulnerability were selected. These papers come from journals published worldwide to cover the work done across the globe. These studies were further searched on the basis of filter for a decade and were classified into three time periods, namely, 1991-2000, 2001-2010 and 2011-2019. The papers obtained from both these searches were compared and repetitions were removed. Among the top searches which were directly related to the subject, 130 research papers were selected to further understand the approach used to assess soil erosion vulnerability in the past two decades (1991-2019) and distribution of the studies conducted in the Himalayan and Non-Himalayan region. Further, each study was categorised according to the used methods to identify the most popular methods in the past two decades.

ADVANCEMENT IN VARIOUS APPROACHES FOR ASSESSING SOIL EROSION VULNERABILITY
Various models have been used for assessing soil erosion vulnerability such as Revised Universal Soil Loss Equation (RUSLE), Water Erosion Prediction Project (WEPP), Soil & Water Assessment Tool (SWAT), Erosion Potential Method (EPM), Principal Component Analysis (PCA) and Multi Criteria Decision Method (MCDM)-based qualitative assessment methods like Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS), Compound Factor (CF), Višekriterijumska Optimizacija i Kompromisno Rešenje (VIKOR), with morphometric, soil and water quality parameters as input. The results showed that RUSLE/USLE models are the most used methods to assess soil erosion vulnerability. Studies related to soil erosion vulnerability using different methods ( Figure  1) have increased in each decade since 2000. In models, different parameters like land use and land cover, morphometric parameters, soil quality, and combinations of these parameters have been used to study the soil erosion vulnerability. It is evident from the results that no study among 130 studies has incorporated all three parameters i.e. soil quality, water quality and land use together to study soil erosion vulnerability ( Figure 2).
Over the years, the number of studies more than doubled from 2011 to 2019 (155) in comparison to the period 2000-2010 (41) (Figure 1(a)). In the past three decades, the regions of study distribution were mainly non-Himalayan regions (75.38%) and the Himalayan region being one of the most fragile ecosystems accounts for only 24.62 percent of studies. The review also suggested that RUSLE is the most frequently used model having been used in 51.4 percent of the studies (Figure 3), while all other methods together account for 48.6 percent of the studies.
A total of fourteen methods were used so far among which ten methods provide qualitative status of soil erosion vulnerability and four methods (RUSLE, SWAT, EPM and WEPP) provide a quantitative estimation of soil erosion. Both qualitative and quantitative methods involve various parameters like topography, land use, soil quality etc, whereas water quality parameters were not considered as indicators. Methods used to assess soil erosion will be briefly discussed in the following section.

Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS):
This method was first introduced by H wa n g and Yo o n (1981) and Am eri et. al., 2018 used it to study the soil erosion susceptibility in a watershed (Table 10). This method shares 0.7% of the total studies reviewed being among the least used MCDM methods so far ( Figure 3). It is a distancebased method and the main source of calculation is based on a positive ideal solution (PIS) and negative ideal solution (NIS) for identifying decision making alternatives. In the model, the preferred alternative is the one that has the least distance from the positive ideal solution (PIS) and a higher distance from the negative ideal solution (NIS). The results of the two distances are expressed in the form of a closeness coefficient. The highest closeness coefficient indicates the most preferred alternative (L i o u and Wa n g , 1992); (Ka n n a n et al., 2009). In this method, different parameters like morphology, geology, slope; soil quality and land use can be taken depending upon the data availability and behaviour of the parameter. This method has been used less frequently in the past decades (2000-2010) but is a useful method for a qualitative ranking of watersheds with respect to soil erosion (Figure 2 & 3).

Višekriterijumska optimizacija i kompromisno rešenje (VIKOR):
VIKOR is a very well known MCDM technique which emphasizes the selection and ranking of an alternative set of conflicting criteria (O p r i co v i c and Tze n g , 2004). It was first introduced by (Opricovic and Tzeng , 2004). The VIKOR is a less used method in assessing and ranking sub watersheds (Table 10). In this method, normalized decision matrix based on criteria and their behaviour towards the alternatives and weighted decision matrix is computed. The advantage of this method is that the evaluation of all the criteria does not require expert review; but, raw data can also be used for assigning weights to the criteria. Methods

Compound Factor (CF)
This is the most commonly used MCDM methods with a varying number of parameters. This model provides a comparative estimation driven by scientific knowledge and understanding of a qualitative phenomenon (To d o ro v s k i and D že ro s k i , 2006). This is the most used MCDM method so far (9.5%) as shown in Figure 2 & 3 in which the total number of ranks assigned is based on the number of options. In this average of the ranks of all the parameters is designated as a compound value and represents the collective impact of all the parameters (A l taf et al., 2014). This method is flexible with a number of parameters selected for the study and weight assignment and if required can be performed based on expert review. The drawback of this method is that it has no provision for the normalization of variables of different scale and sizes and also, in case of same values, an assignment of the same rank can lead to over or underestimation of the state.

