Necessity of integrated methodology for hazard mitigation of quake lakes: case study of the Wenchuan Earthquake, China

Abstract

The massive 8.0-magnitude Wenchuan Earthquake triggered huge landslides, avalanches, and debris flows that blocked rivers and created 34 important quake lakes, including the Tangjiashan Quake Lake on the Tongkou River. More than half of these lakes were identified to be of moderate or high hazard levels, so activities needed to be undertaken for hazard mitigation of potential flooding. This paper presents the mitigation processes of quake lake hazards, which involve various techniques such as rapid hazard analysis, scenario-based mitigation planning, and real-time forecasting of outburst flooding for implementation actions. The shortage of hydrologic and geological data and the nature of emergency situations raise substantial challenges in the hazard mitigation of quake lakes. This paper suggests a potential approach in dealing with quake lake hazards, which integrates the automatic monitoring network, hydrologic models, and hydrodynamic models with a comprehensive indicator for hazard levels. The necessity of improving the integrated methodology is highlighted.

Keywords

quake lake / hazard mitigation / dam failure / emergency / Tangjiashan

Introduction

Quake lakes are created by landslides, avalanches, and debris flows that run into valleys and block rivers with barrier dams during an earthquake. Since barrier dams are usually made up of loosely consolidated materials, such as soil, gravel, and broken stones, quake lakes create a high risk of catastrophic dam failure, endangering downstream lives and properties [1]. With successive inflows, the water level of a quake lake rises rapidly, creating overtopping or seepage flow in a short period of time; this causes the loosely consolidated barrier dams to fail under the coupled actions of fluvial and gravitational erosion. Meyer et al. [2] investigated 63 failed barrier dams and found that few of those barrier dams were stable; in detail, 22%, 44%, 50%, 59%, 83%, and 91% of them failed within 1 day, 1 week, 10 days, 1 month, 6 months, and 1 year after their formation, respectively. Moreover, in suddenly formed earthquake-hit regions, floods caused by dam failures occur without sufficient disaster preparedness, ruining downstream houses, farmlands, and infrastructures around the flow paths. For example, three barrier dams on the Min River and nine barrier dams on its tributaries were triggered by a 7.5-magnitude earthquake near the Diexi Town of Mao County, Sichuan, on August 25, 1933. The highest barrier dam with a height of 160 m at the Diexi Town collapsed 45 days after the earthquake and the resulting flood rushed down the valley for a distance of 250 km, killing at least 2500 people [3,4].

On May 12, 2008, the massive 8.0-magnitude (MS) Wenchuan Earthquake in Sichuan Province triggered 34 important quake lakes. Over half of them were in moderate or high risk of dam-failure floods to downstream inhabitants and rescue personnel. However, the complicated safety conditions and the widespread quake lakes made it extremely difficult to obtain the necessary information about these lakes and dams (hydrological, geotechnical, and water regimen data) to create an effective disaster mitigation plan. This paper presents hazard mitigation processes that integrate various data and techniques, including rapid hazard assessment, scenario-based mitigation planning, and real-time forecasting of outburst flood. With the substantial challenges raised by the lack of hydrologic and geological data as well as the nature of emergency situations, this paper seeks to highlight the necessity of improving the integrated methodology to support hazards mitigation of quake lakes.

Quake lakes triggered by the Wenchuan Earthquake

At 2:28 PM on May 12, 2008, a massive earthquake of magnitude (MS) 8.0 struck Sichuan Province, China, resulting in tens of thousands of deaths and hundreds of billions of RMB loss. The main shock and numerous aftershocks triggered many huge geohazards, such as landslides, rock avalanches, and debris flows. Nine rivers were dammed and 34 quake lakes with a potential of dam failure flooding were created. These quake lakes clustered approximately along a straight line, as shown in Fig. 1, forming several quake lake cascades. For example, within the 6 km long reach of the Tongkou River upstream from Beichuan City, the earthquake created its largest quake lake, the Tangjiashan Quake Lake, and three other quake lakes, as shown in Fig. 1. With successive inflows and frequent aftershocks, these quake lakes with loosely consolidated dams had posed an increasing risk of uncontrolled outburst floods in the coming flood season. (The flood season starts in June and ends in October and occupies 80% of annual precipitation.) The clustered/cascaded quake lakes could have induced a chain of dam failures, which would have resulted in exaggerated flooding, if it had not been dealt with in time.

