|The Himalayan Dilemma: Reconciling Development and Conservation (UNU, 1989, 295 pages)|
|5. Mountain slope instability: natural processes or human intervention?|
Any attempt to understand the significance of rates of denudation that have been determined for the Himalaya not only requires an appreciation of the status of mountain geomorphology, as has been outlined in the preceding section, but also must be founded on at least a general overview of the region's physiographic divisions, bedrock geology, and tectonic structure. To keep this section within reasonable limits the discussion is restricted to the Himalaya sensu stricto. Reference also should be made to Chapter 2, and particularly its treatment of climate and vegetation.
The entire central part of the mountain system can be divided into nine strike-oriented physiographic units that trend approximately from westnorthwest to east-southeast (cf. Figure 2.3 page 20). The four northern units include the southern portion of the Tibetan Plateau, the Tibetan Marginal Range (6,000-7,000 m), the Greater, or High, Himalaya (with numerous summits exceeding 7,500-8,500m - maximum 8,848 m), and the Inner Himalaya (sometimes referred to as Trans-Himalaya), an intervening system of high plateaus and valleys lying between the two great mountain ranges. These four units contain the 'high mountain' belt and, except for very small areas in the more deeply cut valleys, extend above the montane forest belts and timberline at approximately 4,000 m. This is an extremely complex mountain landscape, heavily sculptured by glacial erosion during the Late Cainozoic and carrying a considerable cover of snow and ice today. Lithologies range from sedimentary and metamorphic rocks to granite intrusives: the whole has been subject to extremely complex faulting, folding, and overthrusting. The higher, north-facing slopes tend to be semi-arid to extremely arid because of high altitude and rainshadow effects. The south-facing slopes intercept the summer monsoon although, as discussed in Chapter 2, precipitation decreases with increasing altitudes, and the very large precipitation totals are generally confined to the outer ranges and plains. Until recently Cherrapungi in the plains of Assam claimed the world's record annual rainfall total (11,615 mm). However, longterm records from high-mountain stations, or any precipitation data at all, are rather scarce. Several more recently established stations are showing totals in excess of 5,000 mm/yr; nevertheless, these are all located on south-facing slopes highly exposed to the monsoon.
The southern flanks of the Greater Himalaya merge via a transition belt with the so-called Middle Mountains. The latter have been the traditional centre of high population densities for much of the Himalaya. The transition belt, the Middle Mountains, and the northern flank of the next, outer unit, the Mahabharat Lekh in Nepal, or the Lesser Himalaya in Uttar Pradesh, are collectively referred to as the pahar in Nepal (pahad in Mahat et al., 1986a). The upper northern sections of this tripartite division remain largely under upper montane forest (2,900-4,000 m) below which is the belt of intensive agriculture. Lithologies are extremely varied, including sedimentaries, metamorphics, and granites. However, there are extensive areas of phyllites and schists; these are deeply weathered and this, together with the prevailing steep slopes, renders them highly susceptible to erosion.
The Mahabharat Lekh, together with the Siwaliks (or Churia Hills), constitute the Lesser Himalaya, or outer ranges, bounded against the Ganges Plain (Terse) by the Main Frontal Thrust (a major tectonic translocation). Highest summits of the Mahabharat Lekh approach 3,000 m; the Siwaliks are much lower. Between and within them occur the famous 'dun' valleys, tectonic depressions which support some of the richest agricultural land, ranging from Dehra Dun in the west and including the Rapti Valley at Chitwan in Nepal. The lower, outer duns are locally referred to as the Inner Terai. The Mahabharat Lekh and Siwaliks are composed of the youngest, Tertiary strata and contain some of the most easily erodible lithologies (including unconsolidated sands and gravels) of the entire Himalayan Region. Where these ranges have been extensively deforested, as in parts of Uttar Pradesh and Himachal Pradesh, there has been a catastrophic production of sediments and development of badland topography.
The Siwaliks abut the Ganges Plain abruptly at an altitude ranging from about 200 m in the east to 500 m in the vicinity of Dehra Dun. The alluvial sediments of the Ganges Plain can be divided into the high Terai (Nepal), barbar, or porous place) composed of massive coarse-grained alluvial fans and torrent fans, and the low Terai, underlain by finer sediments of the Ganges flood plain proper. As mentioned above, these sediments are up to 5,000 m thick in places and represent the vast accumulations of Himalayan weathering products over the past several million years.
Central to an understanding of the efficiency of erosional processes in the Himalaya, as well as the significance of the available estimates of denudation rates, is the concept of slope-channel coupling (Brunsden and Thornes, 1979). This concept was introduced in the previous section (pp. 96-97) in the discussion of the lack of complete transfer of the products of alpine (that is, the altitudinal belt above timberline) mass movement from the mountain slopes and out through the fluvial system (Barsch and Caine, 1984). Given these qualifications, and the problems of representativeness of the study sites in both time and space, we can proceed to a review of the denudation rates that have been derived for the Himalaya region.
Table 5.1 provides selected denudation rates for the Himalaya as a whole, or for various watersheds and areas of differing size. Despite the wide range (from 0.5-20 mm/ yr), and the inherent errors discussed above, these figures are very high when compared with rates from other parts of the world. The Alps, for instance, have provided an overall rate of I mm/yr; figures above about 6-7 mm/yr are amongst the highest ever recorded or estimated. At the outset, therefore, they are suggestive of a very dynamic environment, especially when coupled with the high rate of tectonic uplift and the intensity of seismic activity.
The work of Brunsden et al. (1981), Starkel (1972a and b), and Ramsay (1985, 1986) represent almost the only systematic attempts in recent years to derive denudation rates from actual field measurements. The study by Brunsden et al. provides especially valuable insights for a section of the Lesser Himalaya in eastern Nepal, including observations from two field surveys for a road alignment between Dharan on the Terai to Dhankuta at 2,200 m. Theirs is the first comprehensive description of current hillslope and fluvial processes. They concluded that the relative relief of their study area has been increasing throughout the past million years or so because stream downcutting has exceeded lowering of the local interfluves. Valley side slopes, therefore, have progressively lengthened and their parallel retreat is being effected by stream undercutting and landsliding. Thus, according to Brunsden et a/. (1981), debris mobilized on the steep slopes flows directly into the rivers giving a very high sediment-delivery ratio, and is subsequently moved out onto the Ganges Plain by fluvial transport.
Table 5.1 Selected denudation rates for the Himalayan Region (from various sources, after Ramsay (1985:9).
|Location||Denudation (mm/yr)||rate Comments||Author|
|Ganges/Brahmaputra watershed||0.70||From present rate of influx to Bay of Bengal Fan||Curray and Moore, 1971|
|R. Hunza watershed||1.80||From sediment yield||Ferguson, 1984|
|R. Tamur watershed||5.14||From sediment yield||Seshadri, 1960¹|
|R. Tamur watershed||4.70||From sediment yield 1948-50||Ahuja and Rao, 1958¹|
|R. Tamur watershed||2.56||Williams, 1977|
|R. Arun watershed||1.90||1947-60||Pal and Bagchi, 1974¹|
|R. Arun watershed||0.51||Williams, 1977, after Das, 1968|
|R. Sun Kosi watershed||2.50||Pal and Bagchi, 1974¹|
|R. Sun Kosi watershed||1.43||Williams, 1977, after Das, 1968|
|R. (Sapta) Kosi watershed||0.98||From suspended sediment||Schumm, 1963, based on Khosla, 1953|
|R. (Sapta) Kosi watershed||1.00||Williams, 1977, after Das, 1968|
|R. Karnali watershed||1.50||UNDP, 1966|
|Darjeeling area²||0.50-5 00||Forested/deforested||Starkel, 1972a|
|Darjeeling area²||10.0-20.00||In catastrophic storms||Starkel, 1972a|
¹Brunsden e' al. (1981).
²These figures are estimates and are not based on precise measurements.
The two-period fieldwork of Brunsden and his colleagues was serendipit ously timed to bracket a summer monsoon period with high rainfall amounts and a flood peak with an estimated ten-year recurrence interval. Thus they were able to appreciate the impact of such a condition on the steep slopes of their field area. They concluded that the Lesser Himalaya of eastern Nepai have one of the highest rates of denudation in the world. They regard the present landforms as characteristic forms in equilibrium with current processes and controlling tectonic, climatic, and base-level conditions. Storage and transport processes and the linkages between them were identified. In particular they conclude that one of the consequences of the very efficient slope-to-stream channel sediment-transfer system is that delivery of weathering mantle to the valley floor is not a continuous process. Debris tends to move in waves during storm events, causing pulses of heavily silted water to move downstream. This conclusion indicates that the system of erosion in the Lesser Himalaya of eastern Nepal contrasts with Barsch and Caine's alpine (mid-latitude) situation in terms of high sediment-delivery ratio for the former and low for the latter.
However, quantification of the various elements of the Brunsden et al. (1981) system, and especially the development of even a preliminary sediment budget for a single small watershed is not yet attainable.
Ramsay (1985, 1986) undertook the most exacting and effective study yet available of a specific watershed in the Himalaya. His assessment of erosion processes in the Phewa Valley, near Pokhara, Nepal, is discussed in the following section. Within the context of his attempt to convert erosion process data, together with an extensive review of the geomorphological literature, into denudation rates, however, his conclusion is an appropriate summing-up of the current status:
The author . . . would like to emphasize that his own figures for failure age, volume, and frequency (and hence the estimate of surface lowering by landsliding [2.5 mm/ yr]), should never be used without the prefix 'based on a small sample.' They are probably as valid, or as invalid, as the estimates made by Caine and Mool (1982), and Starkel (1972a and b). (Ramsey, 1985:130)
As emphasized above, the ubiquity of landslides and other catastrophic slope failures throughout the Himalayan system has been not only widely reported but viewed with alarm because it is assumed to be the direct consequence of human land-use changes over recent decades. This section will review some of the more recent work on mass-movement processes and attempt a preliminary conclusion concerning their causes, natural (geological) or accelerated (maninduced).
Kienholz et al. (1983, 1984) produced detailed maps of land use, geomorphic processes, and mountain hazard assessment on a scale of 1:10,000 for a small section of the Middle Mountains (Kakani-Kathmandu). Caine and Mool (1981, 1982) examined the channel geometry and stream-flow estimates and the landslide activity of the same area, as part of the United Nations University Mountain Hazards Mapping Project.
Caine and Mool (1981) concluded that the two mountain streams that they studied in detail were similar in their dynamics and form to streams in other parts of the world. This implies, they believe, that hydrological concepts and models derived from long records and detailed studies in mid-latitude regions can be applied to problem resolution in tropical and sub-tropical mountains, where few direct observations are available. They also demonstrated that the direct impact of man on the two fluvial channels is limited to a constraint on channel width in the uppermost 8 km. This results from the construction of agricultural terraces, especially masonry walls, which cause the stream to adjust by cutting a deeper channel. It would appear that as the stream size increases beyond about 8 km from the source, local landscape modification by the subsistence farming community is limited by the available technology and is not adequate to influence the lower reaches.
Caine and Mool (1982) also conclude that landslides presently occupy about I percent of the land surface of the map area. They have a total volume of more than 2.2 x 106 m³ and a mean age of about 6.5 years, that is, number of years since initial failure. This suggests a current rate of lowering of the entire surface (denudation rate when data from other erosive processes are added) of about 12 mm/yr (recalculated by Ramsay (1985) as 11.3 mm/yr).
For the Phewa Valley, Ramsay (1985, 1986) calculated that landslide density was 1.6 per km² with 95 percent of the landslides being small, shallow failures. The total area affected is 0.7 km², or 0.5 percent of the area. If the area of the landslide deposits is added (as in the case of the Kakani fieldwork), this gives a figure of 3.25 km², or 2.7 percent of the catchment area. Ramsay challenges Caine and Mool's estimate of the rate of landslide expansion per annum and notes that many landslide scars in the Phewa Valley appear to be healing. This view is supported by Ramsay's observation that the age of the small, shallow failures tends to cluster at slightly less than eight years, suggesting that they may heal within a decade. Caine and Mool's estimated mean age of 6.5 years is close to Ramsay's estimate. When this is taken in conjunction with Ives's (1987) repeat photography (1978-87), it would appear that the contribution of landsliding to overall denudation, regardless of the actual cause of the landsliding, has been overestimated. Nevertheless, at present it is not possible to take into consideration the positive impacts of the farmers (cf. also Kienholz et al., 1984).
Ramsay's larger areas of failure, which he terms mass-movement catchments after Brunsden et a/. (1981), contrast markedly with the larger number of small failures. Their average age is about twenty-four years and they appear to be enlarging rapidly due to feedback mechanisms, such as the extension of bare ground and overgrazed areas that produce a much higher proportion of immediate runoff during heavy rainstorms. These are certainly influenced to some degree by human intervention in the form of mismanagement of land for agricultural purposes.
Ramsay (1985, 1986) calculates an annual denudation rate of 2.5 mm/yr from landslide activity alone. This does not include the effects of soil creep, sheetwash, rifling, gullying, and solution. However, neither does it take into account the sediment-delivery ratio and, as Caine and Mool (1982) and Kienholz et al. (1984) have noted, a large proportion of the material transported by mass movements remains within the small watersheds and is not transferred to the larger rivers. Thus it appears that there is either a difference in interpretation by Caine and Mool (1982) on the one hand, and Brunsden et al. (1981) on the other, or else a difference in actual field conditions between their two study areas.
Several other studies have been undertaken with a view to determining the regional importance of landsliding, and most of these have been reviewed by Ramsay (1985, 1986) and Carson (1985).
Laban (1979) carried out an airborne reconnaissance of most of Nepal. Using a light aircraft he made observations on the number of slope failures per linear kilometre of flight as viewed from one side of the aircraft. These were categorized according to ecological province. Despite these limitations, he was able to conclude that road and trail construction was associated with 5 percent of all landslides observed; that some of the most densely populated and extensively terraced areas contained the smallest frequency of landslides; and that specific regions and lithologies displayed the highest frequences - the Siwaliks and Mahabharat Lekh, and the deeply weathered phyllites. He concluded that geological structure and Ethology were far more important (accounting for more than 75 percent of all landslides in Nepal) in terms of landslide occurrence than were human land-use changes. A similar conclusion was reached following an extensive field and remote-sensing survey in a section of the Garhwal Himalaya (Josh), 1987).
Wagner (1981, 1983) used a statistical analysis and the general characteristics of a hundred landslides, mainly along roads in the Middle Mountains. He sought to develop a site-specific landslide hazard assessment and mapping approach based upon use of 'equatorial Schmidt projections,' a geometric tool for studying the intersection of geological planes. Wagner concluded that the over-riding cause of landslides and rock slides was geological (natural) and not human. These conclusions, derived from the work of Laban (1979), Wagner (1981, 1983), Joshi (1987), and the synthesis of Carson (1985), need to be carefully qualified. It could be misleading to refer simply to the stipulation that landslides are caused primarily by the impact of rainstorms on certain lithologies (i.e. that the cause is overwhelmingly natural); frequently, at issue will be the relationship between land use and Ethology.
Brunsden et al. (1981) and Caine and Mool (1982) both attributed human activities as an important cause of gullying. However, they also emphasized the importance of the material, particularly the brittle behaviour of the weathered, untransported bedrock. Caine and Mool (1982) did not find a direct relationship between heavy rainstorms and landslide occurrence, noting that a high water-table (within less than 2 m of the surface) was a primary controlling factor. Thus heavy rainfalls, early in the summer monsoon season, did not produce many landslides. As ground-water recharge progressed and the watertable approached the surface, landslide intensity accelerated. An opposing view is that landslides are most common early in the summer monsoon season because that is the time of minimum surface vegetation cover. This serves to emphasize not necessarily conflicting viewpoints, but possibly differences between specific areas, a common obstacle to any balanced overview. For the Kakani area at least, the conclusions of Caine and Mool ( 1981) have been verified by several seasons of observation and supported by many casual expressions of opinion by informants in the Kathmandu area.
Carson (1985) sums up by noting that many Himalayan landscapes appear to experience a period of relative slope stability when mass-wasting is reduced. This is followed by a brief period of excessive instability when large numbers of landslides occur almost simultaneously on otherwise longundisturbed slopes. Frequently the triggering mechanism is a catastrophic rainstorm with a long recurrence interval, such as the Darjeeling 1968 disaster (Ives, 1970; Starkel, 1972a and b) or a somewhat smaller event, such as the 10year rainstorms of the Dharan-Dhankuta area of eastern Nepal (Brunsden et al., 1981). It has also been assumed that major landslide cycles may be triggered by large-scale seismic shocks, with or without rainfall'. Carson (1985: 11) observes that villagers often tell stories about extreme precipitation events, occasionally coincident with earthquakes, that trigger large-scale slope failures. He cites the area southwest of Banepa in the Nepal Middle Mountains where the villagers maintain that all the landslides still visible had occurred during two heavy rainfall events, one in 1934 and the other in 1971. They maintained that, in spite of more recent heavy rains, no major landsliding has occurred since the 1971 summer monsoon. Another case is the heavy rainstorm of September 1981 where mountain slopes near Lele, Lalitpur district, Nepal, failed en masse in spite of a thick natural regrowth of shrub. Our photograph (Figure 5.5) taken in October 1987 shows that already, within six years, natural revegetation has all but concealed the landslide scars from the unpractised eye.
Certainly, the above commentary provides the impression that much of the large-scale landsliding not only requires a major triggering event, but that it is followed by a long period of inactivity. This is presumably required for the accumulation of a new weathered slope mantle as a necessary condition for the next major failure. Seismicity is undoubtedly important, but lack of adequate earthquake records precludes any quantification of its influence. There have also been several recent statements to the effect that presence or absence of a forest cover will have little influence on the scale of a major landslide cycle (Carson, 1985). It may even be conjectured that forested slopes merely postpone the occurrence of a major cycle and, by so facilitating the production of a deeper debris mantle, ensure larger-scale events with longer recurrence intervals. In contrast, deforested areas given over to rain-fed agriculture, and especially grazing and overgrazing, are destabilized more frequently but by small, shallow failures. Thus the integrated effect over a longer time-scale may be comparable. Carson (1985:35-6) concludes that 'Mass wasting processes are not usually directly related to man's activities. Consequently, intervention by man to reduce mass wasting can be very expensive with less clear cut results.' He also concludes that 'Flooding and sedimentation problems in India and Bangladesh are a result of the geomorphic character of the rivers and man's attempts to control the rivers. Deforestation likely plays a minor, if any, role in the major monsoon flood events on the lower Ganges.' The second point, however, will be taken up more fully (see pp. 135-144) after we have considered some other specific and important questions related to the mountains.
Erosion Rates in the Greater Himalaya
There is a dearth of precise information on rates of erosion and massmovement processes in the Greater Himalaya. The prevailing assumption is that, because the high mountains display some of the greatest relief on earth, including very high-angle glaciated slopes, as well as being influenced by tectonic instability, erosion, and hence denudation rates, must be high.
The mapping of mountain hazards in Khumbu Himal (Zimmermann et al., 1986) led to the conclusion that, with the exception of limited areas characterized by severe tectonic disturbance together with susceptible lithologies, the high-angle slopes were remarkably stable. Zimmermann et al. (1986) noted that the single most hazardous process (for humans) was sudden peak river discharges resulting from the infrequent but precipitous break-out of ice-dammed and moraine-dammed lakes. This topic is discussed separately (p. 111).
In part to offset this paucity of information, and in part to test the assumptions that high rates of erosion and mass wasting could be substantiated, Byers established thirty-five soil erosion study plots in Khumbu Himal and made precise observations at each of them at weekly intervals throughout the entire 1984 summer monsoon (Byers et al., 1985; Byers, 1986, 1987c). Total precipitation recorded for 1984 was slightly above the 28-year mean of 1,048 mm for Namche Bazar, so that the data obtained are reasonably representative.
Byers employed a standard geomorphic approach to the problem, using unsophisticated instrumentation that could be maintained easily in an isolated high mountain area. The thirty-five study plots of 5 x 5 m extended through a 1,000 m altitudinal interval (3,440-4,412 m) above and below upper timberline. They were arranged in a stratified, replicated design to permit comparison between (1) north-facing forest (Abies spectabilis/Betula utilis/Rhododendron spp.), (2) south-facing scrub grassland (Cotoneaster microphyllus/R. lepidotum/grass forte), and (3) variations in elevation. The fixed instrumentation consisted of a rain gauge, plastic sediment trough, erosion pins, and two sets of painted marker pebbles. Ground cover was determined for each plot by direct measurement at the time of instrumentation, and seasonal vegetation changes were assessed on adjacent 400 m² quadrats using a point-intercept method (800 points). Weekly observations also included air and ground temperature, soil capillary pressure, notes on plot changes and disturbance caused by local people, and weather.
Preliminary analysis of the data leads to the following findings. Summer monsoon precipitation was modest (Khumjung: 3,790 m, 725 mm compared with a 15-year average of 773 mm) and decreased with increasing altitude. However, local rain-shadow effects are marked, so that the simple relationship of diminishing total precipitation with altitude is frequently masked (compare the Khumjung total with that of 1,002 mm for Tengboche which is located in the open main valley of the Dudh Kosi-Imja Khola at 3,867 m). More significantly, from a geomorphic point of view, rainfall intensities were low (mean value of 29 measured 30-minute maximum rainfall intensities for Khumjung was only 3.03 mm/hr; and between 1968 and 1984 24-hour intervals exceeded 25 mm on only thirteen occasions with an absolute maximum of 54 mm). Cumulative precipitation increases sharply during the summer monsoon but, at most of the study plots, sediment yield was slight (Figure 5.6). The one exception was the plot at Dingpoche (4,412 m, in an overgrazed pasture) that demonstrated a high sediment yield. This was representative of an alpine meadow-juniper shrub area with a sandy soil subjected to heavy human disturbance, including up-rooting of juniper shrubs. Partial explanation for the low sediment yields is provided by the vegetation cover data. Many areas had SO percent or less ground cover in the pre-monsoon season, which partly explains why trekking-season visitors have made alarmist reports of environmental degradation and anticipated severe erosion during the summer monsoon that they did not stay to witness. In contrast, Byers observed that pre-monsoon showers and light rains early in the summer monsoon season produce a rapid spread in ground vegetation cover and hence considerable protection against rain-drop impact (Figure 5.7).
Although this study monitored only surface erosion and did not consider such processes as debris flows, talus creep, and the higher-altitude glacial and fluvial systems, when the findings are coupled with the hazards mapping results it suggests that one of the world's highest mountain regions is far less active geomorphically speaking than might be suspected and than has been generally assumed. Similar results are anticipated from a parallel installation and datacollection programme for the 1985 summer monsoon season in the Yulongxue Shan area of the Hengduan Mountains, northwestern Yunnan (Ives, 1985).
The remaining topic for consideration in Khumbu Himal is the catastrophic outburst (Icelandic: jokulhlaup) of glacial lakes. A moraine-dammed lake drained catastrophically on 5 August 1985 in a tributary valley of the Bhote Kosi (Langmoche Glacier: Galay, 1986; Ives, 1986; Vuichard and Zimmermann, 1986, 1987). This provided a remarkable opportunity for beforeand-after assessments and prompted a preliminary enquiry into the recurrence interval of such events. Hagen (1963), Hewitt (1964, 1982), Gansser (1966), Xu (1985), and others, have commented on such phenomena in the HimalayaKarakorum region. In Khumbu Himal, at least, three jökulhlaup, and possibly five, have occurred within living memory (Vuichard and Zimmermann, 1986; grower, personal communication, 1986).
While these catastrophic events, with their point sources in the Greater Himalaya, are just being recognized as serious threats to water resource development (Ives, 1986) they may be of considerable importance also in terms of sediment transfer onto the Ganges Plain over the longer geological period. The largest-known glacier lake outburst, which deposited 5.5 km³ of material in the vicinity of Pokhara, Nepal, some 600 years ago, has been referred to above (Fort and Freytet, 1982). The progressive retreat and thinning of most Himalayan region glaciers during the present century is resulting in the formation of new moraine-dammed lakes, or the enlargement of pre-existing ones between the retreating glacier front and the most recent Neoglacial moraines. Small ponds on the surface of lower glacier tongues are also enlarging and coalescing. A dramatic example of this process is a new lake that has formed since 1956 on the lower part of the Imja Glacier below Lhotse. It is now about 0.5 km², in extent; should a sudden drainage occur, the geomorphic, and human-destructive, impacts could be of catastrophic proportions (Ives, 1987, unpublished; Hammond, 1988). A cursory inspection of the metric camera imagery of the Dudh Kosi and Arun catchments in eastern Nepalsoutheastern Xizang, revealed the presence of at least fifty ice-dammed and moraine-dammed lakes (Ives, 1986). While prediction of the frequency of collapse of their ice and moraine dams must await detailed field survey, their sheer number is impressive.
Vuichard and Zimmermann (1987) developed a sediment budget for the Khumbu outburst of 5 August 1985. They concluded that most of the material moved was redeposited within about 25 km in the stream channel and only about 10-15 percent (finer fraction) was transported out of the area. Their data indicate that 900,000 m³ material was removed from the moraine dam but most of this was redeposited within the first 2 km below the breach. Much more material was picked up from the stream channel and valley sides further downstream. The peak discharge was calculated at 1,600 m³/sec some 3 kw below the source and attenuated downstream.
It is extremely difficult to develop long-term sediment-transfer averages from the scanty information available. Nevertheless, the magnitudes of these events are sufficiently high and their recurrence interval, on an areal basis, quite small (5-15 years) so that they may prove a significant source of sediment for deposition at lower altitudes. Also, even if only a fraction of the material is fartravelled during the actual event, rivers subject to periodic jökulhlaup may well prove major source areas for subsequent sediment flux during peak monsoon rainstorms as the coarse material dumped in their channels is carried further downstream. The spectacular dynamics of the (Sapta) Kosi, for instance, and the development of its vast alluvial fan which occupies much of Bihar State, may be influenced to a considerable degree by the occurrence of jökulhlaup in several of its headstreams (Dudh Kosi, Sun Kosi, and Arun).