Principal component analysis (PCA):
Principal component analysis (PCA) is a powerful multivariate statistical technique to segregate parameters contributing to observed criteria to assess soil erosion vulnerability in a sub-watershed (K h a l e d i a n , 2016). It may be employed to explore the most influential parameters of different criteria like morphometric parameters, soil and land use parameters etc. based on the parameters which are highly correlated with important components. Further, this can be used to rank sub-watersheds to assess soil erosion vulnerability. This method has been used in few studies (1.4%) as shown in Figure 2 & 3 and is flexible for criteria selection, but requires a large number of parameters. Farhan (2017) has suggested that prioritization of watershed based on the Compound Factor is more consistent as compared to the PCA approach.

FUZZY:
The Fuzzy method can provide an optimum solution in which the uncertainties associated with evaluating criteria while assessing soil erosion vulnerability status using various multi cri-  teria approaches. The prioritization by the fuzzy analysis technique can be performed using various related criteria like morphometrics, land use etc. or even single criteria with different parameters. According to Chang's extent analysis of the FAHP method (C h a n g , 1996), each criteria can be evaluated through the formation of pairwise comparison matrix-based on the fuzzy linguistic scale and weightage could obtained through the normalization of fuzzy measures. The prioritization of each sub-watershed can be carried out on the basis of an FAHP analysis score where the first rank will be assigned to the sub watershed having the highest analysis value which will indicate the most vulnerable zone. Fuzzy techniques are mostly useful for carrying out ranking with overlapping parameters and there is a lot of scope for the application of this method for the ranking of sub watersheds (Table 10). This method is used in 5.4 percent of the studies reviewed ( Figure 3).

Water Erosion Prediction Project (WEPP):
WEPP (Water Erosion Prediction Project) are process-based models for runoff and soil erosion prediction (Laflen et al., 1997). Similar to the empirical models such as Revised USLE (RUSLE), the models have been widely used to model agricultural land (Renard et al., 1997). WEPP predict soil loss in a range of environments, e.g. rangeland and forest. WEPP model is capable of simulating runoff and sediment yield from the untreated watershed with good accuracy. It can be used to control soil loss and runoff by formulating structure based management strategies for watersheds. The WEPP software includes an erosion prediction model, a climate generator program and a Windows interface (Flanagan et al., 2007). Data required for this model are climate files, soil input and practices and management scenarios. This model requires parameters like the amount & duration of rainfall, maximum & minimum temp., solar radiation, organic carbon, texture and land use management practices in the watershed. This model has been used in only 2.7 percent of the studies (Figure 3 and Table 10), which may be due to its non-GIS interface and specific data requirements.

Soil and Water Assessment Tool (SWAT)
The Soil and Water Assessment Tool (SWAT) is a river basin model developed by the United States (US) Department of Agriculture in collaboration with Texas A&M University. SWAT version 2012 has been released in combination with ArcGIS (version 10.4) and ArcSWAT interface. Within the Geographic Information Systems (GIS) environment, SWAT is a distributed modelling as a watershed is delineated into sub-basins and subsequently into hydrologic response units (HRUs), which represent homogeneous combinations of land use, soil types, and slope classes in each sub basin. The physical processes associated with water and sediment movement, crop growth, and nutrient cycling are modelled at the HRU scale to assess the runoff generated from streams (Table  10). SWAT provides two methods for surface runoff estimation. The first one is based on the Soil Conservation Service curve number and the second one estimates runoff height using the Green and Ampt infiltration method. SWAT calculates the surface erosion caused by rainfall and runoff within each HRUs with the Modified Universal Soil Loss Equation (MUSLE) (Equation 2) (Williams, 1975). Soil erosion caused by rainfall and runoff is computed by the Modified Universal Soil Loss Equation (MUSLE).

sed = 11.8 × (Qsurf × qpeak × areahru)0.56× K USLE × C USLE ×P USLE × LS USLE ×C FRG
Where: sed is the sediment yield on a given day (metric tons); Qsurf is the surface runoff volume (mm H 2 O ha -1 ); qpeak is the peak runoff rate (m 3 s -1 ); areahru is the area of the HRU (ha);K USLE is the USLE soil erodibility factor; C USLE is the USLE cover and management factor; P USLE is the USLE support practice factor; LS USLE is the USLE topographic factor, and C FRG is the coarse fragment factor This model simulates hypothetical, real and future scenarios and has been proven an effective method of evaluating alternative land use effects on runoff, sediment and pollutant losses. SWAT model is one of the appropriate watershed models for long-term impact analysis. This method is among the most used (4.5%) quantitative methods used worldwide (Figure 2 & 3).