Among the 34 quake lakes, the Tangjiashan Quake Lake was the largest one with a storage capacity of about 316€million m³ and a drainage area up to 3550 km². Its huge barrier dam had a maximum height of 120 m and a maximum width of 611.8 m perpendicular to the river, with a total volume of 20.37€million m³. The quake lake was only 6 km upstream from the seriously ruined Beichuan City, where almost 20000 people had been living, and 70€km from Mianyang City, the second largest city in Sichuan Province with a population over 1.3 million. Before its final man-controlled breaching on June 10, the lake water stage rose at a rate of about 1 m/day. A potential flooding disaster threatened more than 1.3 million people downstream. After the earthquake, it became urgent to assess and mitigate the hazards of quake lakes, especially for the hazard of the Tangjiashan Quake Lake because of its large storage capacity.

Integrated methodology

Quake lakes threaten downstream inhabitance because of the potential of barrier dam breakage and subsequent flooding. However, hazards assessment and mitigation planning are hindered by the lack of topographical, geological, and hydrological data of these suddenly formed lakes. Following the disaster of the Wenchuan Earthquake, detailed risk analysis of dam failure flooding was infeasible. Data from various sources could be used to facilitate carrying feasible analysis of quake lakes with the aim of giving guidance to disaster mitigation planning. Alternatively, a simplified approach that depends on key data could be used to analyze quake lakes. However, both of these ways might have difficulties in emergency conditions due to significant uncertainties. Therefore, an integrated methodology that utilizes various data resources and methodologies was adopted for a comprehensive analysis.

The framework of the integrated methodology is presented in Fig. 2, which demonstrates in time sequence of the necessary elements: data collection, hazards assessment, hazards mitigation planning, and implementation of those mitigation plans.

Data collection: The data collection unit formed the basis of quantifying the properties of a quake lake and of predicting its behavior. Three approaches, i.e., remote sensing inspection, on-site survey, and field monitoring, were adopted to collect data at various spatial scales. Remote sensing inspection provided overall information of a quake lake, including its location, topography, flooding area in the upstream, and flood pathway in the downstream; on-site surveys produced the geometrical and geological information about its barrier dam. The two approaches supplied the fundamental data for hazard assessment and mitigation planning. In contrast, field monitoring was used to obtain the real-time data of rainfall, reservoir inflow, lake water stage, and dam overtopping/seepage flow. The real-time data facilitated detailed planning of hazards mitigation and early warning of impending flooding.

Hazards assessment: Hazards assessment can be achieved through using either a simplified tabulation approach or a detailed risk analysis that covers the probability of dam failure, the amplitude and duration of flood, and the vulnerability of downstream inhabitants and infrastructures. In the emergency case where there are insufficient data and not enough time for a detailed analysis, a rapid matrix tabulation approach is preferred and adopted to classify the level of each quake lake hazard. This assessment allows the hazard level of each quake lake to determine each quake lake’s priority.

Hazards mitigation planning: The planning tasks involve the hydrological/hydraulic analysis of potential dam failure flood and the corresponding optimal design of mitigation measures. The hydrological/hydraulic analysis is also subdivided into the forecast of reservoir inflows, characterization of dam failure flood, and flood routing and inundation analysis downstream. To meet the requirement of emergency analysis, two approaches with different levels of complexity are applied to each task. The simplified or empirical approach is easy to use but has large uncertainties in its modeling. The alternative approach is more rigorous but also more time-consuming, and its accuracy depends significantly on the quality of the surveyed/monitored data. An effective way to improve the reliability of mitigation planning could be the combination of the two modeling results from the two different approaches.

Implementation of mitigation plans: The implementation of mitigation plans includes both engineering and non-engineering measures. The engineering measure is to excavate a drainage channel on the crest of the barrier dam, which is regarded as the most efficient way to lower the water stage and therefore to produce a man-controlled flood when the quake lake bursts. The implementation of non-engineering measures such as early warning to and evacuation of downstream inhabitants is closely related to the prediction of impending dam failure flood. Therefore, an integrated prediction based on real-time monitoring and numerical simulation is adopted to supply rolling updated data that helps implement specific planned measures prior to outburst flooding.

