Spatiotemporal dynamic of soil erosion in the Roraya River Basin based on RUSLE model and Google Earth Engine

A

Annual Soil Loss, t · ha⁻¹ · year⁻¹

BMKG

Meteorological, Climatological, and Geophysical Agency

C

Vegetation Cover, dimensionless

CA

Cellular Automata

CHIRPS

Climate Hazards Group InfraRed Precipitation with Station data

DEM

Digital Elevation Model

EUROSEM

European Soil Erosion Model

FLUS

Future Land Use Simulation

GEE

Google Earth Engine

GIS

Geographic Information Systems

IntErO

Intensity of Erosion and Outflow

InVEST

Integrated Valuation of Ecosystem Services and Trade-offs

K

Soil Erodibility, t · ha · h · ha⁻¹ · MJ⁻¹ · mm⁻¹

LS

Slope Length and Slope Steepness, dimensionless

MODIS

Moderate Resolution Imaging Spectroradiometer

MUSLE

Modified Universal Soil Loss Equation

NCSL

Non-Cumulative Slope Length

NDVI

Normalized Difference Vegetation Index

P

Conservation Measure, dimensionless

R

Rainfall Erosivity, MJ · mm · ha⁻¹ · h⁻¹ · year⁻¹

RS

Remote Sensing

RUSLE

Revised Universal Soil Loss Equation

SDR

Sediment Delivery Ratio

SRTM

Shuttle Radar Topography Mission

TM

Thematic Mapper

USLE

Universal Soil Loss Equation

WEPP

Water Erosion Prediction Project

Soil erosion causes massive changes and damage to river beds, and land surface structures and reduces soil productivity. This event has become a concentration of studies in environmental science, agriculture, soil conservation, water resources, and various other disciplines (Vanwalleghem 2017). The destructive nature of soil erosion has impacted human survival throughout the world (Nearing et al. 2017). Several countries in Southeast Asia experience serious soil erosion problems. Data show that around 70% of land in Southeast Asia (Vietnam, Philippines, Indonesia, Laos, South Korea) experiences severe erosion (>11 · t · ha⁻¹ · year⁻¹) (Sartori et al. 2019). This amount is greater than the amount that can be tolerated by the soil (Zhao et al. 2007). The severity and frequency of soil erosion depend greatly on soil variations, climate, landforms, and various socio-economic factors (Borrelli et al. 2017; Poesen 2018; Jiang et al. 2020; Wuepper et al. 2020). In several studies, soil erosion can cause serious problems related to sediment in sedimentation areas as an off-site effect on the quality of soil in the discharge area (Mekonnen et al. 2015; Liu et al. 2020). For example, soil erosion causes a decrease in crop yields on agricultural land due to the loss of the topsoil layer (Panagos et al. 2018; Yang et al. 2019) and affects soil degradation (Pena et al. 2020; Tsymbarovich et al. 2020) and loss of nutrients (Farkas et al. 2013; Bashagaluke et al. 2018). Furthermore, soil erosion will cause the loss of storage reservoirs and reduce the life of dams and hydroelectric power plants resulting from the accumulation of large amounts of sediment in downstream areas (Kondolf et al. 2014; Bai et al. 2020). This will worsen the quality and use of surface water (Sthiannopkao et al. 2006; Issaka & Ashraf 2017). Furthermore, it threatens archaeological sites (Huisman et al. 2019; Agapiou et al. 2020) and urban areas (Knox et al. 2000; Shikangalah et al. 2016), causes internal migration (Bilsborrow 1992; Zhang & Zhuang 2019), and affects gaseous emissions (Liang et al. 2018; Lal 2020). Considering the threats and losses caused, soil erosion is a major challenge for world ecosystems (Tamene & Le 2015; Borrelli et al. 2017; Naipal et al. 2018; Fenta et al. 2020b). Based on this description, it is important to identify dynamic changes due to soil erosion (spatially, quantitatively, and qualitatively) to realize comprehensive management of soil and water conservation and ecological civilization in combating land degradation, restoring eroded soil, and reducing the risk of the entry materials into the ecosystem.