Erosion Potential Method (EPM)
The Gavrilović method (Erosion Potential Method, EPM) is an empirical, semi-quantitative model which has been extensively applied to erosion and torrent-related problems in the Balkan countries. The method encompasses erosion mapping, sediment quantity estimation, and torrent classification. The method does not explore physics of erosion processes and it is difficult to predict it efficiently with minimal data and lack of previous erosion research (Ko s t a d i n o v et  al., 2014). The outputs of this model are based on the multiplication of the model and categorisation of the model parameters such as Average annual temperature, Average slope of the study area, and Drainage density etc. Not all parameters are included in the model through multiplication, e.g., most of these parameters are categorized as high-or medium-sensitivity, whereas those in the multiplication form are classified as very high-sensitivity parameters. This method has been widely used in the European continent making it the second (8.1%) most used quantitative method ( Figure  2&3). The following table represents the parameters used in this method by (G av r i l ov i ć et al., 2005) and also by other researchers provided in Table 10.

W=T×H×F×π√Z3
(1) Where, W is the total annual erosion (m 3 /year), T is the temperature coefficient, H is the mean annual precipitation (mm), Z is the erosion coefficient, and F is the basin area (km 2 ) The temperature coefficient (T) is calculated by Equation (2): T=√t10+0.1 2) The soil erosion coefficient (Z) can be calculated from the following Equation (3): Where, X is the soil protection coefficient which reflects the type of landuse, Y is the coefficient of soil resistance which depends on soil and geology, φ is the erosion and stream network developed coeffi-cient that includes the type and extent of erosion, and I is the average slope(%) of the watershed.

Revised Universal Soil Loss Equation (RUSLE)
Being the most used model, RUSLE has been modified by various researchers retaining the basic soil loss equation (B e n av i d ez et al., 2018). This method has been used in 51.4 percent of the studies is the most promising method for quantitative soil erosion vulnerability assessment. RUSLE is a straightforward and empirically based model that has the ability to predict long term average annual rate of soil erosion on slopes using data on rainfall pattern, soil type, topography, crop system and management practices (P ra s a n n a ku m a r et al. 2011). Although it is an empirical model, it not only predicts erosion rates of ungauged watersheds using knowledge of the watershed characteristics and local hydro climatic conditions, but also presents the spatial heterogeneity of soil erosion that is too feasible with reasonable costs and a better accuracy in larger areas (A n g i m a et al. 2003). RUSLE has been widely used for both agricultural and forest watersheds to predict the average annual soil loss by introducing improved means of computing the soil erosion factors (Wischmeier and Smith, 1978;Renard et al., 1997). Digital elevation model (DEM) along with remote sensing data and GIS can be successfully used to enable rapid as well as detailed assessment of erosion hazards (J a i n et al. 2001al. , Ko u l i et al. 2009). The emergence of soil erosion models has enabled the study of soil erosion, especially for conservation purposes, in an effective and acceptable level of accuracy. In order to estimate soil erosion and to develop optimal soil erosion management plans USLE/RUSLE has been widely applied worldwide to predict soil loss because of its convenience in application and compatibility with GIS (Millward and Mersey, 1999); (Jain et al., 2001) (1) Where, A = soil loss (t ha -1 yr -1 ) R = rainfall erosivity (MJ mm ha -1 h -1 yr -1 ) K = soil erodibility (t h MJ -1 mm -1 ) LS = topographic factor (dimensionless) C = soil use and management factor (dimensionless) P = soil conservation practice factor (dimensionless)  Where, X is the average annual rainfall in mm and P is monthly precipitation in mm, while P 2 is annual precipitation Brief of each factor as follows:

• Rainfall Erosivity (R)
It is one of the factors that used to quantify the soil erosion and it is the potential ability to erode. Rainfall erosivity represents the potential of rain to cause erosion in an exposed and unprotected soil surface, whose physical definition is the product of rainfall kinetic energy and the maximum rainfall intensity in a 30-minute consecutive (EI 30 ) (Zhang et al., 2013).

