Risk dynamics modeling of reservoir dam break for safety control in the emergency response process

Dam break is an accident that may heavily threat downstream residents’ life and property safety, especially in China. As revealed by accident investigation statistics, both flawed organizational behavior and inadequate downstream resident risk awareness have affected the safety risk of reservoir dams. Multiple information transferring mode and dynamic processes perform with the characteristics of social-technical systems. Based on the system dynamics approach, this study proposed a risk causation model aiming for factor interactions involving organizational, human, and technical system levels. The derived simulation model represented the historical risk evolution process of Gouhou reservoir in China and the rationality of the proposedmodel was verified. To further improve the efficiency of the organizational response and monitor real-time dam safety, a software tool called Dam Emergency Response Aids (DERA) was constructed to evaluate the potential safety benefits of risk control measures, and to overcome the defects of static emergency plans. By integrating relevant professional modules and data, the mobile application (APP) has been applied on the Jinniu Mountain reservoir dam in Nanjing of China and helped to maintain its excellent safety operation until now. It shows that the risk dynamics model proposed can improve the abilities of dam operating management organization for more effective responses under emergency circumstances.


INTRODUCTION
Reservoirs are water conservation projects that retain water, block floods, and regulate water flow in the flood season. Reservoirs can play an important role in flood control, irrigation, water supply, power generation, water source protection, etc., and their sizes vary greatly (Li et al. ). In China, artificial reservoirs are usually formed by constructing barrage dams at the narrow mouths of ravines or rivers. China is the country with the largest number of reservoirs in the world. There are 98,000 reservoirs of various types, with a total storage capacity of more than 930 billion m 3 , accounting for nearly 35% of the country's total surface runoff. Facilities are also an important part of the flood prevention and disaster reduction heavy engineering system (Wayne ). At present, with engineering technology development, China has entered the rank of countries with low dam failure rates. However, reservoir dam emergencies behave in a manner of a social-technical system. The law of dam break risk causation is more complex, and it heavily affects the dam break emergency response. In this case, traditional qualitative analysis of dam break accidents and derived regulatory revisions have made it difficult to accurately assess and prevent the impact of socio-technical system factor risks on dam safety (Sheng et al. ). Reservoir dam break accident is a dynamic operation process of a social-technical system with multiple information flows such as water regime, disaster situation, material transfer, and information interaction (Zhang et al. a, b). Its risk dynamics transmission process covers organizations, personnel, and dam hydraulic systems. The interaction of risk factors forms a high-order dynamic feedback loop. To model the evolution process to cover its time-serial dimension of the emergency decision-making process, it is significant for constructing a quantitative risk assessment tool on dam break prevention and crisis handling. problems by analyzing the information feedback, dealing with the dynamic structure and feedback mechanism between the factors of the complex system, to obtain the overall cognition and problem solving of the system. In the field of system safety, system dynamics has been used as an important supplement to analyze organizational accidents and proposed safety policy in the field of aviation, astronautics, and chemical industries (Bouloiz et al. ; Shin et al. ; Yu et al. ; Lu et al. a, b). Especially in view of the social-technical system, organizational accidents are increasingly being studied by using a system dynamics approach. This approach helps to model the risk interactions of organization safety with conceptual description, causation analysis, and time-domain simulation tools.
The risk archetypes are constructed from three basic building blocks: the reinforcing loop, the balancing loop, and the delay.

Feedback loop
The reinforcing loop refers to a particular behavior that encourages similar behavior in the future, and it corresponds to a positive feedback loop in the control theory.
As Figure 1

Modeling process
In this study, the critical risk factors embedded roots in reservoir dam routine and emergency operating processes.
The data supporting risk analysis include: • engineering assumptions grounded in organizational experience and accident investigation related to reservoir dam operating processes flaws; • behavior modes and safety features proposed in literature reviews, such as accident and risk models; • accessible safety data, such as system flaws and human error identified in accident investigation reports.
For the feedback causation-based dam operating risk dynamics modeling, the Causal Loop Diagrams (CLDs)

Accident investigation and data collection
In the reservoir dam break emergency decision-making system, based on the risk evolution process of the social-

RESERVOIR DAM OPERATING RISK CAUSATION MODELS
The system dynamics method is used to analyze the risk causation of the emergency decision system for reservoir dams in the form of the causal loop diagram. The causal loop diagram is a basic method to model the feedback structure of the risk dynamics. It captures the process of information transmission and feedback, that is, the process of a risk variation in the social-technical system, which affects the variable itself in turn through a series of causal relationships.