Quantitative analysis is a common method for estimating soil erosion (Li & Zheng 2012). Several models have developed in understanding soil erosion caused by water, for example the Universal Soil Loss Equation (USLE) (Wischmeier & Smith 1978), Modified Universal Soil Loss Equation (MUSLE) (Williams & Berndt 1977), European Soil Erosion Model (EUROSEM) (Morgan et al. 1998), and Water Erosion Prediction Project (WEPP) (Flanagan et al. 2001). The USLE model is applied as a direct empirical model in estimating the rate of soil loss over a relatively long period. Subsequently, this model was revised and became known as Revised Universal Soil Loss Equation (RUSLE) which was first introduced by the United States in 1986 and was optimized based on the Universal Soil Loss Equation (USLE) model (Wischmeier & Smith 1978; Renard & Ferreira 1993). The RUSLE model is considered more comparable to a computer program than sidemode plots. The USLE and RUSLE models are among the 25 most widely applied soil erosion prediction models worldwide from a quantitative point of view (Renard et al. 1997; Zha et al. 2015; Borrelli et al. 2021). The RUSLE model also provides a more reasonable solution between accuracy, deployment stability, and moderate data demand. Therefore, the intensity of its use has increased in the last few decades along with improvements in Remote Sensing (RS) technology and Geographic Information Systems (GIS). RS and GIS can facilitate the process of improving the RUSLE model locally (Haregeweyn et al. 2017; Yesuph & Dagnew 2019), regional (Tamene & Le 2015; Fenta et al. 2020a), continental (Bosco et al. 2015), and global scales (Borrelli et al. 2017; Wuepper et al. 2020). Considering data preparation and processing, modeling soil erosion on a large scale is considered to take longer and be inefficient.

The Roraya River Basin is very vulnerable to soil erosion due to the high intensity of rainfall and the power of surface runoff. The high intensity of rainfall followed by the fragmentation of forests and land into agriculture, settlements, and mining is considered responsible for the high intensity of soil erosion (La Baco et al. 2017). However, previous research only focused on soil erosion at the plot scale (La Baco et al. 2017), relatively small river basins (Maga 2018), or the upstream part of the Roraya River Basin (Gamoro 2010). In addition, regional soil erosion studies are considered to have not described specific estimates or drivers of soil erosion rates throughout the Roraya River Basin (Gamoro 2010; La Baco et al. 2017; Maga 2018). This gap will hinder understanding the contribution of individual soils in the Roraya River Basin to the total rate of soil loss. So, it is important to understand vulnerability to soil erosion as an effort to mitigate risks in the Roraya River Basin to reduce the impact of erosion on downstream areas.

To overcome this, Google, Carnegie Mellon University, and the United States Geological Survey jointly built a cloud-based platform that provides satellite imagery and observation data like Google Earth Engine (GEE). GEE can store large amounts of data and retrieve and process RS data quickly and efficiently (Gorelick 2012; Patel et al. 2015; Gorelick et al. 2017; Aldiansyah & Saputra 2023). This method is considered efficient compared to traditional image processing tools which require a download process, devices with high specifications, and quite large storage. GEE has become the most advanced cloud-based spatial data management platform known to the world (Yunfeng et al. 2018). This platform can perform cloud computing to quickly illustrate Earth Data on a global scale over a very long period (Gorelick 2012). Researchers have reported the accuracy of this platform in vegetation monitoring (Johansen et al. 2015; Wilson et al. 2017; Tao & Hong 2018), land use (Farda 2017; Huang et al. 2017), plant classification (Lavreniuk et al. 2017; Shelestov et al. 2017), water quality (András et al. 2017; Chen et al. 2017). Therefore, GEE has been successfully used in several fields ranging from regional to global scales (e.g., Tian et al. 2019; Zeng et al. 2019, 2020; Elnashar et al. 2021). However, the application is still limited to the study of soil erosion by water such as Landsat data extraction or map cover management factors, while all processing is beyond GEE (Papaiordanidis et al. 2019; Wang & Zhao 2020). In this study, soil erosion assessment was implemented entirely on the GEE platform. This aims to increase the ability to prioritize areas according to their vulnerability to erosion risk. This study uses open-access primary input data that is fully accessible in GEE, to introduce RUSLE in a spatiotemporal explicit model related to continuous monitoring and assessment of erosion hazards. For the reasons above, this study aims to model and assess soil erosion using the RUSLE model based on RS data in a GEE environment in relatively larger areas where observational the data is limited. It can be used as a rapid assessment tool in decision-making and land management in developing countries to determine priorities for where soil and water conservation projects should be built.