Hazards assessment and mitigation

Data collection

The shortage of geometrical, geological, hydrological, and other technical data of the sudden-formed lakes created by the Wenchuan Earthquake significantly hampered the disaster mitigation efforts. Both remote sensing inspection and on-site survey were carried out soon after the earthquake to quantify/qualify the properties of the 34 quake lakes [3,5–7]. To support effective mitigation planning of quake lake hazards, real-time hydro-meteorological data were collected through established monitoring systems [8]. These data were updated in a rolling way, monitoring the instantaneous state of the quake lakes.

Remote sensing inspection and on-site survey

Many government agencies and research institutions were engaged in field data collection through remote sensing inspection and on-site surveys. The inspected/surveyed data were used to characterize and then classify the quake lakes. Xu [6] characterized the 34 quake lakes in detail in terms of the reservoir storage capacity, the geometry of the dam body (such as volume, height, and thickness to height ratio), and the composition of dam materials and structure. Similar works were contributed by other researchers, including Cui et al. [3], Xu et al. [5], and Yang et al. [7]. According to Xu [6], the 34 quake lakes featured as follows:

(1) Reservoir storage capacity: This parameter is related to the properties of the drainage basin of the river that the barrier dam blocked. Among 20 dams that had the surveyed data of storage capacity, 14 lakes occupied a storage capacity of 1.0–10 million m³ and three lakes had 10–100 million m³. The Tangjiashan Quake Lake was the largest reservoir, having about 314 million m³ of storage capacity; the second largest reservoir, the Xiaojiaqiao Quake Lake, had a storage capacity of 20 million m³.

(2) Height: The height of each barrier dam was related to the volume of landslides, avalanches, or debris flows that blocked the river. Thirty-three quake lakes had their dam height recorded. Among them, 14 barrier dams had a height ranging from 30 to 70 m; 7 dams had a height of 15 to 30 m; and 6 dams were higher than 70 m, including the dam of the Tangjiashan Quake Lake.

(3) Thickness-to-height ratio: This ratio is an important index that describes the stability of a barrier dam. Among the 32 dams with these data, 8 dams, 6 dams, 2 dams, and 16 dams occupied a ratio less than 4.0, between 4.0 and 6.0, between 6.0 and 8.0, and greater than 8.0, respectively. Among them, 6 dams even had a ratio greater than 14.0.

(4) Composition of dam materials and structure: The barrier dams were generally made up of soils, boulders, or unbroken bedrocks. Accordingly, the dam structure can be distinguished into three types: single layer consisting of mostly soils or boulders; two layers consisting of soils or boulders in the upper part and unbroken bedrocks in the lower part; and three layers with soils in the upper, boulders in the middle, and unbroken bedrocks in the lower. The barrier dam of the Tangjiashan Quake Lake had a structure of three layers, while that of the Xiaojiaqiao Quake Lake had a structure of two layers.

Hydro-meteorological monitoring system

The monitoring systems collected real-time hydro-meteorological data for forecasting and early warning of potential outburst floods [8]. In a dangerous environment with frequent aftershocks, automatic, instead of artificial, monitoring systems are more feasible. Hence, automatic monitoring systems were designed and established on the basis of the existing network of hydrological gauging stations; many of which were seriously damaged in the earthquake.

Four principles were followed in the establishment of the monitoring systems:

(1) The existing gauging stations should be fixed, if damaged, or re-equipped as soon as possible to automatically monitor the rainfall in the catchment or water level of the lake. Video cameras should be installed to visually record the real-time state of the quake lake and its barrier dam.

(2) The spatial density of gauging stations should meet the requirement for carrying out hydrological process simulation and forecast. The measuring span should cover the possible maximum discharge and water level when the lake bursts.

(3) Secondary facility (or artificial system as an alternative) should be equipped to keep on continuously monitoring at important sites in case the primary facility fails.

(4) Evacuation passages should be planned at the sites where an artificial system is used.

