Natural and Artificial Rockslide Dams

The formation and behaviour of natural and artificial rockslide dams are reviewed to update the well-known work of Costa and Schuster [1]. Rockslide dams block surface drainage to form upstream lakes. They may occur naturally due to landslides or as a result of engineered rock slope failure. As evidenced by the 2010 Hunza event (Pakistan), the stability of rockslide dams is a major consideration in landslide risk assessment in mountain terrain, particularly with respect to the possibility of a destructive downstream flood resulting from a breach of the dam. The damming of a river by a rockslide may require immediate engineering response to mitigate the hazard. However, failure by breaching is less frequent than long-term stability. These issues are examined with reference to nine case histories of rockslide dams and rockslide-dammed lakes; Gohna (1894), Rio Barrancas (1914), Condor-Sencca (1945), Mayunmarca (1974), La Josefina (1993), Tsao-Ling (1999), Yigong (2000), Tangjiashan (2008), and the Hunza (2010). The case histories also illustrate the utility of digital terrain data (especially the SRTM-3 data set obtained in February 2000) and remote sensing imagery to obtain accurate estimates of the impoundment volumes and other geomorphic data on rockslide-dammed lakes. Methods of estimating peak breach discharge and downstream flood effects exist but are still largely empirical in nature. Measures to mitigate hazard associated with rockslide-dammed lakes include the construction of a spillway over the rockslide debris, a by-pass tunnels through the abutments of the debris dam, the implementation of dam and lake-level monitoring and failure warning systems to mitigate downstream damage. A review of some well-documented examples show that these measures have had been applied with mixed success in the past. Natural rockslide dams are commonly used for foundations for conventional constructed dams. Artificial rockslide dams are created by rock slope failure induced by large-scale explosion (blast-fill dams). The largest blast-fill dam yet constructed is the Medeo Dam, a debris flow retention structure near Alma-Ata, Kazakhstan. Rockslide dams and their geomorphic effects may create an important legacy in the landscape through massive accumulations of lake sediments, impact on river channels, and effects on river long-profiles.

Landslide dams cause two types of floods: (1) upstream flooding as the impoundment fills, and (2) downstream flooding resulting from catastrophic failure of the dam. A landslide dam may last for a few hours or for thousands of years, depending on: (1) rate of inflow to the lake (2) size and shape of the dam (3) physical character of the materials that comprise the dam, and (4) rate of seepage through the dam. This paper discusses (1) modes of flooding resulting from landslide dams (2) characteristics of failure of landslide dams (3) mitigative (i.e., remedial) measures used to reduce flood hazards from landslide dams, and (4) case histories in which mitigative measures have been used, both successfully and unsuccessfully. Mitigative engineering measures can reduce the hazard associated with landslide dams by preventing the failure of most landslide dams, or at least by reducing the severity of possible flooding.

The historical presentation of some 12 major landslide dam cases in the Alps which occurred between the first century AD and the present time, and for which significant information is available, allows the formulation of specific conditions which are generally met with respect to the development of the phenomena, their direct and indirect consequences on the population and the possible prevention and mitigation actions that can be carried out. Three major recent cases in Switzerland and South America, in which human preventive action was essential to save many lives and property, are then briefly commented, especially with respect to the need for timely and well-coordinated action by the Civil Defence, the Army and the local authorities, in order to avoid more dramatic consequences due to the formation and subsequent breaching of the landslide dam lake. This analysis points out the necessity of assessing the potential landslide scenarios and the related risks properly so as to limit the possible consequences of such dramatic events in the future.

Large (>10⁶ m³) rockslides and rock avalanches are common causes of river blockage in the tectonically active Southern Alps of New Zealand. Geomorphic data compiled from 43 examples show that evidence of rockslide dams occurs in comparable number on both sides of the main divide despite differing lithology, rates of uplift, erosion, and sediment yield. These controls however constrain the preservation of few rockslide-dammed lakes in the Southern Alps. In general, well-defined cross-valley dams are between 10 and 400 m high, 1.2 km wide on average, and formed by catastrophic failures. In contrast, many deep-seated low-displacement failures >10⁷ m³ formed ephemeral dams and occlusions, i.e. channel diversions around their toe zones. Rockslide and rock avalanche dams are often major sources of sediment to alpine rivers, though concentrated on a few percent of catchment area. Averaged long-term (>1,000 year) discharge rates from eroding dams can attain 8 × 10⁵ m³/year. Despite the low population density in the Southern Alps, some hazards from rockslide and rock avalanche dams, i.e., outburst floods and prolonged river-bed aggradation from eroding dams remain for downstream communities. The currently high probability of a large earthquake makes synchronous blockage at several locations in the region likely.

Landslide dams are frequent phenomena in the Argentine Andes. We studied 20 landslide dams in NW Argentina and 41 landslide dams in northern Patagonia. These examples show that most of the landslide dams in both regions have longevity of several hundred to several tens of thousands of years. In those cases where the mode of dam erosion/breach was reconstructable it was either related to climatic variability influencing the inflow of water into the landslide-dammed lake or by landsliding into the landslide dammed lake causing a tsunami wave which overtopped the dam crown and caused its erosion. However such tsunami waves not always lead to dam failure. There is one case where flood deposits downriver a dam exist and the landslide dammed lake contains a voluminous landslide deposit, however the dam did not breach. Hence the flood deposits are related to the tsunami wave but not to a breach. In addition, our examples indicate the necessity of expanding the well established dam classification system used globally in the past 18 years. Here we define 4 further dam types which are related to (a) the diversion of the river away from the valley over bedrock (b) the diversion of the river into the neighbouring catchments (c) the deposition of the landslide in a drainage divide, and (d) the formation of multiple dams by the breach of a landslide dam itself. Furthermore, ponds on top of landslide deposits are frequent and depending on their size a catastrophic release may cause damage.

Out of more than 320 late Quaternary rock avalanches identified in the Upper Indus Basin, some 161 formed cross-valley barriers impounding one or more rivers have been investigated. At least 228 lakes were associated with them. Only a few small lakes exist today, but many former lakes exceeded 20 km long, a few, 90 km. More than half the dams had an initial, effective barrier height of over 25 m, 33 exceeded 100 m, and two 1,000 m. Today, impoundments exist in every conceivable stage of infilling and degradation, and 117 barriers are still not fully breached. Given the number and diversity of events, and different states in which they are found, a comparative analysis is undertaken. A revised classification of landslide dams is proposed to take account of the large fraction of the events that involved complex emplacement of the rock avalanche. In addition, 26 cases are identified as hybrid dams, the barriers composed of two or more materials due to the interaction of the rock avalanche with substrates, or large uptake of moisture. Most of the newly discovered dams impounded lakes for many years, some for centuries, and the evidence suggests relatively gradual or phased breaching. Some aspects of breach histories can be reconstructed from field evidence, notably former, abandoned spillways. Dam stability is discussed in terms of the combined effects of barrier size, shape, composition, sudden emplacement and consolidation. The hybrid dams raise some special problems of the response of different materials and relations between them. The incidence of large rock slope failures relates to the region’s late Quaternary history, especially slopes over-steepened by ice action and debutressed with deglaciation. Landslide barriers have been decisive in post-glacial landform development along the Upper Indus streams, and continue to exercise control over sediment delivery and routing of floods from other causes. This newly emerging picture of landslide damming suggests an urgent need for contemporary risk assessment. In areas of potential inundation or outburst floods, there are more people, wealth and infrastructure than ever before. Most rock avalanche deposits have settlements on or near them. Much of the critical arable land is on valley fill sediments associated with landslide dams. Some implied risk profiles are reviewed, and the evidence needed to confirm or refute them.