This research aims to integrate the RUSLE model with GEE to monitor and calculate average soil erosion in 2001, 2005, 2010, 2015, and 2021, as well as analyze the spatial and temporal characteristics of soil erosion in the Roraya watershed. These main findings can be useful in controlling land degradation and managing land resources in the Roraya River Basin. Apart from that, the concept of this study can also be applied in similar areas.

Data in this study include: (1) soil texture and organic matter for a depth of 0–200 cm at a height of 250 m obtained from the United States Department of Agriculture (https://www.openlandmap.org/); (2) annual and monthly average rainfall data in 2001, 2005, 2010, 2015, and 2021 obtained from the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) (https://www.chc.ucsb.edu/); (3) Land use data 500 m obtained from Moderate Resolution Imaging Spectroradiometer (MODIS) Land Cover Type (MCD12Q1) ver 6.1 (https://lpdaac.usgs.gov/news/lp-daac-releases-modis-version-61-land-cover-type-data-products/); (4) SRTM-DEM (Shuttle Radar Topography Mission-Digital Elevation Model) 30 m data originating from The Land Processes Distributed Active Archive Center which is from a collaborative effort by the National Aeronautics and Space Administration and the National Geospatial-Intelligence Agency (https://cmr.earthdata.nasa.gov/); (5) Landsat Satellite image data is also used for analysis. Landsat 7 Collection 2 Tier 1 and Real-Time data Raw Scenes imagery for 2001, Landsat 5 Thematic Mapper (TM) Collection 2 Tier 1 Raw Scenes for 2005 and 2010, Landsat 8 Collection 2 Tier 1 and Real-Time data Raw Scenes for 2015 and 2021 processed from GEE (https://earthengine.google.com/).

According to field data from the Sampara Watershed and Protected Forest Management Center, Southeast Sulawesi Province, the total amount of soil erosion was 86.86 × 10⁷ t in the Roraya River Basin. We then tried to calculate the total erosion in each series of years and compared it with field measurements. This study indicates that the total amount of soil erosion in the Roraya River Basin in 2001, 2005, 2010, 2015, and 2021 was 98.44 × 10⁷, 80.84 × 10⁷, 73.12 × 10⁷, 82.36 × 10⁷, and 117.9 × 10⁷ t, respectively. Referring to the results of field measurements, the results of this research show that the total erosion value is quite close to the measurement results so that it can be analyzed further.

The range of soil erosion in the Roraya River Basin varied in 2001, 2005, 2010, 2015, and 2021 in some areas, dominated by moderate erosion. The most severe areas occur in the upstream and middle parts which stretch across the highlands in the western part of the Roraya River Basin (Figure 8). Toubal et al. (2018) found that many areas with high erosion occurred at an altitude of 600–1,000 m above sea level with a slope of 25% in the Wadi Sahouat Basin, Algeria. Zhu et al. (2013) also found the same thing at the mouth of the Danjiang River, a tributary of the Yangtze River. In 2001, 2005, 2010, 2015, and 2021, the area of moderate erosion was respectively 42,332.42, 31,689.12, 30,943.99, 35,552.00, and 45,356.88 ha, or the equivalent of 28.85, 21.60, 21.09, 24.23, and 30.91% of the total area region (Tables 3 and 4). The erosion areas with the very severe class and above are respectively 37,200.23, 35,057.49, 28,975.08, 32,191.07, and 48,058.73 ha, or the equivalent of 25.35, 23.89, 19.75, 21.94, and 32.75% of the total area (Tables 3 and 4). These findings indicate that the height and slope factors are aspects that will not change in basic topographic features unless there is influence on a geological time scale, their impact on soil erosion is relatively constant in different years, but it should be remembered that different areas also have different levels of soil erosion different due to differences in topographic threshold levels.

Vegetation index and rainfall are considered to be the main factors in the dynamics of soil erosion. The NDVI values in 2001, 2005, 2010, 2015, and 2021 in the Roraya River Basin were 0.50, 0.48, 0.79, 0.65, and 0.62 (Figure 10(b)). If the NDVI value is higher, the growth of vegetation cover will be better so that the effects of soil erosion can be suppressed and reduced. When compared with the soil erosion modulus, the resulting trend is opposite to the NDVI trend. Therefore, vegetation is the reason that influences the decrease in soil erosion modulus in the Roraya River Basin. The average annual rainfall erosivity factor for the Roraya River Basin in 2001, 2005, 2010, 2015, and 2021 was 866.51, 815.37, 679.02, 753.70, and 911.49 MJ · mm · ha⁻¹ · h⁻¹ · year⁻¹, respectively (Figure 10(c)). The higher the intensity of rainfall, the higher the erosion of the soil. The rainfall trend shows a decrease from 2001 to 2015 and an increase in 2015–2021. Southeast Sulawesi received more rainfall in 2015–2020 (BMKG 2023). This finding is consistent with the trend in soil erosion modulus in the Roraya watershed. So, changes in vegetation and rainfall indices are direct causes of changes in soil erosion in the Roraya River Basin. These findings are consistent with those reported by De Paola et al. (2013) that in the Tusciano River catchment, vegetation cover can reduce the impact of rainfall and topography on soil erosion. However, Hu et al. (2010) found the opposite. Diodato (2004) also found that soil erosion in mountainous areas is strongly influenced by topography and high rainfall erosivity. When rainfall increases, soil erosion will increase. In contrast, topographic pressure and rainfall in the lowlands do not have as significant an impact as they do in mountainous areas. Due to the inconsistency of the influence of factors on soil erosion which is most likely influenced by the location and the temporal and spatial scale of the study area, as well as the type of factors that influence it need to be considered.

To understand the rate of erosion over time, a transfer matrix was created and analyzed further. From 2001 to 2005, the stability levels of very slight erosion, slight erosion, moderate erosion, severe erosion, and very severe erosion in the Roraya River Basin were 30.55, 39.48, 33.06, 23.04, and 70.99%, respectively (Table 5). From 2005 to 2010, the stability levels of very slight erosion, slight erosion, moderate erosion, severe erosion, and very severe erosion in the Roraya River Basin were 36.45, 47.02, 35.39, 21.99, and 64.08%, respectively (Table 6). From 2010 to 2015, the stability levels of very slight erosion, slight erosion, moderate erosion, severe erosion, and very severe erosion in the Roraya River Basin were 28.24, 52.41, 45.06, 32.15, and 79.01%, respectively (Table 7). From 2015 to 2021, the stability levels of very slight erosion, slight erosion, moderate erosion, severe erosion, and very severe erosion in the Roraya River Basin were 10.42, 29.52, 42, 30.15, and 92.65%, respectively (Table 8). From 2001 to 2021, the stability levels of very slight erosion, slight erosion, moderate erosion, severe erosion, and very severe erosion in the Roraya River Basin were 7.75, 36.76, 50.57, 37.68, and 87.24%, respectively (Table 9). Overall, the trend of change in the area of soil erosion is the least in the very slight class, while the area of change in very severe soil erosion is the largest in the Roraya River Basin. These findings are different from several findings reported by researchers, where the dominance of soil erosion is in the very slight to slight class. It is known that most of the Roraya River area is an area for mining, plantation, and other purposes, which is thought to be contributing to the rate of erosion. According to Rijal et al. (2019), Southeast Sulawesi is ranked second with the second highest rate of deforestation in Sulawesi which is triggered by interference from human activities. The results of this study support the statement that the trend of soil erosion modulus is inversely proportional to the vegetation index.

In the Roraya River Basin from 2001 to 2021, the proportion of areas removed from very slight erosion to slight erosion, moderate erosion, severe erosion, and very severe erosion were respectively 20.49% (3,926.86 ha), 31.66% (6,066.93 ha), 14.85% (2,845.72 ha), and 25.25% (4,839.36 ha). The proportion of areas removed from slight erosion to very slight erosion, moderate erosion, severe erosion, and very severe erosion were respectively 7.36% (2,143.66 ha), 42.47% (12,365.89 ha), 9.63% (2,803.46 ha), and 3.78% (1,099.95 ha). The proportion of areas that moved from moderate erosion to very slight erosion, slight erosion, severe erosion, and very severe erosion were respectively 3.32% (1,404.96 ha), 16.87% (7,140.84 ha), 20.92% (8,854.46 ha), and 8.33% (3,525.58 ha). The proportion of areas that moved from severe erosion to very slight erosion, slight erosion, moderate erosion, and very severe erosion were respectively 2.34% (443.22 ha), 3.9% (737.77 ha), 23.6% (4,462.27 ha), and 32.47% (6,141.31 ha). The proportion of areas that moved from very severe erosion to very slight erosion, slight erosion, moderate erosion, and severe erosion were respectively 1.64% (611.32 ha), 0.61% (227.45 ha), 2.84% (1,055.21 ha), and 7.67% (2,853.71 ha) (Table 9).

The integration of the RUSLE and GEE models as spatial analysis models was evaluated regarding soil erosion rates. This model calculates annual soil loss. Considering the predictive role in understanding the development of areas affected by soil erosion (Chalise et al. 2019). This model is a useful tool for effective coordination, for example in the case of agriculture when analyzing soil erodibility and quantifying soil loss to assess the potential of land to agricultural output. So, the RUSLE model is very suitable for evaluating soil erosion for this assessment (Benavidez et al. 2018). Considering that this method was known about a quarter of a century ago, updated concepts and empirical formulations for calculating each parameter should be considered.

Erosion estimation models are not limited to the RUSLE model, there are other models such as Intensity of Erosion and Outflow (IntErO) (Mohammadi et al. 2021; Sestras et al. 2023) and Integrated Valuation of Ecosystem Services and Trade-offs-Sediment Delivery Ratio (InVEST SDR) (Marques et al. 2021; Qiao et al. 2023). The IntErO model can understand the rate of erosional material lost from river basins. However, RUSLE is considered to have significant advantages over IntErO because it is a spatially explicit model, whereas IntErO calculates sediment yield at a more global scale. Despite its shortcomings, the IntErO model is worth considering when erosion assessments are carried out on the resulting erosion material. The volume of soil discharged is calculated from the watershed area represented by the measured sediment. The IntErO model can calculate the maximum outflow from the river basin, the asymmetry of the river basin and the coefficients of the river basin form, watershed development, the river basin tortuousness, the region's permeability, and vegetation cover. It becomes important when studies have to determine soil loss in a landscape and this model is recommended. The RUSLE model only calculates gross soil erosion rates, which may or may not include soil lost from watersheds (Chalise et al. 2019).

Although its use is very common, RUSLE is limited to inter-groove erosion processes (Renard & Ferreira 1997). To overcome this, the InVEST SDR model was developed based on the RUSLE model to understand soil erosion, sediment export, and soil conservation at various spatial and temporal scales widely (Ougougdal et al. 2020; Marques et al. 2021). This model is known for its structural peculiarities in testing the impact of soil erosion on land use (Zhou et al. 2019). However, there are concerns when parameter estimates are only based on physical geographic characteristics. This may cause calculation errors when applied to other countries or regions (Parsons 2019). However, this process is not represented in this model. The SDR model has limitations (Sharp et al. 2018) where it is very sensitive to and ⁠, which are not physically based. and are two calibration parameters that determine the shape of the relationship between hydrologic connectivity and the sediment delivery ratio. Additionally, the model produces NoData pixels in the stream network caused by inattention during in-stream processing. When this process takes place, the sediment will be moved down the slope and stop the calculation process when the sediment reaches the river, so that in estuary areas pixel errors occur where this area has a larger water body. Given its simplicity and relatively small number of parameters, this model requires quite a long time to process and adjust the model. Despite its limitations, the model has undergone full calibration and verification in other regions around the world with different climates, topography, soil types, and land use types (Zhou et al. 2019; Aneseyee et al. 2020; Duan et al. 2020; Gashaw et al. 2021).

Erosion prediction models are often combined with land cover change models such as Future Land Use Simulation (FLUS). Various approaches have been used to model land use changes such as mathematical models, system models, statistical models (regression), cellular models (cellular automata (CA) and Markov chains), evolutionary models (neural networks), and agent-based models (Guan et al. 2011). This was done because several studies have reported that vegetation cover can greatly influence the intensity of soil erosion (Zhou et al. 2008; Mohammad & Adam 2010; Anh et al. 2014; Ferreira & Panagopoulos 2014; Li et al. 2014; Sun et al. 2014; Wang et al. 2016; Xiong et al. 2019; Borrelli et al. 2020; Lins et al. 2023; Moisa et al. 2023). Vegetation influences soil erosion through its canopy, roots, and felt (Gyssels et al. 2005). Thus, soil erosion is directly influenced by land use (García-Ruiz 2010; Leh et al. 2013; Ferreira et al. 2016; Mohammadi et al. 2021). Immigration and conversion of forest land into agricultural areas, grazing land, and residential areas have been followed by development on the coast. Therefore, land use modeling and simulation are important for predicting future soil erosion and land degradation and for various watershed planning and management problems (Deng et al. 2008). The FLUS model is quite good at predicting land use, especially river areas, and is very helpful in predicting changes in soil loss and sediment drainage in various river area development scenarios (Qiao et al. 2023). This model is a useful tool for predicting the impact of land use policies and urban growth on watershed management (Zare et al. 2017) thereby providing a basis for policymakers to change programs related to land conservation policies (Prazan & Dumbrovsky 2011).

Erosion that occurs threatens natural resources, so fast and efficient tools are needed to support regional conservation management practices in erosion-prone areas. This study attempts to assess soil erosion trends by integrating the RUSLE model in the GEE environment in the Roraya River Basin. The RUSLE and GEE model estimates of this study show that the average soil erosion rate in 2001–2021 throughout the Roraya River Basin is 307.22 t h⁻¹ year⁻¹. About 25% of the total area should be placed under urgent soil conservation measures. This area includes the upper and middle parts that stretch across the plateau in the western part of the area. Moderate erosion was the most common form of hydraulic soil erosion during the study period, indicating a downward trend rather than an improvement in regional soil erosion conditions. The trend of change in the area of soil erosion is the least in the very slight class, while the area of change in very severe soil erosion is the largest in the Roraya watershed. It can be seen that two static factors (topographic and slope) show regular impacts and are said to predict soil erosion trends in this study, but vegetation cover and rainfall erosivity act as dynamic factors responsible for substantial interactions. Both influence soil erosion and are equally important.

This paper has shown that to investigate multi-year spatiotemporal dynamic changes in soil erosion, the integration of RUSLE and GEE modeling techniques provides a simple, efficient, and reliable method. In addition, this study quantitatively analyzes the interactions between altitude, slope, vegetation cover, and rainfall erosivity, as well as soil erosion. Vegetation cover factors and conservation measures in the RUSLE model can also be handled well. Based on these findings, this study recommends immediate implementation of soil conservation practices such as bench terraces, flat terraces, culverts, embankments in mining areas, planting cover crops, and fast-growing plant species in the most threatened areas, especially in the northern part of the study area. Further research is expected to carry out experimental plots to validate the RUSLE model. Meanwhile, decision makers should implement future development plans to cover a wider scale in a more scientific manner for water and soil conservation in the study area, especially in developing countries.

The authors would like to thank the University of Indonesia for supporting this research, in addition to the Konawe Selatan Regency Government and the Community and Sampara Watershed and Protected Forest Management Center, Southeast Sulawesi Province for allowing and facilitating the collection of data.

This paper was composed by collaboration among all authors. S.A. and F.W. designed this study, F.W. helped improve its progression and clarity, S.A. and F.W. wrote this paper, and F.W. helped in revising the paper.

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

The authors declare there is no conflict.