Spatial-Temporal Trend Analysis of Rainfall Erosivity and Erosivity Density of Tropical Area in Air Bengkulu Watershed, Indonesia

A discourse analysis of several studies based on the research location for a contrastive study of rainfall erosivity and ED has been carried out. Erosivity modelling of annual precipitation in Europe is presented on a regional scale to analyze soil loss and soil erosion (Panagos et al. 2017). Furthermore, the spatial distribution of rainfall erosivity trends shows an increasing trend of rainfall erosivity from winter to summer in Western to Eastern Europe (Ballabio et al. 2017). Subsequent research also highlighted that the monthly rainfall erosivity map that depicts the spatial and temporal variability could explain the concentration of erosive events and identify the monthly periods to assess soil loss and conservation action measures (Diodato, Bellocchi 2012).

Research on the Himalayan watersheds in Asia shows that the study of rainfall erosivity uses parametric statistics with correlation and linear regression for trend assessment. The study results of this research on rainfall erosivity with annual rainfall have a moderate to a strong relationship, and there is an insignificant increasing or decreasing trend (Ma et al. 2014). However, according to Singh and Singh (2020), these findings show that July is the most vulnerable month for soil erosion in the Suketi River watershed, the Himalayan followed by August and June, and the lower reaches (Northern) being the most vulnerable. Then, research in the Baram River in Sarawak (Malaysian, Borneo) by Vijith and Dodge-Wan (2019) explained that the spatial pattern of rainfall erosivity trends approaches with statistical parametric and nonparametric (linear regression, Superman's Rho, and Mann Kendall). The highest observed has an increasing trend in the Northern Zone and a decreasing trend in the Southern Zone. However, the Baram River in Sarawak research is limited to the spatial distribution of rainfall erosivity and ED based on seasons and years.

Since 1974, Indonesia has been conducting experiments to study rainfall parameters with soil erosion for land criticality assessments. It is found that the rainfall intensity value of 30 min is a show of kinetic energy from the rainfall erosivity map for the island of Java, Indonesia (Asdak 2014). This value would certainly not be appropriate if it is applied to all Provinces in Indonesia. The importance of this research is because Indonesia ranks third in the world for critical watershed land conditions (Asdak 2014). Furthermore, research in the Merawu watershed, Central Java, Indonesia, highlighted that the different equations of rainfall erosivity models could explain actual erosion with an accuracy of 81.13% (Sulistyo 2011).

Erosion research at the study site found that the period of soil erosion occurred. The highest surface erosion rate of the Air Bengkulu watershed occurred in February (0.421 mm · month⁻¹) and the smallest in September (0.023 mm · month⁻¹) (Tunas 2005). This is exacerbated by the damage to the watershed that occurs due to coal mining activities accompanied by changes in the function and designation of Protected Forests in the upstream part of the Air Bengkulu watershed into Permanent Production Forests (Satmaidi et al. 2018). Then the discussion in the downstream watershed stated that there had been erosion and sedimentation in general, which if not handled can experience inundation due to extensive floods, such as the Bengkulu City flood in 2019 (Mase 2020).

Although there has been researched on the potential for erosion with rainfall erosivity based on the research location regionally and the watershed discussed above, several literature studies that have described the assessment of rainfall erosivity are presented with trend analysis and spatial-temporal distribution on an annual, seasonal and monthly basis. However, there is no current comparative research on the development of soil erosion, especially regarding the erosivity approach of monthly rainfall erosivity in the watershed which is tropical Indonesia. This is because the dynamics of the atmosphere are influenced by rainfall caused by trade winds, monsoon winds, and the influence of various local conditions (Eddy, Lestari 2007).

The Air Bengkulu watershed has changed dramatically in recent years due to the rapid expansion in coal mining activities, intensive agricultural practices with sharp increases. In addition, the soil has been exacerbated by tracks, paths and poor access for coal truck mobilisation as the impact of soil compaction because the presence of the trail will have a much greater impact on increasing water flow. Forests continue to be cut down for agricultural activities and infrastructure development and the expansion of mining areas. These anthropogenic activities greatly influenced the rate of soil erosion along with natural agents such as topography, slope, amount, duration, and intensity of rainfall, soil types, geological formations, and types of land use. Based on these observations, areas prone to soil erosion in the fragile Air Bengkulu watershed need to be studied.

