Estimating the Useful Life of the Sempor Reservoir Using Erosion Modelling

The Sempor Reservoir is located in Kebumen District, Central Java, Indonesia. The Sempor Reservoir catchment area is included in the South Serayu Mountains zone. Physiographically, the South Serayu Mountains are included in the anti-clinorium zone, which is a series of structural hills (Van Bemmelen 1949). The Sempor Reservoir Catchment Area includes the Karangsambung Formation, Penosogan, Halang, Waturanda and Watuaranda Tuff.

The Sempor Reservoir has an inundation area of ±2.3 km² and a catchment area of 40.63 km².

This reservoir supplies agricultural irrigation with a service area of 6,478 hectares and drinking water, generates hydropower electricity, controls flooding from the Jatinegara River and is utilised for tourism and inland fisheries (Fig. 2).

Fig. 2.

Land use of the Sempor Reservoir Catchment Area is dominated by forests, both conservation and plantation forests with a fairly homogeneous pattern. Meanwhile, the central part of the catchment area has quite complex land-use variations. Paddy rice and settlements tend to dominate in the western part, which is a colluvial plain. Meanwhile, mixed cropland and dryland farming dominate in the central and northern parts, which are a complex of slightly sloping to steep hills. Mixed cropland in the Sempor Reservoir Catchment Area is generally planted with bamboo, banana, coconut and papaya. Meanwhile, dryland farming is dominated by cassava.

The data used in this research were obtained from official institutions and field surveys, as described in Table 1.

The USLE model was employed to estimate potential erosion rates by integrating four influencing parameters to simulate erosion in a catchment. The parameters used in this study are limited to only a few years due to the unavailability of data from the responsible authorities. The USLE was computed using the following formula: Ea=R×K×LS×CP$${\rm{Ea}} = {\rm{R}} \times {\rm{K}} \times {\rm{LS}} \times {\rm{CP}}$$ where: –

Ea is erosion potential (t·a^-1),

Rainfall erosivity (R) plays a crucial role as the primary driver of erosion in that it directly affects the breakdown of soil aggregates (Oliveira et al. 2012). To spatially analyse the R factor in the Sempor Reservoir Catchment Area, data were obtained from four rain gauge stations: Kedungwringin, Sampang, Sempor and Somagede. In addition, the recorded rainfalls were analysed temporally on a monthly basis from 2009 to 2020 to determine the average condition of rainfall in the study area. The Thiessen polygon method was used to determine the station’s area of influence based on rain gauge positions and observed rain depths. This method is particularly used in areas with low rain gauge network density (Goovaerts 2000). The R factor was calculated using the Bols equation. According to the Indonesia Ministry of Forestry (2009), the equation is considered representative as it has a good correlation to the incidence of soil loss in Java Island. The formula is suitable for use in areas that have a wet climate. R=6.119×P1.211×D−0.474×MaxP0.526$${\rm{R}} = 6.119 \times {{\rm{P}}^{1.211}} \times {{\rm{D}}^{ - 0.474}} \times {\rm{Max}}{{\rm{P}}^{0.526}}$$ where: –

P is the average monthly rainfall (cm),

Soil erodibility (K) is defined as the reaction of soil particles to the process of detachment and transport by raindrops or runoff (Renard et al. 1997). It is difficult to determine soil erodibility due to the numerous influencing factors involved, including the spatial variations and dynamics of soil properties and human activities (Yang et al. 2018). For the Sempor Reservoir Catchment Area, the K factor was determined based on landform units from which soil samples were collected.

The landform units were demarcated using pedogeomorphological approaches, which describe the strong correlation between landform and soil morphological properties (Christanto et al. 2018, Reddy et al. 2003). For instance, soils on sloping and flat terrains are formed through different processes (Gerrard 1981), resulting in divergent soil characteristics, particularly permeability (Putri et al. 2017). Besides, landforms are the best reference and indicator in areal delineation (Park, Burt 2002) due to their low temporal variations (Keshavarzi et al. 2019). The K factor was determined using the formula from Wischmeier and Smith (1978). 100K=[ 2.713M1.14×10−4×(12−a) ]+[3.25×(b−2)]+[2.5(c−3)]$$\matrix{ {100{\rm{K}} = \left[ {2.713{{\rm{M}}^{1.14}} \times {{10}^{ - 4}} \times (12 - {\rm{a}})} \right] + } \cr {[3.25 \times ({\rm{b}} - 2)] + [2.5({\rm{c}} - 3)]} \cr } $$ with M (particles percentage), M=(% slit +% sand )×(100%−% clay )$$M = ({\rm{\% slit}} + {\rm{\% sand}}) \times (100{\rm{\% }} - {\rm{\% clay}})$$ where: –

a is organic matter content (%),

Slope configuration substantially determines the erosion rate as it influences the quantity and velocity of water flows, the major driver of erosion, which segregate soil particles (U.S. Department of the Interior 2006). Slope data were derived from the country’s digital elevation model (The National DEM). LS values were classified according to Regulation No. P.7/DAS-V/2011 issued in 2011 by the Indonesia Ministry of Forestry (2011) on Technical Guidelines for the Standard Operating Procedure (SOP) System for Flood and Landslide Mitigation.