• Soil erodibility factor (K):
It is a measure of the erodibility of soil. The soil erodibility factor (K) relates to the rate at which different soils erode. The factor is rated on a scale from 0 to 0.7, with zero indicating soils with the least vulnerability to erosion and those with 0.7 as most vulnerable. According to various researchers the following methods have been adopted to calculate this factor:

• Topographic factor (LS)
It is a factor of slope length (L) and slope gradient (S). According to various authors (Table  4) different equations has been used in the model. Where, λ is the field slope length (m), and m assumes a value between 0.2 and 0.5 sl is slope length of the site (m) and S is the slope factor, A is the upslope contributing factor, B is the slope angle, β ® is the land surface slope in degrees, m and n constants equal to 0.6 and 1.3, A® is up-slope contributing area per unit width of cell spacing (m 2 m −1 ).The digital elevation model (DEM) can be used to obtain the accumulated flow and slope map

• Crop Cover Management Factor (C)
Crop cover management factor is the ratio of soil loss from land with specific vegetation to the corresponding soil loss from a continuous fallow land (Wischmeier and Smith, 1978). To derive the crop factor, imagery can be downloaded from relevant websites, and further the C factor value can be derived as per table provided in Table 6.  Literature (Table 6)

• Conservation Practices Factor (P)
It is the ratio of soil loss with specific support practice to the correspondence soil loss with upslope and downslope cultivation. The value of practice factor will be assigned according to the table given below. For determining annual soil loss with the help of Arc GIS tools, downloaded digital elevation model and classified land use and a land cover map, each factor of the RUSLE equation can be generated in the form of a raster image. After combining the entire raster image of five factors using the Arc GIS tool, soil erosivity map can be made and annual soil loss can be calculated for each sub-watershed sensing. After calculating annual soil loss, soil erosion vulnerability assessment can be done on the basis of a category defined in literature by different researchers as provided below:

Other land
All 1 Fallow land 0.9 Bare soil/barren land Apart from the above mentioned methods, researchers have used single components and combination of components such as morphometry, land use, water quality and soil quality of watersheds. Based on these components watersheds have been prioritised using a combined component i.e. land use & water quality; land use and soil quality and a few studies have calculated it based on only one components such as morphometric parameters, land use and land cover ( Table  9). A general approach is to assess soil erosion vulnerability using the Revised Universal Soil Loss Equation (RUSLE) model inclusive of land use and by and discretely analyse the effect of land use on water quality. The existing approach ( Figure 5) for understanding the influence of land use includes study on soil erosion and water quality separately, which does not establish a relation between soil erosion vulnerability and water quality, although it has been reported by various studies that water quality is a strong indicator of soil erosion. Therefore, an advanced approach is required to manage watershed that could be taken into consideration which could analyse the link between soil erosion vulnerability and its relation with stream quality in a watershed. This advanced approach will provide assistance in developing better watershed management plans ( Figure 5).

CONCLUSION
There is need to recognise sites which are prone to erosion in watersheds. There are many methods adopted worldwide to implement the best management practices and suggest mitigation measures to overcome problems i.e. flooding landslides and sedimentation. As there is lot of scattered information on methods which are being used to assess soil erosion vulnerability, it is important to understand the methods which are being used and improved with time. The review suggested that so far RUSLE as a quantitative method has been the most widely used method to study soil erosion, yet other methods like MCDM SWAT and FUZZY etc. are also advancing with time. It was also observed that most of the less frequently used methods provide a qualitative output and researchers should encourage the use of such methods to identify erosion prone regions. Because of its compatibility with ArcGIS, RUSLE has emerged as the most accepted quantitative model to estimate soil loss so far. All factors of the RUSLE mod-el have been modified to make it more convenient and user friendly with time, since many studies on water quality and soil erosion vulnerability under different land use have been reported separately using the RUSLE model and water quality analysis. Presently there is a lack of studies which establish the relationship between soil erosion vulnerability and water quality, since water quality acts as an indicator for the prioritization of a watershed. The determination of such relationships could be suggested as an advanced approach. A discrete effect of land use on soil or water quality will not provide adequate information about the characteristics of a watershed. The paper suggested how different methods are frequently being used and how different methods are varying across the Himalayan and non-Himalayan region. There has been a noticeable increase in study related to soil erosion assessment in the past decade. Studies such as this one may have possible effects on the understanding and estimation of the water quality status based on soil erosion conditions in watersheds.