Therefore, it is essential to study the spatial-temporal trends of the monthly rainfall erosivity. Because during the last decade, no research on the potential for soil erosion has been carried out in the Air Bengkulu watershed based on rainfall erosivity. This is done to analyse the monthly erosive period for spatial distribution rainfall erosivity and ED and to improve our understanding of the evolution of rainfall erosivity in the Air Bengkulu watershed, particularly concerning rainfall events.

The main objective of this study was to assess the rainfall erosivity of precipitation and moreover detailed studies on the distribution of spatial and temporal rainfall erosivity and monthly erosivity in Indonesia in general and the Air Bengkulu watershed, in particular, are lacking. The followings are questions as guidelines to achieve the goal: –

How is the rainfall erosivity based on the monthly rainfall data in the Air Bengkulu watershed?

The Air Bengkulu watershed astronomical location stretches between longitudes 102°14′48.962″ to 102°35′5.992″ East, and between latitudes 3°37′8.705″ to 3°50′30.802″ South. The distribution of annual rainfall between 2900 and 3600 mm can be seen in Figure 1. The rainfall occurs throughout the year and is based on climate classification classified as climate type A (Wet Tropical) with a humidity of 70–87% (Case et al. 2007). The mean annual temperature is 25–32°C with monthly rainfall 230–620 mm, and the number of rainy days ranges from 10 to 23 days. Geographically, the Air Bengkulu watershed is located in Bengkulu Province, precisely on the west coast of Sumatra Island, Indonesia. Administratively, Central Bengkulu Regency and Bengkulu City have an area of 50,049 ha (Sulistyo et al. 2020). The watershed consists of three sub-watersheds, namely Susup River, Rindu Hati River, and Bengkulu River (Supriyono, Yanmesli 2016). Topographically, it is located on hills with a steep slope approaching the summit. The upstream of the Air Bengkulu watershed occupies the Western side of the Bukit Barisan which stretches from the Northern tip to the south of the island of Sumatra. Meanwhile, the downstream is a plain area directly opposite the Indian Ocean.

Fig. 1

In this study, the monthly rainfall data for the recording period from 2006 to 2020 were retrieved from 19 weather stations in the Air Bengkulu watershed. Monthly rainfall is the amount of daily rainfall in 1 month of observation at a specific weather station. Monthly temporal rainfall is obtained from data provided by the Class I BMKG Station in Pulau Baai Bengkulu. The data variation of each weather station shows the monthly rainfall measurement period, presented in Table 1. Data analysis of rainfall calculations was carried out using the Microsoft Excel program. Then, the data is processed to obtain the monthly rainfall erosivity value, statistical trends and mapped with a geographic information system (GIS).

Based on the average of 19 weather stations, the highest rainfall in the Air Bengkulu watershed in November is presented in Figure 2. Above-average monthly rainfall data is shown in November, December, and April (NDA), namely the average rainfall 352.60–432.42 mm. The sea surface temperature influences this in the tropics vary in both space and time scales. The intense interaction between the atmosphere and the oceans in the Indian Ocean region produces a dipole mode phenomenon which is defined as a symptom of anomalous sea surface temperature.

Fig. 2

The Western part of the Air Bengkulu watershed, which directly faces the Indian Ocean, is heavily influenced by sea breezes. It experiences humid air for most of the year. This dipole mode phenomenon can also be found in the Southern Indian continent. High rainfall is inversely proportional to the sea surface temperature anomaly in Indonesian waters in West Sumatra (Hermawan, Komalaningsih 2008). The Indian Ocean plays a vital role in determining the supply of water vapour during the monsoon cycle, which can affect the rainfall pattern in the Air Bengkulu watershed. The lowest monthly rainfall in June, July, August, September (JJAS) shows a rainfall of 162.75–192.82 mm.

The rainfall data were tested for homogeneity using the Bartlett test, and the same method was used to evaluate the homogeneity of variance for the data obtained from the data difference stations for 5–15 years of rainfall data. The residuals were checked for normality and homogeneity to ensure that they met the analysis of variance's assumptions. The Shapiro-Wilk (SW) test (Shapiro, Wilk 1965) was used to confirm normality, while the Bartlett test was used to confirm homogeneity (Bartlett 1992). The application of the Bartlett (B) and SW tests made it possible to evaluate the homogeneity of variance and normality for rainfall, erosivity rainfall and ED of data on the weather station scale presented in Table 2. However, the rainfall data for each extreme weather station is for that. Test the data transformation using the Two-Step Approach for Transforming Continuous Variables to Normal (Templeton 2011).