The crop management (C) factor shows the effect of vegetation presence and soil surface conditions on total soil loss. The soil conservation (P) factor represents human interventions that can affect erosion processes. Information on CP was derived from land-use data (Asdak 2020). The data used are land use in the year 2020 sourced from the Indonesia Geospatial Information Agency (BIG). CP values of different units of land use were classified according to Arsyad (2010). This classification is the one most closely resembling land use in the study area.

The fraction of solid particles that are transported into the reservoir and deposited at the bottom can be identified using the TE formula. After studying various TE approaches for several reservoirs on Java Island, where the Sempor Reservoir is, Susilo (2001) found that the modified Brune formula had the smallest difference index when compared to the observed data. Therefore, it is also the most suitable for the Sempor Reservoir.

The amount of sediment entering the reservoir, termed sediment yield (G), strongly depends on total soil loss in the catchment area. To determine G, erosion potential calculated by using the USLE was multiplied by the SDR. In a study by Olii et al. (2018), the Sempor Reservoir had a mean SDR of 0.2, with high accuracy (R² = 0.96) and a low rate of error (0.2%), which can thus be used for further analysis. Table 2 shows the formulas used for TE (modified Brune formula) and G.

where: –

C is the reservoir’s capacity (m³),

Interpolation produces simulated or predicted values, which should be compared with the data measured in the field (observed values). In this research, the comparison was made with cross-validation by removing some sample points (observed values) and using the remaining sample points to project the value at the location of the removed ones (predicted values) (Chang 2007). The recommended tests for hydrological modelling are simple regression analysis or coefficient of determination (R²), Nash–Sutcliffe efficiency (NSE) and mean absolute percentage error (MAPE) (Moriasi et al. 2007, Tian et al. 2018), as follows: R2=[ (n)(∑ op)−(∑ o)(∑ p) ]2[ (n)(∑ o2)−(∑ o)2 ][ n(∑ p2)−(∑ p)2 ] NSE =1−∑i=1n(oi−pi)2∑i=1n(oi−o¯i)2 MAPE =100n∑i=1n| oi−pi |oi$$\matrix{ {{{\rm{R}}^2} = {{{{\left[ {({\rm{n}})\left( {\sum {{\rm{op}}} } \right) - \left( {\sum {\rm{o}} } \right)\left( {\sum {\rm{p}} } \right)} \right]}^2}} \over {\left[ {({\rm{n}})\left( {\sum {{{\rm{o}}^2}} } \right) - {{\left( {\sum {\rm{o}} } \right)}^2}} \right]\left[ {{\rm{n}}\left( {\sum {{{\rm{p}}^2}} } \right) - {{\left( {\sum {\rm{p}} } \right)}^2}} \right]}}} \cr {{\rm{ NSE }} = 1 - {{\sum\nolimits_{{\rm{i}} = 1}^{\rm{n}} {{{\left( {{{\rm{o}}_{\rm{i}}} - {{\rm{p}}_{\rm{i}}}} \right)}^2}} } \over {\sum\nolimits_{{\rm{i}} = 1}^{\rm{n}} {{{\left( {{{\rm{o}}_{\rm{i}}} - {{{\rm{\bar o}}}_{\rm{i}}}} \right)}^2}} }}} \cr {{\rm{ MAPE }} = {{100} \over n}\sum\limits_{{\rm{i}} = 1}^{\rm{n}} {{{\left| {{{\rm{o}}_{\rm{i}}} - {{\rm{p}}_{\rm{i}}}} \right|} \over {{{\rm{o}}_{\rm{i}}}}}} } \cr } $$ where: –

n is the sample depth data,

The bathymetric survey results in 2023 were used to determine the reservoir’s capacity using the Frustum formula, as follows: V=13×H×(An+A(n+1)+An×A(n+1))$${\rm{V}} = {1 \over 3} \times {\rm{H}} \times \left( {{{\rm{A}}_{\rm{n}}} + {{\rm{A}}_{({\rm{n}} + 1)}} + \sqrt {{{\rm{A}}_{\rm{n}}} \times {{\rm{A}}_{({\rm{n}} + 1)}}} } \right)$$ where: –