Table 1 summarizes the automatic monitoring system for the largest Tangjiashan Quake Lake according to the above principles. Upstream of the barrier dam, there are two meteorological gauging stations (i.e., Piankou and Baishi stations); hence, three new stations (i.e., Youfang, Tumen, and Zhicheng stations) were added for a reasonable representation of rainfall in the catchment, as shown in Fig. 3. Meanwhile, two hydrological stations were installed at Zhicheng and the dam site, respectively, to monitor the rising water level resulting from the successive inflow. These hydro-meteorological data facilitate the forecast of lake inflow and potential outburst floods. Downstream of the barrier dam, two new hydrological stations (i.e., Tongkou and Xiangshui stations) were added to measure both water level and flow discharge. The two original hydrological stations (i.e., Beichuan and Fujiangqiao stations) were put into operation to measure water level only because the flow discharge could still be specified accurately from the stage-discharge curve. The four stations formed a monitoring system of potential flood routing in the Tongkou River and assisted in inundation analysis of the flooding area.

Hazards assessment

Immediately after the Wenchuan earthquake, rapid assessment of quake lake hazards was essential due to the urgency of the situation. However, assessment near the quake lake was difficult due to the damaged infrastructure of the area; roads were damaged and hillsides were unstable due to frequent aftershocks. As a result, in this paper, a simple but rapid matrix tabulation approach was developed instead of a detailed risk analysis.

The three key features of quake lakes, including storage capacity, dam height, and dam material/structural composition, were selected as the assessment indexes. Dam thickness-to-height ratio was not selected because the barrier dams were usually thick enough. The criteria for assessing the hazard level were tabulated in Table 2 and four levels were distinguished, namely, extremely high, high, moderate, and low. For each of the three indexes, a hazard level was defined and the majority of the three levels defined its overall hazard level. With this approach, the hazards posed by the 34 quake lakes were evaluated [6], respectively. One quake lake (i.e., the Tangjiashan Quake Lake) hazard level was assessed as extremely high, while five quake lakes, including the Xiaojiaqiao Quake Lake, was classified as high. The other 28 quake lakes were grouped into the moderate or low levels.

Note that Cui et al. [3], Xu et al. [5], and Yang et al. [7] used similar matrixes for the assessment. In contrast to the present approach, the former two research groups used a slightly different criterion for the index of dam height, and the latter group added two more indexes, including the quake lake catchment area and endangered lives in the downstream. It should be stated that it would give more reasonable classification of risk level to take into account the catchment area and endangered lives. In an emergency situation, however, considering that the quake lakes clustered in the seriously hit region, one could assume an equal importance of downstream flooding prone area for a rapid assessment. Nevertheless, the present research and the other three groups achieved similar conclusions, e.g., the Tangjianshan Quake Lake was the only one with an extremely high hazard level [3,5,7].

Hazard mitigation planning

While completing the assessment of quake lake hazards, mitigation planning was carried out for the quake lakes with an extremely high or high hazard levels. The Tangjiashan Quake Lake, the only reservoir with an extremely high hazard level, was given first priority to develop mitigation strategies. The following section reports the planning process for the Tangjiashan Quake Lake.

Forecast of reservoir inflows

Inflow forecasting is intended for estimating the time of lake water overtopping. It is based on the meteorological and underlying surface properties of the catchment. Subject to insufficient information about the catchment, two different approaches were adopted to improve the quality of inflow forecasting [8]. One approach was based on the empirical P (precipitation)~Pa (annual mean precipitation)~R (runoff) correlation diagram, and the other was a hydrological process simulation using the Digital Watershed Model (DWM).

In the empirical approach, runoff yield and flow concentration were calculated by using the P~Pa~R correlation diagram and the unit hydrograph method [9], respectively. A forecast made at 8:00 on May 30 by the Bureau of Hydrology, Changjiang Water Resources Commission, suggested that the lake water level would reach the elevation of 740.00 m on June 4. At 20:00 on June 3, the forecasted occurrence time was modified to 2:00 on June 7, as the following five forecasts suggested, and agreed well with the observed time of 0:00 on June 7. From what was observed, the forecast lead-time reached 72 h [8].