Understanding the rainfall erosivity and the ED in space and time is very important to predict the potential risk of water erosion. This study calculates the rainfall erosivity using an equation developed by Wischmeier and Smith (1978), furthermore this equation is also developed using monthly and annual rainfall data (Arnoldus et al. 1980). The Arnold equation shows that reliable rainfall erosivity studies explain the relationship between rainfall and soil erosion processes Yu (2005) also uses similar methods to (Mondal et al. 2016, Singh, Singh 2020). Eq. (1): (1) R=∑i=1121.735×101.5 log(Pi2P)−0.08188 R = \sum\limits_{i = 1}^{12} {1.735 \times {{10}^{1.5\,\, log \left({{{P_i^2} \over P}} \right) - 0.08188}}} where: –

R – rainfall erosivity (MJ · mm⁻¹ · ha⁻¹ · h⁻¹ · a⁻¹);

The ED is expressed as the ratio of erosivity to precipitation to rainfall, which measures erosivity per unit of rainfall (mm) (Panagos et al. 2015). The assessment results then can explain the occurrence of rainfall, which shows the erosivity of the most erosive rainfall. Eq. (2): (2) ED=RiPi ED = {{{R_i}} \over {{P_i}}} where: –

ED – monthly erosivity density (MJ · ha⁻¹ · h⁻¹),

After producing a monthly rainfall erosivity assessment of each weather station, different statistical analyses were carried out to identify trends and their statistical significance, using linear regression, Spearman's Rho test, and Mann Kendall test. Linear regression is the simplest and best parametric method for estimating the relationship between variable rainfall and rainfall erosivity. This regression line is a straight line contained in a scatter diagram, which shows a relationship between the two variables (Morgan, Nearing 2011, Ballabio et al. 2017, Singh, Panda 2017). In most statistical analyses of trend determination, the Spearman's Rho test statistic is supported by the best method of test results. The z_sd value will provide whether the observed data has a statistically significant trend. If z_sd >0, that indicates a positive trend and <zero for a negative trend based on statistical significance. For most statistical analyses of trend determination, the Spearman's Rho test statistic is supported by the Mann Kendall test statistic, the best complementary method available for cross-validation of test results. Mann (1945) and Kendall (1948) proposed a nonparametric test based on ratings to detect linear and non-linear trends and are widely used to detect randomness in rainfall erosivity and ED data (Mohtar et al. 2015, Shin et al. 2019, Liu et al. 2020, Zhu et al. 2021).

The monthly erosivity value from 19 weather stations was used to obtain the spatial distribution in the Air Bengkulu watershed using the interpolation method. Interpolation is a mathematical function algorithm to estimate the value at locations where data is not available or often called resampling (Tokola 2000, Hanel et al. 2016). The inverse distance-weighted (IDW) interpolation analysis was chosen because this model is often used in estimation exploration activities, and the calculation process is simpler and easier to understand (Hanel et al. 2016).

IDW is an interpolation method to estimate a rainfall erosivity value and ED at the research location which is not sampled based on monthly rainfall erosivity data and has good accuracy (Vijith, Dodge-Wan 2019). Monthly erosivity values were mapped spatially using GIS (Xu et al. 2019). The analysis was carried out by observing the gradation of the resulting map colour so that the erosive period could be identified (Borrelli et al. 2017). However, the value of the unsampled point is assumed to be the distance-weighted average of the values of nearby sampled points in IDW interpolation. Except that nearer points should be closer to the value of the interpolated position than more distant points, this approach does not consider any spatial relationships, including the effects of topography. This has been done by (Xu et al. 2019) choosing the IDW method to calculate the spatial distribution of rainfall erosivity at a station based on weather station data.