Safety performance level
From the view of safety performance level (SPL), a reservoir dam break is an emergency event under adverse environmental impacts such as rainstorm or earthquake. The impact of early organizational decision-making behavior on later ones presents a characteristic of path dependence.
The safety performance of the whole system mainly reflects the evacuation process of a risky population to a safe settlement, the process of death, and the effect of population evacuation rates on the total population transferred

Organizational management level
The organizational management level (OML) describes the social-technical structure of the reservoir dam   Even worse, some people returned to the danger area after they thought the risk had been eliminated. In order to reflect these different behaviors in the model, a variable called acceptance of evacuation is introduced to measure the extent to which the people at risk recognize the danger.
Meanwhile, the actions of other people may affect the individual's acceptance of the warning information significantly, and then stimulate him or her to evacuation soon.

Dam system level
As shown in Figure  This model shows that for emergency response organizations, the purpose of full decision-making and implementation is to make the residents' risk awareness   Table 1.

RISK EVOLUTION SIMULATION IN RESERVOIR DAM BREAK Simulation model and testing
Based on the theory of system dynamics, each typical behavior of system must be determined by a certain  Figure 8. It can be seen that the number of people under evacuation shows a peak corresponding to relevant orders, and gradually approaches zero over time (that is, the characteristic of target-seeking). This represents that the residents have arrived at their safe refuges, which is consistent with the conceptual expectation and historical experience.

Representation of historical accident process
After the parameter sensitivity test and variable validity check, the historical dam break event of the Gouhou reservoir in Qinghai Province described previously was chosen as a case to verify the model structure and variable definitions by represent the risk evolution in this accident. Based on the model as proposed in Table 1 and after the corresponding adjustment of the model parameters, the simulated behavior of critical variable people under evacuation is selected for result display partially, as shown in Figure 9.
According to the Gouhou dam break accident report and the on-site investigation, it was found that before the evacuation order was issued by the reservoir operating organization, the people evacuated in advance according to their perceived risk. The first peak of evacuation occurred around 180 minutes (21:50) after the sign of the dam break was found, which also showed that some people did not evacuate ahead of others because of limited risk awareness or insufficient understanding of the consequences of the    Table 2 and the simulation results using the variable number of fatalities as an example is shown in Figure 12.
As Figure 12 shows, the parameters reflecting organization decision-making, resident risk awareness and warning mechanism can be adjusted according to the scenario Intelligent organization, which will defer the dam break process and gain more time for evacuation. As a result, 99.41% of the people downstream can be saved in this scenario. Adjusting the parameters of community cohesion, individual alertness, and understanding the consequences of the flooding will enhance the resident risk awareness, as well as improve the efficiency of evacuation and the   Under the semi-quantitative support of above decisionmaking evaluation in medium-to long-term vision, the emergency response should cover these strategies as follows.
In the aspect of decision-making for issuing early

CONCLUSION
For an effective prevention and emergency response of reservoir dam break that may cause significant casualties and property damage, this paper reveals the dam operating risk dynamic mechanism and it models the general risk evolution process pervading in a social-technical system vision.
A quantitative system dynamics simulation model and a Importantly, professional module data including dam hydrology and rainfall status, early warning mechanism and flood evolution process are integrated into the simulation model.
The validity of the model was verified by representing the historical process of Gouhou reservoir dam break in Qinghai Province. The error between the simulated results and historical data is 2.1%. The non linear feedback mechanism for the risk dynamics can be explained from the time series dimension quantitatively.
Based on the safety control scenario simulation, this research helps to remind the safety responsible organizations to focus their investments on publicizing and downstream resident training, which will increase their awareness of risks, enhance public understanding of the evacuation process and risk pre-judgment criteria, and maintain trust in organization decision-making and community cohesion.