The DWM developed recently at Tsinghua University [10,11] is capable of hydrological process simulation for quake lakes. This modeling system uses parallel computation techniques to save computation time, which is beneficial in emergency situations. Model calibration was carried out with historical data combined with updating real-time data. Based on the official weather forecast, three different rainfall scenarios were designed and the lake inflow discharge and water level were then simulated and reported. A forecast performed at 22:00 on June 4 estimated that the lake water level would reach 739.00 m during the time from 18:00 to 23:00 on June 5, depending on the precipitation condition, which was slightly later than the observed time of 16:00 [12]. Figure 4 presents the forecasted water level performed at 24:00 on June 6. The forecasted and observed data matched well.

Note that both of the two approaches depend on weather forecast significantly and the forecast lead-time was limited to several days. Accordingly, mitigation planning for a longer period is difficult and scenario-based planning should be used. In addition, both of these methods assumed that the underlying surface conditions of the lake catchment did not change considerably during the earthquake, so that the correlation diagram and the historical flood data could be used for model calibration. Otherwise, hydrological process simulation with the real-time monitored data as the sole data resource would be necessary to calibrate the model.

Outburst flow analysis

Dam failure analysis depends on not only the lake inflow and rising water level but also the dam morphological and geotechnical information. In emergency situations, it was extremely hard to make a complete survey of dam geotechnical properties. Accordingly, stability analysis for estimating the dam failure mode became unavailable or had significant uncertainties. For simplification and rapidness, a series of dam failure scenarios were designed. The Tangjiashan dam was assumed to have sudden partial collapse or sudden full collapse. The scenarios of 10%, 20%, 33%, 50%, and 100% failure of dam body covered the cases from partial to full collapse.

Two different approaches were adopted to calculate the outburst flood. One approach was based on the broad-crested weir equation, in which the barrier dam was regarded as a broad-crested weir. The other one was a one-dimensional hydrodynamic model created by solving the Saint-Venant equations. With the broad-crested weir equation, the peak flow discharge Q_max at the dam site was calculated as

(1)Q_{\max }=\delta mb\sqrt{2\mathrm{g}}{H}_0^{3\mathrm{/}2}

where b is the dam width, H₀ is the effective water head, δ is the coefficient of lateral contraction, and m is the coefficient of discharge. The maximum flood discharge at the entrance was estimated as 46000 m³/s for the case of 33% dam failure [13].

With the hydrodynamic model approach, a one-dimensional numerical model based on an improved MacCormack scheme was adopted to calculate dam-failure flood. The geomorphology data were drawn from 1∶50000 DEM supplied by the State Bureau of Surveying and Mapping, and the Manning roughness coefficient was provided by local engineers [14]. The calculated flood discharge was 10000, 29000, 61000, 105000, and 277000 m³/s at the dam site under the scenarios of 10%, 20%, 33%, 50%, and 100% dam failure, respectively. In contrast to the broad-crested weir equation, the hydrodynamic model produced distinctly larger peak flow discharge for the scenario of 33% dam failure. This would be primarily ascribed to the embedded assumptions in Eq. (1) that assumes a steady horizontal flow over the weir. The uncertainties in model parameters specified through expert experiences would also have some contributions.

Flood routing and inundation analysis

Flood routing and inundation analysis estimate potential flooding area and help save lives by suggesting whom to evacuate first and when and where to evacuate. Similar to the analysis of outburst flow, two approaches (i.e., an approximate algebraic equation and a hydrodynamic model) were adopted.

An approximate solution of the peak discharge propagating in a natural channel can be obtained from the diffusive wave equation that ignores the local inertia and convective terms in the Saint-Venant equations [15]. Flood routing in the downstream river channel was estimated through this approximate solution as follows [13]:

(2)Q={\displaystyle \frac{W}{{\displaystyle \frac{W}{Q_{\max }}}+{\displaystyle \frac{L}{K{V}_{\max }}}}}

where L is the distance from the dam site, Q is the maximum flood discharge at distance L downstream the dam site, W is the total storage capacity of the quake lake, V_max is the historical maximum flow velocity at distance L, and K is an empirical coefficient ranging from 1.1 to 1.5 for mountain areas, 1.0 for hilly areas, and from 0.8 to 0.9 for plain areas. Based on Eq. (2), the calculated peak discharge at the Beichuan hydrological gauging station (6 km downstream the dam site), the Qinglian Town (45 km downstream), and the cross-section in Fujiang River (52€km downstream) were 40000, 24000, and 22000 m³/s, respectively, given that the maximum flood discharge at the entrance Q_max = 46000 m³/s [13].