H is the contour interval (m) and

This capacity was later used to calculate the remaining useful life of the reservoir (ΔC) using the Gill (1979) approach, as follows: ΔC=G×TE×Δtγ$${\rm{\Delta C}} = {{{\rm{G}} \times {\rm{TE}} \times {\rm{\Delta t}}} \over {\rm{\gamma }}}$$ where: –

ΔC is the remaining dead storage capacity at the present time (m³),

The sediment’s specific gravity (γ) was determined by laboratory tests. Therefore, the sedimentation rate can be defined by dividing between G and γ. ΔC represents the dead storage capacity known through the echo sounding survey conducted in the year 2023 due to limited data of initial reservoir capacity and sedimentation rates from related institutions.

Table 3 shows the rainfall erosivity calculations based on the Thiessen polygon method with four rain gauge stations in the Sempor Reservoir Catchment Area. Kedungwringin, Sampang, Sempor and Somagede stations showed rainfall erosivities of 2,467.21, 2,437.48, 2,176.53 and 2,828.84 MJ mm ha^-1 h^-1 a^-1, respectively. Rainfall tends to have a maximum destructive impact from January to March and from October to December, during which the erosivity is substantially higher than that during April–September.

Also, it was found that rainfall erosivity has a corresponding pattern to monthly rainfall. As seen in Figure 3, the darker the blue colour on the chart, the higher the rainfall erosivity. The high values are at the beginning and end of the year, and the lighter shows the lowest at midyear. Moreover, it indicates that monsoons control the rainfall in the catchment area. Like most regions in southern Indonesia, the catchment area sees one peak of rainfall from November to March, as influenced by the humid west monsoon, and one trough from May to September due to the dry east monsoon (Aldrian, Susanto 2003). Wheeler et al. (2005) explained that Java Island receives the highest rain from November to March due to convective rain from the formation of cumulonimbus clouds.

Fig. 3.

Morphologically, the catchment area originates from structural landscape. This landscape comprises river valleys, colluvial plains, lower, middle and upper slopes of structural hills, and ridges or hillcrests. Each landform unit in the catchment area represents one K index.

Table 4 shows the K index values for existing landforms and their determinants, including particle percentage (M) and numerical codes for organic matter content (a), soil structure (b) and soil permeability (c). The soil erodibility value in the Sempor Reservoir Catchment Area was derived from Fitryady (2008) and Rachma (2019) that used 6 soil samples on each landform. The landform map and soil sampling points are visualised in Figure 4.

Fig. 4.

Structural hillcrests had the lowest K value. One of the reasons is its substantially lower M value than other landforms. Low M values can be linked to the high clay content in the soil texture. Clay particles have a high adhesion capacity, making them stick tightly together to form large solid aggregates that are resistant to detachment and movement (Jiang et al. 2020). By contrast, colluvial plains and the middle slope of structural hills had the highest M values. Both landforms have the most easily eroded soils in the catchment area due to their low clay and high silt contents (Zehetner, Miller 2006).

Structural hillcrests had the highest organic content (a value). More organic matter increases soil aggregation. This will improve infiltration capacity and decrease surface runoff, reducing the release of soil particles due to raindrops and runoff (Yang et al. 2018). High organic matter in this landform is also attributable to the high clay percentage in soils. There is a positive correlation between organic matter and clay (Sirjani et al. 2019) because clay particles have a large surface area to absorb more organic matter (Qu et al. 2023). Conversely, soils with less clay have lower organic content, such as colluvial plains and the middle slopes of structural hills. However, the same case does not apply to the river valleys in the catchment area. Even though their soil texture has a lower clay percentage than silt and sand, they have nearly as high organic content as hillcrests. The main reason is that river valleys are the potential site for depositing nutrients from sediments entrained by the flow (DeLuca et al. 2013).

Soil permeability differs across the five landforms and shows no regular pattern. The middle slope had the highest b value at code 5, meaning that soil particles allow water to pass slowly at about 0.5 to 2 cm·h^-1. Soils with a high proportion of silt have low permeability and resistance to detachment (O’Geen 2006). By contrast, the lower slopes of structural hills had the lowest b value, indicating rapid permeability. This can be linked to the landform’s high organic content because there is a generally positive correlation between soil permeability and organic matter (Mulyono et al. 2019). Some exceptions include river valleys and structural hillcrests that have organic matter in large concentrations but transmit water slowly. Table 5 shows variations in slope LS expressed as LS values. The Sempor Reservoir Catchment Area is dominated by 35–50% slope gradients, indicating hilly topography with a very steep composition. Moreover, it has no areas with flat terrain or 0–5% slopes. Very steep topography increases the proportion of extreme erosion in total erosion events. As a result, topsoil loses more organic matter quickly, leaving soils with low adhesion and high proneness to erosion (Zhu et al. 2022). The steeper the slope, the faster the sediment movement and the higher the erosion potential (Zhang, Yu 2023).

Land-use response to erosion is expressed as a function of vegetative cover or crop management (C) and soil conservation (P). Table 6 shows CP values as components of erosion potential in the catchment area. Vegetated land usually has a low C index because it is resistant to erosion, for some plant or crop species can affect rainfall partitioning and water flow through interception, infiltration, water absorption and by covering or protecting soil aggregates (Suprayogi et al. 2013). Accordingly, conservation forests were considered the most erosion-resistant land-use type because of the lowest C index (0.001). Plantation forests have a higher C (0.2) because they constitute the product of anthropogenic interventions that most likely cause increased runoff and soil loss (Xiong et al. 2019).

Mixed cropland has a lower C index than dryland farming because crops are planted in a higher density in the former than in the latter. Vegetation plays a crucial role in reducing raindrops’ kinetic energy and particle removal from soil aggregates (Satriagasa, Suryatmojo 2020). With wider crop spacing, a larger surface area is exposed to raindrops. Raindrops will fall directly onto the ground without being slowed down by parts of plants. Therefore, compared to forests with high vegetation density, mixed cropland and dryland farming produce more significant surface runoff due to low interception (Zhu et al. 2022). Further, settlements have the highest C value because they are generally not covered by vegetation. Without vegetation, no objects reduce the flow rate of surface runoff (Hidayah et al. 2022). Information on soil conservation (P) is often scarce or difficult to obtain as it combines detailed data from numerous sources on various scales (Patil 2018). Therefore, the unit of analysis for P calculation is frequently merged with the C index.

Nearly all paddy fields in the reservoir catchment area are traditionally terraced. This engineering can conserve soil better than non-terraced fields as it reduces the amount and speed of surface runoff, thus providing more opportunities for the ground to absorb more water (Arsyad 2010). The soil structure becomes stable with the destructive energy of surface runoff being dissipated (Sutrisno et al. 2011).

Another soil conservation measure is contour farming, i.e., planting crops according to contour lines, as found in non-irrigated fields in the catchment area. Contour farming creates buffer rows that slow runoff, and vegetation on the contour strips can filter and trap sediment effectively (Kuok et al. 2013). Nevertheless, traditional terraces are considered better for soil conservation because they increase the infiltration rate and dissipate the destructive energy of runoff (Idjudin 2011).

Figure 5 presents the distributions of R, K, LS and CP factors in the Sempor Reservoir Catchment Area. These data were overlaid to produce the annual erosion rate (Fig. 6). The 2011 Ministerial Regulation classifies erosion susceptibility into five classes according to erosion rates. Table 7 shows the areal percentage of each susceptibility level in the study area.

Fig. 5.

Fig. 6.

The useful life of the Sempor Reservoir was derived from actual and potential sedimentation rates, as presented in Table 11. The Sempor Reservoir was planned to have an effective age of 50 years since its construction, thus expected to perform optimally until 2028 (Julia 2017). After comparing the capacities, the estimation confirmed that the reservoir’s useful life or optimal performance would last until 2028 or according to the initial design life.

However, there are always room where predictions are not necessarily in line with actual conditions. Therefore, the potential sedimentation rate that assumes all sediments will be deposited at dead storage capacity was also calculated. In actuality, sediments settle throughout the bottom of the reservoir (Putra et al. 2019). Therefore, the potential sedimentation rate is probably slower, and the remaining useful life would be >0.5 years. On the contrary, the useful life can also end earlier than in the next 4.59 years due to the flow dynamics that cause sediments at elevations above dead storage capacity to flow more quickly towards it. Therefore, the actual sedimentation rate can be substantially faster.

The remaining useful life can also be >0.5 years. The reason is that this estimate does not consider sediment control structures as a conservation measure that can reduce erosion and sedimentation rates in the reservoir and its catchment area. However, the erosion potential produced using the USLE model is also possibly overestimated (Dariah et al. 2004) because it only sees rainfall as erosion control but does not incorporate its partitioning or conversion into runoffs (Kinnell 2016, Meinen, Robinson 2021). Therefore, future research should analyse the erosion potential more specifically by incorporating, for instance, hydrological studies and validation in the field.