The value of rainfall erosivity in the temporal period of 2006–2020 shows the average monthly rainfall is presented in Figure 2. A graphical display of the rainfall erosivity summary is shown, where the horizontal central line within the box represents the median, the upper and lower horizontal lines enclosing the box are at the upper and lower quartiles, respectively, and the whiskers extend to the two extreme observations. Shown boxplots, the horizontal line within the box indicates within the present box indicates the mean. The illustrative dataset rainfall erosivity Air Bengkulu watershed at weather station recorded.

The monthly period of maximum rainfall erosivity occurs in November, which is the highest peak of rainfall. Monthly erosive periods occur in November, December, and April (NDA) or wet months, with variations in the rainfall erosivity index of:

840.94 MJ · mm⁻¹ · ha⁻¹ · h⁻¹ · a⁻¹,

Meanwhile, the minimum rainfall erosivity is in July. The lowest erosive monthly period occurs in June, July, August (JJA) with a rainfall erosivity value of 87.55–105.46 MJ · mm⁻¹ · ha⁻¹ · h⁻¹ · a⁻¹.

The results of the rainfall erosivity research link the values of r² and r_s to conclude that the data has a trend. The results show that rainfall erosivity is a relationship between rainfall erosivity and the conditions of monthly rainfall in the Air Bengkulu watershed (Table 3). Statistical analysis with correlation analysis assesses the extent to which the strength between the variables affects. The results of the study obtained significant with a significant level of 0.001, the influencing variable and having a solid relationship between the rainfall erosivity variable and monthly rainfall.

The monthly rainfall erosivity trend for 19 weather stations shows a positive trend, which means that the data is increasing in the Air Bengkulu watershed. This indicates that the results of the slope (b₁) value between 1975 and 3950 from the regression explain the assumption that bulk erosivity has an increasing trend Pagar Jati station with the lowest score and Anak Dalam station with the highest score. The regression parameter value also explains the coefficient of determination (r²), showing 0.91–0.99. This means that 91–99% of the variation in monthly rainfall is related to and affects the rainfall erosivity. Slope assessment (b₁) can be used to detect trends in rainfall erosivity in the Air Bengkulu watershed.

Spearman's Rho analysis can be observed from the lowest to the highest with a Z score a value of 38.109 and 115. 928 at Tanjung Heran and Talang Pauh stations. Then, H_o rejects and shows that the data has an increasing and significant trend at a significant value of 0.001. However, in the analysis of the Man Kendall test, the results show that the Z score is between −0754 and 0.825. H_o was accepted and stated that the trend was insignificant. This test also explains that 11 weather stations are trend increasing and 8 stations are trend decreasing. Therefore, the percentage of monthly rainfall erosivity was 57.89% increasing and 42.11% decreasing.

Data analysis from the interpolation stage can be seen from the colour gradation that shows monthly rainfall erosivity presented in Figure 3. There is a colour gradation from yellow to blue. The colour gradation that has more yellow colour indicates the lower value. In comparison, the blue colour gradation indicates a higher value. The spatial distribution of the monthly rainfall erosivity explains that the upstream watershed of Air Bengkulu, with the rainfall erosivity, continues to increase. The homogeneous colour gradation occurs in November, when it turns blue, and in July it turns yellow.

Fig. 3

This shows that the erosive period, namely November, December, and April are the most vulnerable. But the months of June, August, and September will witness the lowest periods of the erosion process. But the months of January, February, March, May, and October, the spatial distribution of rainfall showed heterogeneous colour gradations throughout the watershed. However, the Taba Penanjung station shows consistency and continues to increase with increasingly green colour gradations in the Northeast area of the Air Bengkulu watershed.

Analysis of the average ED of 19 weather stations in the Air Bengkulu watershed is for graphic comparisons of a ED of monthly data boxplots presented in Figure 4. The ED for the erosive months in November, December and April, (NDA) with ED values between 1.12 MJ · ha⁻¹ · h⁻¹, 1.59 MJ · ha⁻¹ · h⁻¹ and 1.23 MJ · ha⁻¹ · h⁻¹, 1.12 MJ · ha⁻¹ · h⁻¹. This value is directly proportional to the rainfall erosivity and indicates that rainfall is a potential disaster risk.

Fig. 4

Based on Table 4 shows the value of the coefficient of determinant between 0.795 and 0.989. This means that 79.5–98.9% density erosivity strongly influences both variables, namely rainfall erosivity and monthly rainfall. Further test with the Spearman's Rho assessment is directly proportional to a value between 0.904 and 0.994. This also shows that monthly rainfall has a strong relationship with the calculation of rainfall erosion at the rainfall station in the Air Bengkulu watershed.

