Seepage Behavior of Earth Dams Considering Rainfall Effects

Infrastructure Research Center, K-Water Institute, 200 Sintanjin-ro, Daedeok-gu, Daejeon 306-711, Republic of Korea Department of Cadastre and Civil Engineering, Vision College of Jeonju, 235 Cheonjam-ro, Wansan-gu, Jeonju, Jeollabuk-do 55069, Republic of Korea Department of Civil Engineering, Engineering Research Institute, Gyeongsang National University, 501 Jinjudero, Jinju, Gyeongsangnam-do 52828, Republic of Korea


Introduction
e seepage through an earth dam or a rock-fill dam generally correlates with the reservoir water level of the dam.Abnormal seepage problems, such as piping in the core zone of the dam, can be detected by carefully analyzing the relationship between the reservoir water level and the seepage rate.According to the design standards of a dam in Korea [1], seepage barriers and V-notch weirs should be installed downstream from dams.e seepage rate through a dam should be measured and used as a basis of a seepage stability evaluation.Seepage barriers in various forms have been used to control seepage and to mitigate seepage problems in dams.Since the 1970s, rigid barriers consisting of unreinforced concrete, plastic concrete, deep-mixed soil cement, and jetted grout have been used extensively in new dam construction and for mitigation of seepage problems in existing dams [2,3].Seepage-induced piping is the most common cause of dam failures.In previous case studies, 46% of the failures of large dams can be attributed to piping [4][5][6][7].ASTM [8] defines piping as the progressive removal of soil particles from a soil mass by percolating water, leading to the development of channels.Seepage erosion occurs when the water flowing through a crack or defect erodes the soil from the walls of the crack or defect [9][10][11][12].erefore, seepage monitoring is the most important approach to preventing damage due to seepage problems [13][14][15].
Seepage that flows out of a rock-fill dam is typically monitored at the downstream toe, and the measured seepage flow inevitably includes rainfall-induced infiltration into the downstream shell as well as reservoir-specific seepage through a core zone.Subsequently, the rainfall should be excluded from the seepage rate to increase the accuracy of the measurement.Additionally, base flow estimations can be used to investigate the effects of rainfall on the relationship between the seepage rate and the reservoir water levels.It has long been a topic of interest in hydrology because of its importance in understanding hydrologic processes, which play a crucial role in water resources management [16][17][18].In recent years, it has increasingly been used as tracers to investigate groundwater-surface water interaction [19,20] and a separate base flow from storm runoff or annual stream flows [21].Figure 1 shows the components of the measured seepage flow out of a rock-fill dam considering the rainfall [18].During the dry season, water stored in the dam is removed by soil water drainage.ese processes proceed at different rates in time and space as well as not being readily quantified.e gradual depletion of discharge during periods with little or no precipitation constitutes the drainage or recession rate, which can be graphically presented as the recession curve [18].
To carry out dam safety analysis based on relationships between reservoir-specific seepage and reservoir water levels, the seepage flow measured at a V-notch weir of the downstream toe is adjusted to filter out rainfall-induced infiltration into the downstream shell.To address this problem, this study applied a method to filtering out rainfallinduced infiltration into the downstream shell by using digital filtering at two rock-fill dams in Korea.After which, the seepage behavior and permeability of the core zones were analyzed to determine the relationship between reservoir water levels and adjusted seepage flows.
e hydraulic conductivities of the core zones were also inversely analyzed through a conventional seepage analysis.

Digital Filtering for Calibrating Rainfall Effects
2.1.Digital Filtering Process.Although there is an effort [22] to reduce the subjectivity associated with traditional graphical methods of base flow separation in hydrology, their application is restricted to well-defined, single-peaked, and isolated hydrographs [23].However, the hydrographs measured in humid regions are generally continuous and consist of multiple peaks.In recent years, recursive digital filters have been developed and applied for base flow separations.e recursive digital filters are based on the principle used in signal processing by regarding the base flow and direct runoff as low-and high-frequency signals, respectively.
e digital filters provide an effective alternative to the graphical methods, as digital filters can be easily automated and are capable of providing reproducible results.Many researchers have proposed digital filters [24][25][26][27][28]. Tan et al. [29] compared the hydrograph separation characteristics between Nathan and McMahon [24] and Chapman and Maxwell [26] for a partially urbanized watershed.Nathan and McMahon's method which provided the more reliable separation result was adopted to separate the base flow from the total runoff hydrograph.
In this study, the digital filter proposed by Nathan and McMahon [24] was utilized for the separating of rainfallinduced infiltration from measured seepage through a rockfill dam.Mugo and Sharma [30] showed that the algorithm is capable of producing promising separation results for three humid tropical forested catchments with area from 0.36 to 0.65 km 2 . where during periods where there is no direct runoff, β d � filter parameter (�k), d � direct runoff filter parameter, and Q � measured seepage.

Recession Constant.
To filter out the measured hydrographs using a digital filter, the recession constant k of the study site needs to be estimated.Base flow recession is represented as follows: where Q 0 and Q t � base flows at time 0 and t and τ � turnover time of the groundwater storage, during a period when there is no recharge.In the following equation, a linear relationship between the groundwater discharge and groundwater storage, S, was assumed as follows: Taking the natural log on both sides of (4) yields e traditional approach which uses the semilogarithmic plot of a single recession segment is widely applied in 2 Advances in Civil Engineering estimating the recession constant.is approach plots the natural log of the measured runoff according to elapsed time and takes α as the minimum slope corresponding to the base flow portion of the hydrograph (i.e., the linear portion of the ln Q t plot).Vogel and Kroll [31] found that the traditional approach is able to provide a reliable α estimate by averaging the recession constant values obtained from an ensemble of individual hydrograph recessions.Sujono et al. [32] found that the semilogarithmic approach produces reasonable and comparable estimates with those obtained from more complicated methods such as the master recession curve method [24] and the wavelet transform method [33,34].As shown in Figure 2, once the α value is determined, the recession constant, k, can then be estimated as follows [35]:

Seepage Flow Monitoring
In this study, the relationships among rainfall, water level, and seepage measured from two rock-fill dams (Dam A and Dam B) were investigated.Dam A is a hybrid-type dam combined with a concrete section on the right side and a rock-fill section on the left side as shown in Figure 3. e dam is 53 m height and 496 m length in total, and the rock-fill section is 131.6 m long.e general information of the dam is given in Table 1.

Dam A.
e seepage monitoring device was installed downstream from the dam as shown in Figure 3. e collected seepage water behind the core zone is transported by steel pipe (300 mm diameter) to a seepage monitoring device.It consists of a V-notch weir and an ultrasonic-type sensor to measure the overflow height.
e measured seepage flow of Dam A during 1981 to 1985 is shown in Figure 4.It indicates that the seepage flow is strongly in agreement with the reservoir water level and that there are

Dam B.
Dam B is a rock-fill dam with a central core.It is 58 m in height and 330 m in crest length.e general information of the dam is given in Table 1.
e seepage monitoring device for Dam B was also installed downstream from the dam as shown in Figure 5.It consists of a V-notch weir and an ultrasonic-type sensor to measure the overflow height.Dam B has a problem with direct runoff due to rainfall flowing into the seepage monitoring system through the downstream of the rock-filled zone.Figure 6 shows the measured seepage flow from 2002 to 2007.
e seepage followed the reservoir water level during the dry season, but it increased sharply during the wet season (June to September).
e maximum seepage flow is 2,465 l/min with a reservoir water level of 103.7 m.In this case, it is difficult to know the seepage flow through the core zone.

Determination of Recession Constant.
Two cases which include the big rainfall events were selected to determinate recession constant, α, for Dam A during the study period.Figure 7(a) shows an event on August 29, 1981, in which the rainfall and seepage rate were recorded at 93.3 mm/day and at 193 l/min, respectively.Figure 7(b) shows an event on August 17, 1985, in which the rainfall and seepage rate were recorded at 69 mm/day and at 313 l/min, respectively.Even if the rainfall of the event in 1981 was larger than that in 1985, the seepage rate of the 1981 event was smaller than that of the 1985 event.
e reason for this discrepancy was that the reservoir water level in the event of 1985 (78.06 EL. m) was larger than that of 1981 (74.98 EL. m).According to the digital filter [24], the recession constant k � 0.843 is determined as summarized in Table 2.
In the case of Dam B, rainfall events of twelves cases were selected to filter out rainfall-induced infiltration into the downstream shell.Figure 8           Advances in Civil Engineering rate-time curves.As shown in Figure 8(a), the seepage rate of 1,206 l/min was recorded when rainfall was 63.4 mm/day and the reservoir water level was 106.1 EL. m.On the contrary, in the case of October 25th, 2007, a seepage rate of 188 l/min with a rainfall of 33.1 mm/day occurred even though the reservoir water level was almost the same (106.1 EL. m). is indicates that Dam B was largely affected by the rainfall into seepage rate compared to Dam A. e recession constant was estimated for each event as shown in Figure 8 and Table 3.It was then determined that k � 0.9925.

Calibrating Seepage Flow.
e measured seepage flows of Dams A and B were adjusted against rainfall-induced infiltration using ( 1) and ( 2) proposed by Nathan and McMahon [24].Figure 9 shows the comparisons between the measured and adjusted seepage flows for Dam A. e results are strongly in good agreement which indicates that the rainfall effects are very minor and the seepage flow that occurred through the core zone was also found to be correlated with reservoir water level.On the contrary, the comparisons of Dam B in Figure 10 show that the measured seepage flows were filtered out and the peak values were removed by the digital filters. is indicates that the seepage flow of Dam B was strongly affected by rainfall.Although the maximum measured seepage of Dam B was 2,455 l/min on August 14, 2007, the adjusted seepage rate was 152.5 l/min when rainfall effect was excluded at the same period.[36] had studied seepage through various earth dams with the heights of 5, 10, 20, and 50 m under steady-state conditions.

Prediction of Permeability. Chapuis and Aubertin
ese studies have been numerically analyzed with a twodimensional finite element method, and they suggested a simplified method to predict the seepage flow rate as follows: where Q � total seepage flow rate, k � permeability of core zone, Δh � the total head difference between the pond (constant head reservoir) and the toe of the downstream draining-filtering blanket, L max � the core width (horizontally measured) at the bottom elevation of the downstream draining-filtering blanket and corresponds to the largest core width, L min � the core width (horizontally measured) at the elevation of the pond surface and corresponds to the smallest core width below the pond surface, and α 1 ∼α 3 are constant coefficients.e values of parameters α 1 and α 2 depend on the range of Δh 2 /L as shown in Table 4.
erefore, the predictive equation ( 7) and α 1 ∼α 2 values in Table 4 were used to estimate permeability of core zones for Dams A and B.
Figure 11 shows the relationship between seepage flow and reservoir water level for Dam A. e permeability of the core zone for Dam A was calculated from (7).As shown in Figure 11, it was estimated at 8.5 × 10 −5 cm/sec and the watertightness of the core zone is judged to be fully secured so that they may serve as a seepage barrier.e relationship between seepage flow and reservoir water level for Dam B is shown in Figure 12.In the case of Dam B, adjusted seepage flow conducted in this study was used for predicting permeability of a core zone because seepage flow of Dam B was largely affected by rainfall and the permeability of the core zone was predicted as 2.7 × 10 −5 cm/sec.It indicates that the core zone is also judged to be fully secured.

Conclusions
In this study, a method to filter out rainfall-induced infiltration into the downstream shell using a digital filtering method was applied to two rock-fill dams.
e seepage   Advances in Civil Engineering behavior and watertightness of the core zones were analyzed by determining relationships between reservoir water level and adjusted seepage flow.Moreover, the permeability of the core zones for each dam was predicted through the conventional seepage analysis [36].
e following shows the summary of the consequent findings.
e differences between measured and adjusted seepage flow of Dam A was very small, and the effect of rainfall was found to be very minor.e seepage flow of Dam B was strongly affected by rainfall.e maximum measured seepage of Dam B was 2,455 l/min, and the adjusted seepage was 152.5 l/min when the rainfall effect was excluded at the same date.
From the comparisons with measured and adjusted seepage flow, a digital filtering method to filter out rainfallinduced infiltration was used for the purpose of effectively analyzing the seepage behavior of dams.
e seepage flow through the core zones of Dams A and B was found to be correlated with the reservoir water level.
is suggests that the seepage behavior of the core zone of both dams is in a stable state condition.Also, the permeability of the core zones for each dam was predicted as 8.5 × 10 −5 cm/sec and 2.7 × 10 −5 cm/sec, respectively.Also, the watertightness of the core zone of both dams is judged to be fully secured and so they may serve as a seepage barrier.
Finally, a catchment for a dam should be constructed in the inner body of a dam such as in Dam A because this will exclude rainfall effects to improve the accuracy monitoring of seepage flow.Advances in Civil Engineering

Figure 1 :
Figure 1: Components of seepage flow of rock-fill dams considering rainfall.

Figure 3 :
Figure 3: Aerial photo and seepage flow monitoring for Dam A.

Table 1 :
Summary of investigated dams.no effects due to rainfall.e maximum seepage flow is 328 l/min at the reservoir water level 78.8 m which is more than normal high water levels (76.5 m) and less than flood water level (80.0 m).

Figure 4 :
Figure 4: Time histories of (a) rainfall and reservoir water level and (b) measured seepage flow of Dam A.

Figure 5 :
Figure 5: Aerial photo and seepage flow monitoring for Dam B.

Figure 6 :
Figure 6: Time histories of (a) rainfall and reservoir water level and (b) measured seepage flow of Dam B.

Figure 9 :
Figure 9: Comparisons between measured and calibrated seepage flow rate for Dam A.

Figure 10 :
Figure 10: Comparisons between measured and calibrated seepage flow rate for Dam B.

Figure 12 :
Figure 12: Predicted permeability of the core zone for Dam B.

Figure 11 :
Figure 11: Predicted permeability of the core zone for Dam A.

8
shows representative seepage

Table 2 :
Recession constant values of Dam A.

Table 3 :
Recession constant values of Dam B.