The remains of rockslide dams are widespread in the river valleys of the northwest Himalayas (Pakistan and India) and the adjacent Pamir Mountains of Afghanistan and Tajikistan, Central Asia. The region contains in excess of two hundred known rockslide deposits of, as yet, unknown age that have interrupted surface drainage and previously dammed major rivers in the region in prehistoric time. In addition, the region contains (1) the highest rockslide dam (the 1911 Usoi rockslide, Tajikistan) in the world that dams the largest rockslide-dammed lake (Lake Sarez) on Earth, (2) the largest documented outburst flood associated with a historical rockslide dam outburst (the 1841 Indus Flood, Pakistan), and (3) the world’s most recent (2010) rockslide-dammed lake emergency, the Attabad rockslide dam on the Hunza River, in the Upper Indus basin of northern Pakistan. We show that some prehistoric rockslide dams in the northwest Himalayas impounded lakes with volumes in excess of 20 Gm³, significantly larger than present-day Lake Sarez. We use SRTM-3 digital terrain data and satellite imagery to analyse four major historical rockslide damming events (1) our analysis of the 1841 Indus rockslide-damming event indicates that the volume of the impoundment and subsequent outburst was 6.5 Gm³, the largest outburst from a rockslide-dammed lake in historical time, (2) the 1858 Hunza Valley rockslide dam impounded about 805 Mm³ before catastrophic outburst in August 1858, (3) the development of the 2010 Hunza rockslide-dammed lake is described in detail for the first time. It reached a maximum volume of 430 Mm³ before stable overflow of the rockslide debris began on May 29, 2010. This remains the situation as of July 25, 2010 (200 days after impoundment), (4) the filling of Lake Sarez was conditioned by excessive outflow seepage through the debris as the lake filled to the extent that a freeboard of ca. 50 m is still naturally maintained without engineering intervention. The emplacement of rockslide dams and the behaviour of their impounded lakes are critical hazards to communities and the development of infrastructure, including hydroelectric facilities, in this region of Central Asia.

This paper attempts to work out the stability conditions of 20 landslide dams in the Indian and Nepal Himalayas as well as two in China. All of them formed lakes upstream by damming the main river. The categories of study areas that they created differ in their shape, their volume and composition of the deposited landslide material, as well as in the size and life span of their former dammed lakes. This representative group of case studies affords the opportunity to obtain new results on this frequent phenomenon with the conclusion that besides the so far well known parameters for the stability of lake-damming natural rock blockages, such as landslide mor-phology, volume of debris, composition of the material and lithology, additional factors influencing the mechanical behavior of the landslide material are most important: movement-type, shattering grade, secondary compaction, involved moraine material, size of the catchment area, climatic conditions and rates of sedimentation. Finally a simple diagram, which correlates the grain, boulder and block size of landslide material and the stability of a dam should help in determining whether a lake-damming landslide is stable or not.

The collapse of a volcanic edifice can enhance several secondary effects, including the formation of one or more natural dams. Their duration depends on the volume of the obstructing mass and on its textural characteristics. A block facies of a debris avalanche produces “durable” dams, whereas a mixed facies is easily eroded after overflowing. Analysis of the sedimentological characteristics of different volcaniclastic deposits that formed natural dams indicates that low sorting and coarse grain-size (but significant clay content) can promote a longer duration of a natural dam. It appear that a –1 phi mean grain-size represents the limit between volcanic natural dams, which last just few days (Md >–1 phi) from those lasting more that 1 year or are still existing (Md <–1 phi). By considering the relation between M/G (M: matrix, G: gravel) and S+C/C (S: silt; C: clay) it seems that for a M/G lower than 1, there is an inverse relation between mud content and dam duration. For dams formed exclusively of fine material (phi less than –1, with comparable fine fraction), the stability decreases as matrix content increases.

The M8.0 Wenchuan earthquake of 12 May 2008 triggered over 20,000 landslides in the earthquake affected area of Longmenshan Mountain. It was a common phenomenon that large scale landslides blocked river channels. In total, 257 landslide-dams were identified by field investigation and remote sensing inspection. The dams were distributed in a belt along the rupture zone and in clusters along the river valleys. Emergency risk assessment was carried out due to the rise of water levels behind the dams and catastrophic outburst potential of the landslide-dammed lakes. Under the emergency circumstances, the dam height, dam composition and storage capacity of the landslide-dammed lake were used as indices of the dam’s risk of collapse. Four risk levels were assigned. By analyzing 21 landslide dams in detail, the results of risk assessment were obtained; 1 dam had an extremely high danger risk (Tangjiashan Lake), 7 dams had a high danger, 5 dams medium danger and 8 dams were of low danger. The case of the Tangjiashan rockslide dam that formed the largest landslide-dammed lake, is discussed in detailed. The scenarios analysed were for 20, 25, 33, and 50% of the dam failing, yielding, respectively, flood heights of 4.6, 5.1, 5.7, and 6.2 m, inundation areas of 3.35, 3.84, 4.22, and 4.65 km², and discharges of 6,106, 7,397, 9,062, and 11,260 m³/s, at the Fujiang Bridge in Mianyang, the second largest city in Sichuan Province. The outcome of this analysis was used by emergency managers to plan downstream evacuations. Sluiceways were designed to lower water level of the dammed lakes and to reduce the risk of the outburst floods. Emergency mitigation operations for risk-elimination proved to be effective and successful. On the other hand, debris flows will be the dominating disaster process in future and will likely deposit masses of debris in river channels forming new dams in the coming 5–10 years. It is thus of prime importance to take measures and prevent debris flows from damming main rivers and posing new threats to downstream locations.

In order to explain the long life span of the Scanno rockslide-avalanche dammed lake (Central Italy), whose age of impoundment is older than 2,300 years, and to get indications about its present and future stability conditions, it was assumed of fundamental importance to build a representative geological model aimed at defining (i) the geometry of the boundary surface between the rockslide debris and the bedrock, (ii) the characteristics of the debris with special reference to hydraulic behaviour, (iii) the flownet within the landslide deposit taking into account the complex geological and hydraulic boundary conditions imposed by the palaeovalley morphology and by the numerous springs downstream of the dam which are fed by different aquifers. The reconstruction of the natural dam geometry (volume, heigth and width) was possible by means of integrated geological (site surveys and boreholes) and geophysical (electric tomography and seismic refraction) investigations: the so obtained data were processed in order to assess the stability conditions according to existing geomorphological indexes. As a result, good stability conditions of the analysed natural dam are highlighted on the basis of its geometric features. A campaign of pumping tests, hydrogeological measurements, chemical and isotope analyses was carried out with the aim of defining the flownet within the rockslide-avalanche deposit and its hydraulic properties. The so depicted hydrogeologic characteristics indicate that dam failures by piping processes are very unlikely, thus supporting the dam stability conditions.

Landslide dams can be originated by very different phenomena, each characterized by specific types of movement, which control the final characteristics of the material. We present the case study of the Val Pola landslide dam (Central Italian Alps, Lombardy), formed in 1987 by al large rock avalanche. We describe the main landslide features as well as those of its accumulation. Data from laboratory and in situ tests performed on the landslide deposits are reported and discussed. The collected data are used to describe the internal structure of the accumulation, the behaviour of the landslide dam during the lake infilling and the overtopping phase. These data are fundamental to interpret the mechanisms of put in place of the accumulation and the theories proposed to explain long runout of rock avalanches. We compare the grain size distributions of the deposits and the physical-mechanical properties from different landslide sectors. We stress the relevance of a detailed characterization of the internal structure of landslide dams to perform correct seepage modelling and to perform reliable predictions of landslide dam stability and breaching (geometry, water and sediment discharge, duration). Point-like characterization, often controlled by local conditions and presence of small outcrops, can be misleading both in terms of the evaluation of the stability of the dam and in the understanding of the processes. We used several published relationships and approaches for the assessment of the peak discharge, breaching time and geometry. We confirm the large variability of the results and the difficulties in using collected data for reliable predictions.

Chinese historic records indicated that on June 1, 1786, a strong M = 7.75 earthquake occurred in the Kangding-Luding area of Sichuan, southwestern China, resulting in a large landslide dam that blocked the Dadu River. Ten days later, the sudden breaching of the dam resulted in catastrophic downstream flooding, and reportedly up to 100,000 deaths. Thus, this may well be the most disastrous event ever caused by landslide dam failures in the world. Although a considerable amount of seismological studies has been carried out to determine the location, magnitude and intensity of the 1,786 earthquake, relatively little is known about the landslide dam. In this paper, the dam was reconstructed using historic documents and geomorphic evidence. It was found that the landslide dam was about 70 m high and it created a lake with a water volume of about 50 × 10⁶ m³ and an area of about 1.7 km². The landslide dam breached suddenly due to a major aftershock on June 10, 1786. The peak discharge at the dam breach was estimated using regression equations and a predictive model.

Abstract On March 29, 1993, a large landslide (20 million m³) took place northeast of Cuenca city in Ecuador, damming the Paute river. During the next 33 days, an artificial lake upstream of the blockage (200 million m³) was formed, flooding fertile land and destroying houses, roads, railways and a regional thermoelectric plant. On April 23, the water began to overflow out of the dam through a channel which was excavated in order to reduce the water volume accumulated in the lake and avoid catastrophic damages downstream. On May 1, the breaching of the dam produced a peak discharge of 10,000 m³/s, destroying roads, house, bridges, etc., downstream. This work describes the rupture of the dam and the routing of the flood wave. Models used to predict the rupture time and peak discharge of the flood are also compared. In addition, some aspects of the crisis management are presented.

The Flims rockslide blocked at about 8,200 year BP the Vorderrhein river and so gave rise to the Paleo-Lake Ilanz. Shortly after the filling of the lake a severe flood event happened. The lake emptied not totally, but remained at a lower level for a long time. After important and perhaps even complete filling of the remaining basin with sediments the incision in the dam restarted and progressed in several smaller steps. Based on the detailed field works, the following sequence of the events can be envisioned: The Tamins rockslide occurred first, damming Lake Bonaduz; the Flims slide come down soon after and probably entered in this lake; at the same time the alluvium was mobilized and resedimented as Bonaduz gravel; Lake Ilanz was dammed to a level between 820 and 936 m for very short time; the dam broke partly and a flood happened; the remaining lake survived for a long time and was filled up; the Rhein river eroded the floodplain down to the actual level.

The Usoi blockage, 2.2 km³ in volume formed a World-highest natural dam and the Lake Sarez containing about 16 km³ of water can pose a threat to the downstream Bartang – Pianj – Amu-Daria River valleys. Different view-points on the dams’ stability exist. Modern data on the dam and the potentially unstable lakeside are presented that are used for the Lake Sarez outburst flood risk assessment. The conclusion has been drown that at present the Usoi Dam is stable and can not be completely destroyed either by the internal erosion or by overtopping owing to its huge dimensions, composition and structure. Regarding the existence of the Right Bank Landslide 0.9 km³ in volume, which simultaneous collapse has been considered as the most hazardous phenomena capable for dams’ breach, no sufficient evidences of such scenario have been found out. This slope is undergoing typical post-glacial evolution and only relatively small failures that can not cause the Usoi Dam destruction are expected. Even Lake level raise caused by dams’ body compaction due to strong earthquakes and decrease of seepage or by increasing inflow due to climate changes should not lead to the immediate catastrophic consequences, since composition of the dams’ surface covered by blocky carapace prevents its rapid destruction.

A fundamental quantity affecting landslide impact is its size: we ask how size might be known in advance of failure, and explore conditions under which there might be useful answers. The regional answer comes from probabilistic landslide hazard analysis. We approach the local problem by discussing forms that answers might have taken in some historical New Zealand landslides. Edifice shape seems more important than specific, weakest defects in determining release surfaces, and so the probability-density distribution of potential failure sizes at a site is estimable from topography, general knowledge of rock-mass characteristics, and the probability-density distributions of potential triggers. To understand and predict how far a landslide of known size will travel, and what its internal structure will be, we discuss dynamic grain fragmentation and its role in rock-avalanche and block-slide motion. We then examine how fragmentation affects the stability of rock-avalanche and block-slide dams.

This article examines the possibility of predicting the geometry of landslide dams on the basis of the characteristics of potential landslides. Empirical techniques exist, by which the distance of travel and mean thickness of the landslide debris can be estimated. These methods are suitable for preliminary studies, but suffer from considerable variability. More sophisticated techniques using two or three-dimensional dynamic analyses of landslide motion are then described using several examples involving actual landslide cases. Theory exists for predicting a wide range of types of landslide motion. However, for practical uses, all of the dynamic methods rely on calibration against case histories. A large amount of research work is still required before such methods can be routinely used to make reliable predictions. For the time being, their use requires a large amount of subjective judgement, which should be augmented by back-analysis of cases similar to the case under consideration.

The sedimentology of rock-avalanche deposits, mass movements that commonly cause natural dams is poorly understood. This paper presents the results of detailed research into the grain-size distribution and sedimentology of rock-avalanche deposits. These data show poorly sorted deposits composed of angular, highly-fragmented clasts with preserved source stratigraphy. Grain-size distributions segregate based on rock type and become finer with distance travelled from the source when topographic constraints do not dominate. The results indicate that the inverse grading noted at deposits is often a misconception, the main interior of deposits show fragmentation variation only due to lithology. A facies model is presented consisting of a surficial, coarse openwork carapace facies, a fragmented interior body facies and a basal facies of entrained substrate mixed with rock-avalanche debris. Grain-size prove fractal in nature in direct contrast to Weibull type distribution previously cited for rock-avalanche deposits and suggest confined comminution as the formational mechanism. An application of the data is presented using finite element/limit equilibrium techniques to a simplified rock-avalanche dam. The modelling shows that sedimentology is key to the failure sequence. An idealised homogenous dam took 17.75 days to fail; a more sedimentologically realistic dam failed within 48 h as the lake reaches the carapace facies and phreatic tonguing occurred.

Slope stability analysis of both natural and engineered slopes has changed significantly during recent years with a gradual transition from simple limit equilibrium analyses to the frequent application of numerical modelling. Case histories are presented illustrating the use of numerical modelling to examine the influence of groundwater and coupled hydro-mechanical processes on deep-seated slope failure mechanisms. The application of limit equilibrium and finite-element methods is discussed with respect to the continuum mechanics of soil and debris slopes. Preliminary analyses of the Usoi landslide dam and a shallow soil slide from Switzerland are presented as examples. Although such techniques may be applied to rock slopes, in most cases it is more efficient to utilize discontinuum-based techniques, such as the distinct-element method, in order to simulate the influence of groundwater on the deformation of jointed rock masses. The use of distinct-element modelling to illustrate the efficacy of remedial groundwater drainage on rock slope stability is clearly illustrated using a case study from southern Switzerland.

Volcanically active regions, like the Taupo Volcanic Zone (TVZ) in the central North Island of New Zealand, are an ideal environment for the temporary impoundment of surface water because of the basin-forming properties of explosive eruptions and volcano-tectonic faulting, and the valley-damming potential of pyroclastic and lava flows, lahar deposits, and debris avalanches. Historic break-out floods are known from two lakes, the summit Crater Lake of Mt. Ruapehu, and intra-caldera Lake Tarawera, but careful analysis of the geologic record shows abundant evidence for older floods on much larger scales in other settings. The largest flood so far recognised involved the release of c. 60 km³ of water from intracaldera Lake Taupo following a caldera-forming eruption at 26.5 ka: the recurrent nature of such hazards is shown by a similar 20 km³ break-out in the aftermath of the 1.8 ka Taupo eruption. Multiple prehistoric paleofloods are recognised or inferred at several other volcano-lake centres. Paleohydraulic analyses of these events show that in terms of volume and peak discharge they are amongst the largest known late Pleistocene floods worldwide, only being exceeded by break-outs from pro-glacial lakes impounded by continental ice-sheets, overspilling of pluvial lakes in tectonic basins, or some Icelandic jökulhlaups.

In April 2000 a large-scale rock avalanche dammed the Yigong Zangbo River, forming an extensive rockslide-dammed lake. The impoundment lasted for 62 days before a catastrophic breaching caused a massive outburst flood in the Yarlung Zangbo (Tibet) and the Dihang Rivers (India) that travelled about 500 km downstream to the floodplain of the Brahmaputra in northeastern India. In response to discrepancies in the published literature on the event, a review and re-evaluation of the characteristics of the rock avalanche and associated landslide-dammed lake has been undertaken using remote sensing data. The data included SRTM-3 digital terrain data and LANDSAT-7 imagery obtained before, during and after the impoundment. Analysis of satellite imagery and SRTM-3 terrain data indicates that the volume of the damming rockslide was ca. 105 Mm³ and the outburst volume of the rockslide-dammed lake was ca. 2 Mm³ at a maximum pool elevation of 2,264 m a.s.l (rounded to 2,265 m a.s.l.). These figures differ from those previously published but are believed to be well-constrained credible estimates of the volumes of the 2000 Yigong events. In terms of historical outburst volumes from rockslide-dammed lakes, the volume of the 2000 Yigong event is only exceeded by that of the 1841 outburst flood from the Indus River rockslide-dammed lake, northern Pakistan.

The Tsao-Ling rock avalanche was the largest landslide triggered by the 1999 Chi–Chi earthquake in central Taiwan. In addition to detailed and extensive field observations, several Digital Elevation Models (DEMs) generated from sets of aerial photos have been utilized to document and measure the coseismic and post-seismic morphological changes at Tsao-Ling. The estimated volume of the initial rock avalanche is about 125.5 and 138 Mm³ for the depleted and accumulated zones, respectively, indicating an increase in volume due to fragmentation. The average thickness was about 150–170 m, up to 195 m of the slid mass–over 140 m on the Chingshui River channel while between 30 and 90 m debris covered on the preexisting debris deposit hill. The landslide debris blocked the river channel and formed a dammed lake, with a maximum volume of 45 Mm³. In the deposit area, strong river erosion has removed 72 Mm³ of the debris since the earthquake. By August 2004, the rockslide-dammed lake had filled up completely with sediment. The filling materials originated in numerous landslides in the upstream area.

It is suggested that rockslide – rock avalanche dams are more complex than is recognized in existing landslide dam classifications. A classification system is proposed emphasizing the three-dimensional relations of the landslide deposit to valley morphology. This system combines a three-step classification taking account of: (A) the plan view distribution of the deposit in relation to valley morphology and the impounded water bodies, (B) the cross valley profile of the landslide deposit as it relates to the buried valley morphology, and (C) the profile of the rockslide debris and the underlying substrate along the thalweg of the valley. It is argued this serves to better describe the dams with respect to the hazard of outburst floods, impoundment histories and the subsequent morphological evolution of mountain stream valleys.

The construction of the largest blast-fill dams that had been built in the former USSR in 1959–1989 is described, the Medeo and the Baipaza dams exceeding 1×10⁶ m³ in volume in particular. Explosions had been designed by use of the modeling device allowing estimation of the single blasts parameters that provide the required dams’ size. The experimental Burlykia and Uch-Terek dams are described that have been constructed during the preparation for construction of the unique Kambarata-1 dam which have had to be about 200×10⁶ m³ in volume. Method of this dam construction by avalanche-like motion of debris from the exploded slope is elaborated on the basis of data on these dam constructions and of data on natural blockages.

Blast-fill and, especially, blast-rockslide technology allows construction of large dams with high safety factor and at the acceptable cost at those sites, which geological and seismological conditions complicate or even prevent construction of dams of other types. Large-scale field experiments and study of natural dams of the same size as the designed structures have been carried out to obtain reliable information about the processes that take place during large amount of blasted rock emplacement. Grain-size composition, density and filtration characteristics evolution of the experimental Burlykia dams’ body, and topographic and geological conditions of dam sites preferable for such dams construction are described. Besides, types of across- and along-valley distribution of rock avalanche debris, which, in turn, determines dams’ height and shape, are identified and factors determining either morphological type are proposed. Such critical parameters as dams’ permeability and erosion resistance in case of overtopping depends on blockages’ dual-zone internal structure with wide core composed of shattered debris and coarse carapace. High seismic stability of natural blockages exemplified by numerous case studies shows that large blast-rockslide dams could be considered as one of the most earthquake-proof type of water retaining structures.