The one-dimensional hydrodynamic model for the preceding outburst flow analysis was also used for flood routing in the Tongkou River downstream [14]. The cross-section geometry of the river was simplified as a shape of a power function. In the calculation, the time step was 5 s and the space step was 500 m. The Manning roughness coefficient from the dam site to the Mianyang City along the river channel varied linearly from 0.07 to 0.03, as suggested by local engineers. Figs. 5–7 present the calculated maximum flood discharge, depth, and arrival time under each dam failure scenario, respectively. The calculated peak discharge under the scenario of 33% dam failure at the Beichuan hydrological station, the Qinglian Town, and the cross-section in Fujiang River were 41000, 19000, and 18000 m³/s, respectively, given that the maximum flood discharge at the dam site was 61000 m³/s. The calculated peak discharge at the Beichuan, Qinglian, and Fujing River was close to the one given by Eq. (2), respectively, although the flood discharge at the dam site was distinctly larger. Meanwhile, the maximum flood depth provided by Fig. 6 was used to estimate the maximum inundation zone through marking on the topographical map. An evacuation map could then be designed to provide substantial information of the evacuation time, flood shelter, and evacuation path, along with the arrival time provided by Fig. 7.

Mitigation measures

Mitigation measures included both engineering and non-engineering techniques. As the best approach to prevent potential catastrophes from dam failure, a drainage channel at the dam crest was usually excavated to reduce the water volume and level when dam failure began [16]. The non-engineering measures included giving early warning to local inhabitants and establishing evacuation procedure for downstream inhabitants and rescue personnel on the basis of the scenario-based analysis of the flooding area and the monitoring of river water regimen.

A drainage channel on the crest and downstream face of the barrier dam was designed to produce a man-controlled flood. In the optimal design, the location and dimensions of the channel were related to the topographical and geological properties of the barrier dam, the hydrological properties of the lake, and the field conditions for construction [7]. The designed channel had a gradual slope of 0.6% in its upstream part and steep slopes of 24% and 16% in its middle and downstream parts, respectively. It had a trapezoidal cross-section with a slope of 1∶1.5 on both sides. The dimension of the channel cross-section depended on the elevation of the channel bed at the entrance that was primarily affected by weather and lake conditions. One of the three designed elevations was 742.00 m and the corresponding channel bed dimension was 326 m by 13 m on the dam crest.

Along with the engineering measure of excavating a drainage channel, the non-engineering measures such as warning and evacuation were implemented. Observatory sites were arranged around the dam site and along the downstream river to provide a real-time 24-hour observation and alert system. Various alert methods, including flare launch signals, mobile phone message, telephone, TV, and broadcast radio, were used to inform inhabitants and rescue personnel who should evacuate. Meanwhile, the evacuation of inhabitants and rescue personnel was programmed in detail for the scenarios of 33%, 50%, and 100% dam failure. In practice, the people within the inundation area of the 33% dam failure scenario were evacuated in advance.

Implementation of mitigation plans

Channel excavation and drainage

The disposal of the Tangjiashan Quake Lake started on May 21, 2008, on which the lake water stage was observed for the first time at the dam site. The excavation of the designed drainage channel started on May 26 and finished on June 1. The final channel bed was 8 m wide with an elevation of 740 m at the entrance and 739 m at the exit. The elevation at the entrance was 2 m lower than the designed elevation because the field capacity of construction was much greater than the planned amount. Accordingly, the crest of the barrier dam was finally lowered 10 m through excavation.

At 7:08 on June 7, the excavated channel began to drain at the lake water level of 740.37 m (http://news.xinhuanet.com/newscenter/2008-06/08/content_8325971.htm). The water level reached 742.18 m at 8:00 on June 10, kept for about 1 hour, and then rapidly decreased from 742.17 m at 9:00 to 735.81 m at 12:30 when the peak discharge of about 6500 m³/s occurred. The resulting peak discharge, 6540 m³/s, propagated in the downstream river at the Beichuan station at 13:00, 6210 m³/s at the Tongkou station at 14:24, and 6160 m³/s at the Qinglian Town at 14:28, respectively, which were considerably lower than the flood control capacity of the downstream Mianyang City, suggesting that the engineering measure had effectively removed the risk of dam failure flooding. During the drainage process, the channel was deepening and widening progressively due to sediment erosion and bank failure until the peak discharge appeared. Finally, more than 100 million m³ of water drained and the channel bed elevation was lowered about 20 m again.