ED trend assessment in the regression equation shows that the slope assessment (b1) slope value (b1) is 0.004–0.005. Even though the slop value is <1, it is positive, so there is an increasing trend. Spearman's Rho shows that five rainfall stations have decreased with a Z score of −0.473 to −0.015. This result rejects H_o and accepts H₁, which means that the ED distribution data at weather stations have a strong relationship, and the data has an increasing trend. The test results also explain that 14 stations are trend increasing and 5 stations are trend decreasing. Therefore, there was a 73.68% increasing and a 26.32% decreasing monthly ED.

Based on Figure 5, the spatial distribution of ED shows that the most erosive occurs in November. This is indicated by gradations of red and identified potential hydrometeorological disasters. However, at the Talang Empat and Padang Harapan rainfall stations, there is a yellow gradient in the West and Southeastern locations of the Air Bengkulu watershed. Meanwhile, in December, the red gradient is concentrated in the downstream part of the watershed. January and April heterogeneous spatial distribution of yellow, red, and red colour gradations concentrated in the North and Southeast of the Air Bengkulu watershed. The lowest ED is in June, July, August, and September is shown in blue gradations.

Fig. 5

It can be stated that rainfall erosivity occurs in erosive periods in November, December, and April (NDA). This implies that the erosion of rainfall is directly proportional to rainfall. Meanwhile, the erosive rainfall erosivity is lowest in June, July, and August (JJA). In the months of rainfall events can have a direct influence on rainfall erosivity. However, the difference in the occurrence of rainfall gives rise to the characteristics of typical precipitation (Diodato, Bellocchi 2012). So that the value seems overestimated or underestimated to the value of rainfall erosivity. It has been revealed data information that the erosion of rainfall is underestimated at the rainfall weather station in Karang Tinggi, there is a low value of rainfall erosivity caused by inconsistent indications of rainfall recording by observers or the occurrence of errors in the equipment used.

The occurrence of rain is influenced by several factors, including latitude, altitude, and distance from the sea (Terassi et al. 2020). The Air Bengkulu watershed has a fairly diverse topography. This topographical diversity results in heterogeneous physical conditions between regions and is directly affected by the western part bordering the Indian Ocean. The heterogeneity that occurs at the study site causes an uneven distribution of rainfall events. Therefore, it is necessary to instal rain stations that are evenly distributed and represent rain events in various topographical forms.

Therefore, it is necessary to directly integrate watershed management with rainfall characteristics and pay attention to rainfall erosivity, triggering potential erosion and land criticality disasters. Therefore, management in soil and water conservation planning pays attention to this month. This is done by considering changes in the watershed landscape towards land use activities, including forest (production, protection, and conservation), agriculture, plantations, mining. Furthermore, the information which is generated through this study compiling with other supportive, data will facilitate towards taking up long-term flood/erosion protection measures at the research site along the Padma river (Billah 2018). that the slightest erosion information on the variables studied can be used as material for land conservation planning in the watershed by policymakers.

It is necessary to do conservation because of changes in the watershed landscape to reduce the danger of erosion in the monthly period of rainfall erosivity in November. In accordance with Arsyad (2012) the erosion rate that will occur can be estimated, and the erosion rate can still be tolerated. In that case, a policy on land use can be determined. However, in recent decades, land changes in the upstream watershed have triggered runoff on land without ground cover vegetation (Letsoin et al. 2020). Soil conservation measures with the application of wise land use, especially on steep slopes and riverbanks with lots of plants using candlenut (aleurites moluccanus), jackfruit (artocarpus heterophyllus), breadfruit (artocarpus heterophyllus), and durian (durio zibethinus), can improve the quality of the watershed landscape with a sustainable agroecosystem plan scenario applying organic fertilisers, multi-crop options, small mounds for sediment trapping and water harvesting with small ponds (Barchia et al. 2020). Soil conservation and land-use measures are implemented to reduce the rate of erosion to be less than the erosion rate that can still be tolerated.

