Establishing a Simple-yet-effective Approach of Early Warning System for Storm-Induced Earth-Filled Dam-Break Cases in Data-sparse Region

Historically, the occurrence of dam-break cases has been proven to cause significant loss of life and economical damage. Apart from the catastrophic nature of dam-breaks, the absence of a robust disaster prevention system exacerbates the disasters that can be occurred in any situations. Considering the importance of such prevention system, this study is therefore aimed to propose an Early Warning System (EWS) to mitigate the impact of dam-break disasters. However, predicting the occurrence of such disasters is challenging, specifically in areas like Indonesia, where comprehensive data recording is lacking. While it may be difficult to predict the occurrence of a sunny day break, the storm-induced break is more predictable. In this study, a simple yet effective macro-based EWS is proposed for Earth-Filled Dam-Break Cases using a macro approach based on the Evacuation Clearance Time (ECT). By comparing the ECT value with the arrival time of the floods from the affected areas, additional evacuation time can be obtained, which will be used to determine the EWS. Cengklik Dam, as one of the oldest dams in Indonesia that was constructed in 1931, is then chosen as the case study. The proposed EWS for Cengklik Dam is given in three levels of warning indicated by the reservoir water level at +141.36 m, +141.40 m, and +141.45 m. With the proposed EWS, the results show that 100% of people are expected to reach the evacuation point safely. The case study shows that the proposed EWS can significantly reduce the risk impact of the dam-break events.


INTRODUCTION
Reservoirs are vital in Indonesia for managing water resources, serving various purposes such as irrigation, clean water supplies for municipal and industrial areas, hydropower generation, and flood control.Due to the increasing demand for water resources, the government has been constructing dams aggressively.Since 2015, Indonesia has built 65 earth-filled dams (Portal, 2021) and plans to build more in the future.Despite the benefits, the presence of a dam can pose a significant threat, particularly to downstream areas.Dam-break incidents occur when a dam fails partially or completely, causing a sudden release of enormous water and creating an uncontrolled inundation (Sun et al., 2014).Over the past decade, Indonesia had experienced two such incidents, one in 2009, when the Gintung Dam in Tangerang City collapsed, killing 99 people and flooding more than 10 hectares of downstream area (Tempo News Por-tal, 2021).This event prompted the government to implement better operation, maintenance, and monitoring systems for all dams in Indonesia.Another incident occurred in 2013, when the Way Ela Dam in Central Maluku Regency collapsed, injuring eight people, and severely damaging more than 470 houses, causing 5,227 people to evacuate to shelters for several weeks (Portal, 2013).Froehlich (2008) reported 71 dam-break events worldwide between 1876 to 1994, predominantly due to the overtopping and piping conditions.This significant number highlights the need for a strategic plan to prevent further losses due to dam-break events.
Several studies have focused on dam risk analysis to minimize threats.In 1988, a simple dam risk analysis was conducted using an empirical equation based on 24 dam-break events to estimate the loss of life by considering two factors, namely population at risk and warning time (Brown and Graham, 1988).However, this method is limited as it neglects several factors, such as dam-break failure, physical dam characteristics, downstream social and economic conditions, as well as evacuation processes.This method is based on the 24 dam-break events in the United States which occurred before 1988, there is a need to conduct further studies to ensure the relevance of this method for different places and time.
To address these limitations, a new method based on physical indicators, such as indicator-based dam risk analysis, called the True Risk Factor (TRF), was developed by Shi et al. (2009).The value of TRF can be obtained with a weighting system for several indicators, including dam structure factors, downstream conditions, and flood hazards due to rain.Li et al. (2018) and Ge et al. (2019) stated that other methods include adding indicators such as understanding the failure mode, building vulnerability, and type of downstream land use.A simple and precise method of determining the risk of a dam-break event in the form of quantitative parameters called Flood Risk Level (FRL) was proposed by Yudianto et al. (2021).FRL determines the level of risk into four index values by evaluating several factors such as flood depth, flow velocity, distance to shelter, people demography, and building types in the areas with minimal data availability, such as Indonesia.This approach successfully indicates the most vulnerable area during a dam-break.
Planning for a catastrophic dam-break event must include instructions with respect to the evacuation process through an Early Warning System (EWS).Unlike typical floods in adjacent areas, dam-break floods can flow through highly populated regions that have not experienced flooding before, depending on the topographical conditions.As a result, people living in these vulnerable areas may not be aware of the lurking dangers, and have low disaster preparedness (Yudianto et al., 2021).The EWS is crucial for reducing losses by rapidly generating and disseminating disaster-related information to affected individuals, communities, and organizations.Since dambreak events are highly unpredictable, predicting their risk accurately remains challenging (Sattar et al., 2011).
Dam-break events can generally be categorized into two types, namely sunny day and storminduced breaks.When a dam collapses due to overtopping failure (storm-induced break), predicting the dam-break event involves monitoring the reservoir water level, which will cross the top of the earth-filled embankment dam level.This allows dam operators and residents downstream to prepare and evacuate.Detecting the possibility of a dam-break event is extremely challenging in the case of sunny day piping failure, especially for dams without advanced instrumentation records.Even with complete instrumentation data readings, no standard reference exists to predict dambreach occurrences.Therefore, while predicting the occurrence of sunny day breaks remains challenging, storm-induced breaks are more likely to be predicted.
The available study on early warning systems (EWS) for dam-break events in data-sparse regions is limited.Previous study, such as those proposed by Huaizhi et al. (2012) and Su and Wen (2005), have proposed methods based on the seepage and deformation of dam body or hydrostatic pressure, calculated using the rescaled range analysis and fractal theory combined with a continuous automatic recording.However, these may not be applicable to dams without complete instrumentation records, which are common in Indonesia.
Another study has proposed an EWS model for a concrete arc dam based on hydrostatic pressure (Su et al., 2018).These studies did not include parameters to express community preparedness for downstream areas facing dam-break events.As a result, there is currently no method for determining an EWS for storm-induced dam-break events in data-sparse regions.This paper proposes a simple-yet-effective approach that determines an EWS based on the required evacuation time for downstream communities.Since dam-break events typically occur suddenly, people must have sufficient time to reach the evacuation shelter.The proposed approach is specifically designed for earth-filled dams and is validated using Cengklik Dam as a case study.