Real-time forecast of outburst flood

When the excavated channel began to drain, a real-time forecast of outburst flooding was carried out to estimate the peak flood discharge and its occurrence time, facilitating decision-making on the selection and implementation of specific contingency plans for downstream flooding area. In reality, the dam failure due to lake water overtopping was a progressive process. The real-time monitored hydrological data in the early stage of overtopping flow enabled calibration of a physically based numerical model that has few parameters. Accordingly, this model could simulate the observed flow hydrograph and predict the impending peak discharge and its occurrence time. With the updating data, a real-time forecast could be achieved, i.e., the calibration-prediction operation was performed in a rolling way until a convergent solution was reached.

The real-time forecast of the outburst flood started at 16:00 on June 9 [17]. The time step was 0.5 s and the space step was 50 m in the numerical calculation. Figure 8 presents the forecasted flow hydrograph during the dam breaching process. The forecasted peak discharge and occurrence time were in reasonable agreement with the observed data. Moreover, such forecast results were reported 2 hours earlier before the peak discharge occurred, which gave decision-makers 2 hours to response to the flooding.

Discussion

Estimation of flood characteristics is the fundamental task in the mitigation of quake lake hazards. The results initially depend on the failure mode of the barrier dam but may vary in the different stages of mitigation.

The failure mode of the barrier dam with the excavated drainage channel was neither sudden partial failure nor sudden total failure but a progressive one for the Tangjiashan Quake Lake. Such a failure feature was similar to that of earthen dams, being distinctly different from that of concrete dams. Because of the progression mode as well as the lowered elevation of the dam crest, the peak discharge was significantly suppressed in contrast to the estimated one in scenario analysis of sudden partial failure. This was the main objective: to produce a man-controlled flood to prevent a catastrophic one. However, the effect of lowering the dam crest on the breach progression rate was not well understood.

The mitigation process involved two stages, i.e., planning and implementation, in terms of the flood characteristics considered. The peak discharge at the dam site, those at the downstream observation sections and the corresponding arrival time, had been estimated in both stages. Besides the overestimated peak discharge resulting from the assumed mode of sudden failure, the arrival time estimated in the planning stage was also significantly lower than the observed. From Fig. 7, the peak discharge of the 33% failure would take about 31 min to arrive at the Tongkou station and 61 min to arrive at the Qinglian Town, much lower than the corresponding 116 and 138 min in the real case, respectively. The direct reason was ascribed to the overestimated flood discharge, and the other reasons might include the over-simplification of downstream flood pathways that was drawn from 1∶50000 DEM and the underestimated Manning friction that was used for flood routing. Singh [18] stated that the values of Manning friction are expected to be much higher, possible larger than 0.10, due to turbulence, presence of debris, and other obstructions in actual failures. However, the information of these factors related to both the dam and lake was almost unavailable in the planning and even implementation stages.

The real-time forecast approach used in the implementation stage produced a reasonable flow hydrograph for the Tangjiashan Quake Lake. It seemed promising because the updating hydrologic information from the monitoring network was used to calibrate the Manning friction and other parameters. However, the approach cannot forecast in the planning stage and only applies to the progressive failure of barrier dams. For effective and efficient mitigation of quake lake hazards in emergency situations, the integration of various approaches and data resources would be necessary.

Conclusions

Quake lakes are created by landslides, avalanches, and debris flows that block valleys or rivers. With successive inflows, quake lakes increase flooding upstream and pose high risk of catastrophic floods downstream. In the Wenchuan Earthquake that occurred in China on May 12, 2008, 34 important quake lakes were triggered, and more than half of them had a moderate or high potential of outburst flooding. It became urgent but extremely difficult to perform hazards analysis and mitigation in the seriously damaged environment. An integrated methodology was adopted for emergency mitigation of these quake lakes. Rapid hazards analysis, scenario-based mitigation planning, and real-time forecast of outburst flood were implemented, with the support of on-site survey and remote sensing inspection, automatic monitoring networks, and integrated hydrologic and hydrodynamic modeling approaches.