Furthermore, when compared with monthly rainfall erosivity studies, the findings, such as in the Suketi River watershed at Himalaya, show that farmers know erosive months. That is to support in terms of best agricultural technique management (Singh, Singh 2020). Like agriculture in the Wei watershed in China, knowledge of rainfall erosivity information can help manage erosion prevention in sustainable agricultural systems (Zhang et al. 2021). It is strengthened by the findings made by (Duulatov et al. 2021) in improving soil conservation to ensure the sustainability of future agricultural products with operational management and agricultural planning. Then, the rainfall erosivity is mapped in mainland Africa as material for policy planning in compiling land and water resources management. (Fenta et al. 2017).

The results showed that the rainfall erosivity trend with the linear regression equations and Spearman's Rho test was significant with a significance level of 0.001. However, after doing the Mann Kendall test, it is insignificant. This rainfall data to support this research is incomplete and discontinuous. The incompleteness of this precipitation data can also be used as a separate study related to less efficient data inventory. The quality of precipitation data is very important, if there is a lot of unfilled rainfall data, it will not be good to support research related to rainfall erosivity. This causes uncertainty in determining the trend of rainfall erosivity and rainfall density. Trend analysis is an analytical method aimed at estimating or forecasting future erosivity of (Ma et al. 2014, Schmidt et al. 2016, Vijith, Dodge-Wan 2019, Zhu et al. 2021).

Therefore, to forecast well, it requires a wide variety of information (data), which is quite a lot and observed in a relatively long time of rainfall data and associated with several variables that affect rainfall. The analysis results show the fluctuations that occur and what factors influence it to estimate the rainfall erosivity pattern for spatial and temporal scales (Diodato, Bellocchi 2012). Therefore, in calculating rainfall erosivity, high temporal resolution precipitation data is needed at an intensity of 30 min for an extended period (Panagos et al. 2017). It is suggested that the period for rainfall erosivity data is at least 22 years (Panagos et al. 2016a, b).

Theoretically, in time series analysis, the most determining factor is the quality or accuracy of the information or data obtained and the time or period the data was collected. The more data that is collected, the better the estimation or forecasting will be obtained. Conversely, if less data is collected, the estimation or forecasting results will get worse. Estimating monthly rainfall erosivity based on data is problematic and highly biased since rainfall intensity is usually inconsistent (Nearing et al. 2017, Lee, Hsu 2020). However, the findings of this study explain that rainfall affects the rainfall erosivity which indicates a monthly erosive period and is prone to hydrometeorological disasters. (Baumann et al. 2002, Hanel et al. 2016, Djalante et al. 2021, Duulatov et al. 2021).

Although the two statistical analyses show significance, however, accuracy results show an insignificant trend in the Men Kendall test. The same research was carried out in Johor Bahru, where there was a decline, and there was no significant trend during the study period (Mohtar et al. 2015). However, a map of the incidence of soil erosion that can cause soil loss from hillslopes by water, A subsequent study recently Nearing et al. (2017) highlighted rainfall erosivity trend noted that watersheds in the American Southwest show significant under-predictions of erosivity. However, Zhang et al. (2010) reported a significant increasing trend in rainfall erosivity although an insignificant trend in the total rainfall amount was detected. Then, the same thing is also shown that the Suketi River Western Himalayan catchment watershed, India shows that no trend shows a significant value (Singh, Singh 2020). Therefore, in general, it is necessary to assess the rainfall erosivity in this regard due to changes in the global climate. It is also because of climate change in the United States to update maps and information regularly. The rainfall erosivity values for the Eastern United States were calculated using data from 1936 to 1957 (Yin et al. 2017).

The results showed that the spatial distribution of the Air Bengkulu watershed, the rainfall erosivity of the Northeastern, is an erosive area. However, It has been revealed that the resulting interpolation map tends to be rough and centred on the weather station. This is due to limitations in the number of weather stations, installed weather stations that are not in accordance with the physical conditions of the area, limited installation costs and equipment maintenance that requires continuous observers.

Geographically, the upstream watershed area with a slope of 45% is located on the west side of the Bukit Barisan mountains range along the Sumatra island route. Therefore, this area is most vulnerable to erosive processes. Areas that receive more rainfall as an orographic rainfall climatic phenomenon (Bagwan 2020) increase evaporation at an altitude of 100 m. The surface temperature will drop by 0.6°C. Spatially, the IDW method best explains the monthly rainfall erosivity based on weather station plots. however, it has been revealed that the findings of the Air Bengkulu River Basin in the middle show an underestimating result because the colour gradation of the map presented is drastically rough.

This shows that further studies are needed as to why this is so, namely by adding weather station plots. so it was revealed that the Karangtinggi Weather Station underestimated the results of interpolation of rainfall erosivity. However, this research is still not related to how topographical parameters can affect the value of rainfall erosivity. The colour gradation of the middle of the watershed shows an underestimating value. Furthermore, the map produced at the research site is satisfactory enough to determine the distribution of the rainfall erosivity. Thus the findings of the resulting map can provide a reliable estimation of the rainfall erosivity index and an essential prerequisite for the successful identification of areas at risk of soil erosion where soil conservation practices are required. (Borrelli et al. 2016).

The upstream area of the Air Bengkulu watershed is an area that influences the interaction of rainfall erosivity events downstream. Changes in the landscape of the upstream watershed causing heavy rainfall, causing runoff to carry material over the ground. Deforestation of the upstream watershed is used for mining activities and changes in land cover for socio-economic activities such as the formation of community forests, plantations (Andriansyah, Mustikasari 2011, Supriyono et al. 2017, Satmaidi et al. 2018, Gunawan et al. 2020). The Value 80% of Rindu Hati's protected forest area has been turned into agricultural and plantation land (Belladona, Nasir 2019). Meanwhile, the downstream area of the Air Bengkulu watershed states that the result of the erosion process is sedimentation which causes the river flow to experience a vast inundation impact due to flooding, such as the 2019 Bengkulu Flood (Mase 2020).

In other research locations, the rainfall erosivity is influenced by geographical conditions. For example, in the Wei River watershed in China, it is highlighted that spatial differences change under the influence of topography and geomorphology (Zhang et al. 2021). The coefficient of variation is usually higher in upland areas than in lowland areas, indicating that as elevation increases, the spatial variability of regional rainfall in the horizontal dimension increases, and the erosivity of rainfall changes drastically. Research conducted by Borrelli et al. (2016) of rainfall erosivity trends can determine the spatial distribution of research locations for both soil and water conservation planning.

The findings of the ED trend between 0.26 MJ · ha⁻¹ · h⁻¹ and 1.59 MJ · ha⁻¹ · h⁻¹ were insignificant. The pattern of ED increased by 73.68%, and the pattern decreased by 26.32%. An ED value >1 means that rainfall causes erosivity to the rainfall and triggers an erosive month (Oliveira et al. 2013, Mondal et al. 2016). In addition, areas with high rainfall erosivity and ED may be associated with landslide activity and risk of soil erosion (Panagos et al. 2016a, b). The upstream watershed, with its soil slope and geological conditions, may be more conducive to soil erosion than landslide activity influenced by rainfall and rainfall erosivity. However, the downstream watershed has erosive rainfall, making this area vulnerable to floods and sedimentation disasters (Mase 2020).

Important information about the rainfall erosivity and the ED is the time of year at the maximum and minimum points. The same rainfall intensity produces different effects on soil erosion. It directly depends on different factors, such as land cover in a watershed that changes throughout the month (Vijith, Dodge-Wan 2019). Knowledge of the highest ED is essential for management practice. It allows optimisation of management measures for soil and watershed conservation (Andriyani et al. 2017, Nasir et al. 2020). The spatial distribution of monthly rainfall erosivity in the catchment area has been presented in Figure 5, which shows the spatial distribution of ED. The intensity of rain erosivity occurs in November, December, April, and the lowest is in June, July, August. Based on the data, selective erosion control measures can be done, such as reduced tillage, leaving crop residues in agricultural areas, changes in crop or crop rotation to attenuate the impact of rainfall on soil and vegetation by increasing the soil.

The comparison of the study of rainfall ED trend and spatial distribution can significantly determine the location of the disaster. The conversion of land from rubber plantations to oil palm plantations has led to deforestation (Barchia et al. 2020). Changes occur in the upstream watershed of the Air Bengkulu. Therefore, the high rainfall in April 2019 was flooded (Mase 2020). This proves that rainfall events affect. According to the facts in the field, it shows that the ED of monthly rainfall can explain the vulnerability of erosive processes and hydrometeorological disasters.