METHODS AND CASE STUDIES
The study proposed a method, as shown in Figure 1, that utilizes the Evacuation Clearance Time (ECT) to determine the timing of an early warn- ing system (EWS) for downstream communities in the event of a dam-break.The ECT is comprised of three components, namely Pre-movement Time (PmT), Preparation Time (PT), and Evacuation Process Time (EPT).By comparing the ECT value with the arrival time of floodwaters in the affected areas, the appropriate timing for warning communities downstream can be determined.The EWS is comprised of three warning levels, each corresponding to the reservoir's water level.
Comparing ECT values with flood arrival time data is useful for evaluating the risk of flooding in various villages.Assuming the flood arrival time is slower than or equal to the ECT, the village is considered at low risk or relatively safe.However, the village is at high risk when the flood arrival time is faster than the ECT value.An Extra Time Needed (ETN) parameter is introduced to determine which areas are at high risk.The ECT value for each village can be obtained by subtracting the flood arrival time from the ECT.A smaller or zero ECT value for a village indicates a lower risk.Furthermore, to obtain flood arrival data, a dam-break flood can be simulated assuming the dam collapses (hypothetical dam-break).This hypothetical dambreak can occur during overtopping conditions or when the reservoir water level surpasses the top of the dam during the smallest flood return period.On the other hand, assuming the dam has been verified to be safe against overtopping even during the Probable Maximum Flood (PMF) con-ditions, then the hypothetical dam-break can be assumed to occur when the reservoir water level is at its maximum condition.
The development of an Early Warning System actual storm-induced piping dam-break events may occur before the hypothetical dam-break point.Therefore, to consider this possibility, the EWS is planned to achieve several evacuation success conditions.The first or critical condition is that 100% of the affected population can successfully be evacuated.The second success condition is that 100% of the affected population can successfully be evacuated before the hypothetical dam collapses.The third or ideal condition is that 100% of the affected population evacuates further before the hypothetical dam collapses.In this study, the ideal condition is represented by 2.5 times the maximum ETN value.The constant value of 2.5 act as a safety factor that expectedly can accommodate several possibilities, such as residents who receive the warning late or the slow evacuation process.Considering that the proposed EWS is made with a time/ECT approach, the time required to achieve the three success conditions needs to be known.The equation used to find the time required to achieve each condition is shown In Equation (1) to Equation (3), respectively.Hence, these three values indicate when a warning should be given, e.g., to achieve the first success condition, a warning should be given with a value of maxETN hour(s) before the dam-break event.The passage describes the development of an Early Warning System (EWS) to address the risk of a piping dam break caused by a storm.The EWS includes several evacuation success conditions to account for the possibility of the dam breaking before the hypothetical dam-break point.The first success condition requires 100% of the affected population to be evacuated successfully.The second condition requires 100% of the population to evacuate before the hypothetical dam collapse.The third or ideal condition requires the population to evacuate 2.5 times the maximum ETN value before the dam collapses, providing a safety factor for potential issues such as delayed warnings or slow evacuation.The time required to determine when to issue a warning for each condition is calculated using equations (1) to (3).For example, to meet the first success condition, a warning must be given with a value of max(ETN) hour(s) before the dambreak event, as determined by the EWS's time/ECT approach.
Time needed to achieve the 1 st success condition Time needed to achieve the 2 nd success condition Time needed to achieve the 3 rd success condition It is necessary to establish a physical condition parameter for the dam to facilitate the creation of an easily implementable EWS.This parameter will serve as a reference for the dam administrator to issue downstream warnings.By conducting reservoir routing based on the hypothetical dambreak case, the Water Surface Elevation (WSE) curve before the dam-break event can be determined.Therefore, the WSE is chosen as the physical condition parameter.Combining the known WSE curve with the time required to achieve the first, second, and third success conditions will yield an EWS that comprises three warnings.The first warning is issued to achieve the third or ideal success condition, the second warning is for the second success condition, and the third warning is for the first or minimum success condition.It should be noted that the conditions for warnings and successes are inversely related.The first warning is given when the third success condition is met, while the third warning is given when the first success condition is met.In addition, Figure 2 shows the visualization of this process.

Evacuation Clearance Time (ECT)
This study employs a macro approach to determine the estimated time required for residents to reach a shelter safely in different disaster situations.The definition of PmT, according to Shi et al. (2009), is the time between the first sounding of a disaster alarm and the start of the movement by affected residents.Shi et al. (2009)   (e.g., houses, schools, high rise buildings).However, Shi et al. (2009) reported a 20 to 120 seconds range for PmT values, depending on the preparedness factor and location.This study used a PmT value of 120 seconds to account for worst-case scenarios.

Preparation Time (PT)
The second component of PT is the time needed for the affected residents to prepare before leaving the disaster zone (Lindell et al., 2020).Previous studies have indicated that PT generally includes tasks such as leaving work, commuting from work to home, gathering people for group evacuation, packing essential items, safeguarding property, and turning off utilities (Cheng et al., 2011;Lindell et al., 2020) .To consider these activities during evacuation preparation, Lindell et al.
(2020) empirically estimated the PT distribution, as shown in Figure 3.It also presumes that the PT distribution applies to the case study as no prior analysis has examined a comparable distribution for this particular location, to the author's knowledge.

Evacuation Process Time (EPT)
EPT is the last component, which represents the time the affected residents take to reach the safe zone or shelter from the start of the evacuation.
The EPT value is influenced by the distance from the affected zone to the shelter and the evacuation method.In regions with limited local institutional capacity, infrastructure, and standard operational procedures for evacuation, such as Indonesia, the evacuation process relies heavily on the affected residents due to the lack of facilities, infrastructure, and standard operational procedures.Consequently, this study simplifies the EPT variable by dividing the evacuation method into two categories by foot and motorized vehicles, including passenger cars and motorcycles.
The closest distance between the affected area and the shelter is divided by the walking speed to calculate the EPT for evacuating by foot, as shown in Equation ( 5).
Where EPTWalk is Evacuation Process Time by walking (hour), L is the distance from the affected area to the shelter (km), and V is the walking speed (km/hour).
The L variable for the formulas can be obtained by conducting field measurements or using Geographic Information System (GIS) software.Walking speed varies according to the age range, which is categorized as follows 1.08 m s -1 for children (4-14 years old), 1.27 m s -1 for adults (15-64 years old), and 1.04 m s -1 for the elderly (> 65 years old) (Shi et al., 2009).For the EPT variable with the evacuation method by motorized vehicle, the shortest distance between the affected area and the shelter is divided by the corrected vehicle speed, shown in Equation ( 6).Therefore, the EPT variable with the motorized vehicle evacuation method is calculated using Equation ( 7) (Cheng et al., 2011).While completing these formulas, several constants remain the same as in the original study.Assumptions based on the similar characteristics between the location in Cheng et al. (2011) and the ones in this study are made, such as Lavg is 4.5 m, Lo is 1.5 m, µ is 0.5, C is 0.5, and γ is 0.6.
Where EPTmv is evacuation process time for motorized vehicle method (hour), L is the distance from the affected area to shelter (km), V' is corrected speed (km hour -1 ), Uo is travelling speed when the capacity is null (km hour -1 ), Uo = γ * µ * Vo , Q is traffic capacity (vehicle second -1 ), Km is

Case Study
The Cengklik Dam, situated in the Boyolali Regency of Indonesia, is a homogeneous earth-filled dam constructed in 1931.Its geographical location is 110°43'58,22 "S; 7°31'1,11 " E, and it has a catchment area of 10.69 km 2 and was built on the Bengawan Solo River. Figure 4 depicts the Cengklik Dam's areal photographic, physical feature, and downstream area condition.The dam has a total length of 1,693 m, 14.5 m (the crest elevation of +142.44 m), and a 30 m broad crested spillway (crest elevation of +140.54 m or 1.9 m beneath the crest).The dam's storage capacity is 9.7 million m 3 , mainly serving 1,578 ha of irrigation area (Engineerig, 2020).However, as shown in Figure 4, the Cengklik Dam is located near urban areas, where residential areas and various public facilities, including the Adi Sumarmo International Airport, are the majority of land use types in the downstream area.This proximity of the dam to these high-risk areas increases the danger associated with its presence.
The hypothetical situation where the safety of the Cengklik Dam was examined by Engineerig (2020) and found to be secure against overtopping failure.However, the possibility of the dam-breaking due to storm-induced piping failure is considered.The scenario assumes that the dam breaks at the max-  5 and Table 1 to simplify the presentation.The estimated inundation area is approximately 3,617 ha, and the number of lives at risk is estimated to be 168,097.

RESULTS AND DISCUSSION
Determining the distance between the affected village and the shelter, which could be defined from field measurements or GIS software, is a crucial parameter in the EPT component.However, in areas like the Cengklik Dam, where data is scarce, using pathfinding tools algorithm is not feasible due to the unavailability of road network data (Lahoorpoor and Levinson, 2019).The study uses a simplified approach to address this issue by calculating the average distance between the furthest and closest distances of the affected village to the shelter, as stated in Equation (8).While a single average distance may oversimplify the analysis, it is still a useful representation of the situation in the area.
Where L is the average distance from the affected area to shelter (m), Lmax is the furthest distance    The study added the extra time required to evacuate 100% of citizens in the Bolon Village (max ETN value) to determine the third/critical warning reference time (1.32 hours), as shown in Table 3.For the second warning reference time (2.72 hours), the study used the longest time to achieve 100% evacuation before the dam-break occurred in the Bolon village (max ECT value).Finally, the first/ideal warning reference time (3.3 hours) was calculated by multiplying the extra time required to evacuate 100% of citizens by 2.5 (max ETN value times 2.5).These reference times were then correlated with the water level in the reservoir for the PMF condition, and the reference water level for each warning level is shown in column 2 of Table 3.The process of obtaining these reference water levels is illustrated in Figure 7.By implementing these three warnings as an EWS, the study predicts that 100% of the population affected by the Cengklik dam-break flood can be saved.

CONCLUSION
This study proposed and tested a novel method to determine an early warning system (EWS) in the context of the Cengklik Dam.The potential impact of a dam-break event at the Cengklik Dam could affect 58 villages, including critical infrastructure such as the Adi Sumarmo International Airport.
To determine the EWS, ECT values for each village were generated and compared to the flood arrival times.The analysis results led to the developing of a three-tiered warning system when the reservoir water level reached +141.36 m, +141.40 m, and +141.45 m.It should be noted that in real-world scenarios, when the water elevation has reached +141.36, it is possible that the water level may not continue to rise to +141.40 and +141.45 after the first warning.While the proposed EWS can ensure the safety of 100% of the affected population, further investigations are required to determine its applicability to different cases and locations.
The study has demonstrated the effectiveness of a simple, early warning system (EWS) in mitigating the risk of dam-break events.Nonetheless, further study is needed to refine the ECT calculation and examine its applicability in different settings, such as urban and rural areas, as well as to consider physical factors related to dam conditions.Additionally, future studies may need to incorporate the economic cost of disasters, which is a crucial aspect of disaster prevention planning.

Figure 1
Figure 1 EWS establishment methodology

Figure 2
Figure 2 Implementation of EWS visualized

Figure 4
Figure 4 Cengklik Dam aerial view

Figure 5
Figure 5 Cengklik Dam-Break flood area map

Figure 7
Figure 7 Cengklik Dam early warning system parameter PmT is Pre-movement Time (hour), PT is Preparation Time (hour), and EPT is Evacuation Process Time (hour).
Cheng et al. (2011), Lindell et al. (2020), and Shi et al. (2009) al. (2020), and Shi et al. (2009)but modifies the PmT and EPT parameters to suit the Indonesian study location better.The present study refers to this estimated time as the ECT, representing the duration between receiving a disaster warning and safely reaching the shelter.Equation (4) proposes a formula for calculating the values of ECT, with further details about each index available in subsequent chapters.ECT = P mT + P T + M ax(EP T walk , EP T M V ) (4)Where ECT is Evacuation Clearance Time (hour),

Table 1 .
Affected villages from Cengklik Dam-Break

Table 2 .
ECT summary for Cengklik Dam-Break