The shortage of hydrologic and geological data and the nature of an emergency situation raise substantial challenges in the mitigation of quake lake hazards. For example, the uncertainties inherent with each approach and available data resources and the consistency of results from different models, both of which were not explored in this paper, had complicated the hazard assessment and mitigation processes of the Tangjiashan Quake Lake. It would be desired in the future to improve the integration of the monitoring network and the simulation models for better use of available data and the integration of hydrologic and hydrodynamic models to meet various data resources and requirements.

References

[1] Costa J E, Schuster R L. Formation and failure of natural dams. Geological Society of America Bulletin, 1988, 100(7): 1054–1068

[2] Meyer W, Sabol M, Schuster R. Landslide dammed lakes at Mount St. Helens, Washington. In: Schuster R L, ed. Landslide Dams: Process, Risk, and Mitigation, American Society of Civil Engineers Geotechnical Spec Pub l, 1986, 3: 21–41

[3] Cui P, Zhu Y Y, Han Y S, Chen X Q, Zhuang J Q. The 12 May Wenchuan earthquake-induced landslide lakes: distribution and preliminary risk evaluation. Landslides, 2009, 6(3): 209–223

[4] Chai H J, Liu H C, Zhang Z Y. Landslide dams induced by Diexi Earthquake in 1933 and its environmental effect. Journal of Geological Hazards and Environment Preservation, 1995, 6(1): 7–17

[5] Xu Q, Fan X M, Huang R Q, Van Westen C. Landslide dams triggered by the Wenchuan Earthquake, Sichuan Province, south west China. Bulletin of Engineering Geology and the Environment, 2009, 68(3): 373–386

[6] Xu W J. Landslides-Induced River Blockage and Its Hazards.Beijing: IWHR, 2010 (in Chinese)

[7] Yang X G, Yang Z H, Cao S Y, Gao X J, Li S J. Key techniques for the emergency disposal of Quake lakes. Natural Hazards, 2010, 52(1): 43–56

[8] Zhang J X. Hydrologic Analysis and Emergency Application of Barrier Lake Breaking.Beijing: Tsinghua University, 2009 (in Chinese)

[9] Sherman L K. Streamflow from rainfall by the unit graph method. Engineering News Record, 1932, 108: 501–505

[10] Wang G Q, Liu J H. Digital Watershed Model.Beijing: Science Press, 2006 (in Chinese)

[11] Wang G Q, Wu B S, Li T J. Digital yellow river model. Journal of Hydro-environment Research, 2007, 1(1): 1–11

[12] Gao J, Wang G Q, Fu X D, Zhang J X, Li T J, Wang H, Chen Z H. Forecast of inflows and water levels of Tangjiashan dammed lake based on digital watershed model. Journal of Hohai University, 2009, 37(5): 568–572 (Natural Sciences) (in Chinese)

[13] Liu N, Zhang J, Lin W, Cheng W, and Chen Z. Draining Tangjishan Barrier Lake after Wenchuan Earthquake and the flood propagation after the dam break. Science in China Series E-Technological Sciences, 2009, 52 (4): 801–809

[14] Liu F. The Simulation of One-dimensional Dam Breach and Emergency Calculation of the Sichuan Quake Lake.Beijing: Tsinghua University, 2008 (in Chinese)

[15] Xie R. Hydraulics of Dam Break Flow.Jinan: Shangdong Science & Technology Press, 1993 (in Chinese)

[16] Bonnard C. Technical and human aspects of historic rockslide dammed lakes and landslide dam breaches. In: Italian Journal of Engineering Geology and Environment, Special Issue on Security of Natural and Artificial Rockslide Dams, 2006

[17] Wang G Q, Liu F, Fu X D, Li T J. Simulation of dam breach development for emergency treatment of the Tangjiashan Quake Lake in China. Science in China Series E-Technological Sciences, 2008, 51(Supp II): 82–94

[18] Singh V P. Dam Breach Modeling Technology.Dordrecht: Kluwer Academic Publishers, 1996

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg