Assessment of soil erosion and sediment yield in the Peddavagu watershed, India, using a revised universal soil loss equation model (RUSLE) and GIS techniques Open Access

Soil erosion is the most common form of land degradation caused by a combination of natural processes and human activities (Ganasri & Ramesh 2016; Bhattacharya et al. 2020). Soil erosion is a worldwide problem that has both on-site and off-site impacts (Bogale et al. 2020). It has come to the observation of decision-makers and land managers all over the world that the fight against soil erosion and other kinds of land degradation has been a source of concern (Roder & Hill 2009). Numerous important topography factors have an impact on the mechanism of soil erosion (Ustun 2008; Mitasova et al. 2013; Saravanan et al. 2021). An important factor affecting the runoff process is aspect and slope. As the slope increases, there is a corresponding increase in runoff and a subsequent reduction in infiltration. Due to the slope's rainfall runoff's faster rate, soil erosion will occur. Erosion occurs when wind or water moves soil particles, transferring them to other areas. Erosion is initiated by a confluence of elements, including the presence of steep slopes, climatic conditions, unsuitable land utilisation, patterns of land cover, and so on (Renschler et al. 1999; Mathew & Shekar 2022; Shekar et al. 2023a).

Thlakma et al. (2018) and May & Place (2004) identify water-induced soil erosion as the most prevalent type of soil erosion when compared to other causes. Land-use land cover (LULC), soil type, surface gradient greatly, and other factors affect erosion. Erosion reduces soil nutrients and organic matter (OM), affecting soil performance. The issue of soil erosion has been recognised as a global concern due to the interconnectedness of natural resources, population dynamics, and the phenomenon of climate change (Lal 2003). The utilisation of effective modelling techniques facilitates the acquisition of valuable insights pertaining to the present state of erosion, its patterns of change over time, and the ability to conduct scenario analysis.

Soil erosion models employ mathematical formulations to establish relationships between key characteristics and land surface processes (Jetten et al. 2003). The criteria encompassed in their study consist of meteorological variables, topography characteristics, LULC, and soil properties (Merritt et al. 2003; Saravanan et al. 2010; Ochoa-Cueva et al. 2015). Soil erosion models include detachment, deposition, and transport (Nearing et al. 1994). Soil erosion models serve as pivotal planning tools due to their invaluable ability to anticipate soil loss, as highlighted in the research by Benavidez et al. (2018). The selection of an appropriate model for a specific soil erosion study hinges on a triad of critical considerations: the unique characteristics of the watershed region under examination, and the data availability or insights regarding the model's effectiveness, as elucidated by Keesstra et al. (2014).

Conventional methods for evaluating soil erosion risk are known for their high costs and time demands. A wide range of models (conceptual, physical, and empirical) has been established on a global scale to facilitate soil loss assessment (Maqsoom et al. 2020). Data inputs and complexity differ amongst these models. Soil erosion models, like USLE (Wischmeier & Smith 1978) and the revised universal soil loss equation (RUSLE) proposed by Renard et al. (1997), are used to predict the rate of gross soil loss at the plot scale. One of the models that has achieved widespread acceptance is the empirical RUSLE (Chuenchum et al. 2019; Ghosal & Bhattacharya 2020). Aspects like the availability of data or information and the intended use of the model influence the choice of models for estimating erosion, among other considerations. Moreover, the model demonstrates strong applicability in ungauged catchments. It has a moderate data requirement and, most importantly, effectively integrates with geographic information system (GIS), facilitating the upscaling of soil erosion processes. This method is utilised by scientists worldwide to study soil erosion at various spatial scales (Renard et al. 1997; Fernandez et al. 2003; Gitas et al. 2009; Yue-Qing et al. 2009; Manjulavani et al. 2016; Thlakma et al. 2018; Chuenchum et al. 2019; Fenta et al. 2020).

The evaluation of sediment discharge plays a role in the detection of diffuse source contamination as well as in the strategic development and implementation of construction endeavours. Nevertheless, the measurement of sediment production is typically not directly accessible, and instead, it is estimated through the utilisation of an sediment delivery ratio (SDR) as proposed by Verstraeten & Poesen (2002). Accurately predicting SDR holds significant importance in the management of sediments for the purpose of sustainable development of natural resources and preservation of the environment. Soil erosion represents the initial stage within the sedimentation process, encompassing the sequential occurrences of erosion, transit, and deposition of sediment. A portion of eroded soil is transported via a channel system, contributing to sediment yield (SY), whereas a component of this soil undergoes deposition within water channels. In order to accomplish the purpose of measuring sediment yields, it is possible to make use of the SDR, which provides an indication of the percentage of total soil loss that is caused by water that flows to a certain location within the channel network. This can be done by calculating the SDR at the location in question. Utilising the RUSLE model to quantify spatial gross soil loss is an efficient way to determine the amount of soil that has been lost due to erosion and the amount of sediment that has been moved within a watershed. Furthermore, the estimation of sediment delivery within watersheds can be successfully accomplished through the implementation of the SDR technique. The process of raindrops impacts detaching soil particles from the earth and results in their transportation to rivers, hence increasing the sediment load and SDR of watersheds (Beskow et al. 2009; Colman et al. 2018; Rajbanshi & Bhattacharya 2020; Ahmad et al. 2022). The use of the SDR allows for the estimation of sediment delivery in watersheds that lack sufficient data. The integration of the RUSLE and the SDR has been identified as a major technique for predicting SY and soil erosion in places with limited availability of data (Bhattacharya et al. 2020; Ahmad et al. 2022).

The objectives of this current research were (1) to estimate soil erosion using the RUSLE in a GIS environment; (2) to estimate the SDR of the Peddavagu watershed; and (3) to estimate the SY by combining RUSLE with SDR in the Peddavagu watershed.

The watershed's boundaries were defined through the utilisation of a digital elevation model (DEM) downloaded from National Aeronautics and Space Administration (NASA) website. To create the map representation, remote sensing (RS) data from the Landsat 5 satellite were employed, and this process involved the application of supervised classification methodologies as detailed by Roy et al. (2016). The FAO (1974) launched the Digital Soil Map of the World (DSMW) soil mapping and database project. In addition, the data on rainfall that was particular to the Peddavagu watershed were extensively collected from the IMD in Pune. The major resource for this important meteorological data was the official IMD website (https://mausam.imd.gov.in/), which can be accessed here. The India-WRIS website provided suspended sediment concentration and runoff data for the Bhatpalli gauging station in the Godavari River Basin (http://www.india-wris.nrsc.gov.in/) (Das et al. 2021, 2022).

The SE represents the mean annual soil loss (tonnes/hectare/year). The factors P, C, and LS are dimensionless.

The process of validation holds significant importance in the field of modelling, as it is essential for ensuring the scientific integrity of the models. Without undergoing validation, models lack the necessary credibility and scientific merit. The accuracy of soil erosion was verified using the receiver operating characteristic curve (ROC) and area under the curve (AUC). The present investigation employed India-WRIS data to verify the model (RUSLE)'s SY output.

It measures potential erosion and is impacted by the soil's underlying features. Soil types vary in water-induced erosion susceptibility. The inherent susceptibility of soil loss from rainfall and surface overflow (Wischmeier & Smith 1978; McCool et al. 1995; Thomas et al. 2017). A diverse range of chemical and physical soil factors influence the erodibility of soil. The RUSLE model takes into account various physical features, including the structure of the soil, OM content, primary particle size distribution, and permeability, which are considered to be the key elements that affect the K-factor. The quantification of soil parameters is often conducted in a laboratory setting using established conventional methodologies. Conducting a comprehensive field survey of soils in certain areas within the watershed is unfeasible. Although laboratory-based approaches continue to be widely utilised, their utilisation is accompanied by significant drawbacks. These drawbacks include high costs, time-intensive procedures, and the potential for environmental harm due to the generation of chemical residues (Nanni & Demattê 2006).

Both slope steepness (S) and slope length (L) are important topographic parameters that have a significant effect on soil loss. The LS component in the RUSLE considers these qualities (Datta & Schack-Kirchner 2010; Prasannakumar et al. 2012). Typically, regions characterised by pronounced inclines are more prone to erosion compared to locations that are flat or have gradual slopes. Similarly, as slope length rises, soil erosion increases, and vice versa. The aforementioned analysis demonstrates the susceptibility of topographic parameters to soil erosion, underscoring the importance of obtaining precise measurements for this variable.

The P-factor and C-factor are interconnected as they both aim to assess the efficacy of management practices in mitigating soil loss (Renard et al. 2011). Nevertheless, variable C demonstrates the favourable influence of managerial strategies on erosion, such as the existence of biomass on beneath the soil or the surface, as well as the level of surface roughness. Conversely, the positive effects of management practices on controlling runoff, particularly in terms of altering the direction and velocity of the runoff, are denoted as P (Morgan & Nearing 2011; Benavidez et al. 2018). It is critical to determine LULC characteristics and steepness within a specific catchment region. The variable P has been employed to denote the influence of agricultural practices and land management approaches on soil erosion, with the aim of mitigating the rate of soil erosion through the alteration of runoff flow direction.

The watershed has been partitioned into 13 distinct sub-watersheds (SW) in order to assess soil loss within each individual sub-watershed (Figures A1 and A2). Within the GIS framework, the process of extracting individual sub-watershed shapefiles from a total of 13 SW is accomplished through the utilisation of the export data function. Once all the individual SW shapefiles have been obtained, the Arc toolbox provides a tool called spatial analyst tools. Within this tool, the extraction function, namely the extract by mask feature, is utilised. The input data required for the extract by mask operation consist of a complete soil loss raster file and an individual shapefile representing the watershed. This process aims to obtain the final soil loss of sediment in each individual watershed, measured in tonnes/hectare/year. SW 3 had the lowest mean annual soil loss, 14.10 tonnes/hectare/year, while SW 9 had the highest, 51 tonnes/hectare/year (Figure A3).

Ferro & Minacapilli (1995) suggest that the SDR represents the likelihood of sediment particles eroded from a specific upland region successfully reaching the next stream channel. This parameter calculates the percentage (%) of gross soil loss that reaches the drainage network and watershed outlet, ranging from 0 to 1. The watershed physiology, texture of the eroded material sediment source, land cover, and transport system might affect SDR values (Walling 1988; Richards 1993). After computing all the factors of the RUSLE model, ArcGIS 10.7 software created thematic layers representing each specific RUSLE parameter. Subsequently, these thematic maps were multiplied together using a tool called a raster calculator. From a managerial standpoint, SY holds greater significance than potential soil loss, as it provides a quantifiable measure of the amount of dirt being transported out of a watershed. Therefore, this study has integrated an approach for estimating SY based on the SDR. The SY is the amount of erosion in a catchment area that has not yet been accumulated and is leaving the watershed within a certain timeframe. Typically, the combined effects of erosion and the magnitude of the SDR at a particular position determine the sediment output at that place. SDR denotes the percentage of total erosion that reaches the designated measurement site. This metric represents the effectiveness of sediment transport at the given location.

Equation (2) from the USDA (1979) gives the Peddavagu watershed's SDR of 0.210. According to Mohapatra (2022), there exists an inverse relationship between the area of the watershed and the soil detachment rate. Figure A4's data reveal that the sub-watershed's SDR values range from 0.266 (SW 5 Area: 367.13 km²) to 0.295 (SW 4 Area: 146.56 km²), while the overall watershed value stands at 0.210 (21.0%). These findings reveal that the channels have carried 21.0% of the soil loss. These findings suggest that approximately 21.0% of the eroded soil materials have the potential to reach the channels, while the remaining 79.0% effectively retain and deposit before entering the channel networks.

The watershed is divided into 13 SW to evaluate their sediment yields (Figure A5). Within the GIS environment, the process of extracting individual sub-watershed shapefiles from a total of 13 SW is accomplished through the utilisation of the export data function. Once all the individual SW shapefiles have been obtained, the Arc toolbox offers a tool called spatial analyst tools. Within this tool, the raster calculator is utilised as a form of map algebra. The input data for the raster calculator consist of a specific soil loss raster file for each individual sub-watershed unit, together with the corresponding SDR value for that particular SW unit. These inputs are used to calculate the final SY for each individual SW unit, expressed in tonnes/hectare/year. The study region revealed that SW 9 and SW 5 exhibited the maximum and minimum mean SY, as measured by the SDR, with values of 14.08 and 4.02 tonnes/hectare/year, respectively.

The current investigation provides information about soil erosion classification and shows high soil erosion in the Peddavagu watershed; however, it has limitations. Watershed sediment and soil erosion are major barriers to sustainable land use and stream and lake water quality (Mutua & Klik 2006). RUSLE also ignores gully erosion and only considers sheet/interill and rill erosion. Additionally, the RUSLE model does not directly represent watershed sediment output (Phinzi & Ngetar 2019).

The present investigation incorporates the analysis of rainfall, soil, LULC, and DEM data. The R, K, C, P, and LS factors are first computed within the GIS framework. Subsequently, these factors were utilised to estimate soil loss through the implementation of the RUSLE. The average annual soil loss for the watershed is 17.91 tonnes/hectare/year, indicating high erosion risk. The model's accuracy suggests a very good (82.1%) outcome for the RUSLE model results. Moreover, the study region revealed that SW 9 and SW 3 exhibited the maximum and minimum average soil loss, measuring 51 and 14.10 tonnes/hectare/year, respectively. The Peddavagu watershed's SDR was approximately 0.210. The findings also indicate that the yearly SY originating from the entirety of the watershed amounts to 3.76 tonnes/hectare/year. The study region revealed that SW 9 and SW 5 exhibited the maximum and minimum mean SY, as measured by the SDR, with values of 14.08 and 4.02 tonnes/hectare/year, respectively. The model's performance was assessed by comparing its predictions to gauge data for validation. The results indicate that the observed actual data was 3.66 tonnes/hectare/year, whereas the anticipated value was 3.76 tonnes/hectare/year. Moreover, it offers significant insights for policymakers, decision-makers, stakeholders, and international organisations collaborating to advance the implementation of sustainable watershed management techniques.

The satellite data were made accessible by the USGS, which the authors would like to appreciate for their efforts (https://earthexplorer.usgs.gov/). On their website, which can be found at https://mausam.imd.gov.in/, the Indian Meteorological Department has kindly provided data on the amount of rainfall. We are grateful that the FAO has provided us with a soil data. Also, authors would like to thank India-WRIS for providing daily data of surface runoff and suspended sediment concentration (http://www.india-wris.nrsc.gov.in/).

P.R.S. conceptualised the whole article, developed the methodology, arranged the Software, rendered support in data curation; validated the data, and wrote the original draft. A.M. supervised the article, investigated the data, visualised the process, reviewed and edited the article.

There was no funding for this project

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.