Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
5. Mountain slope instability: natural processes or human intervention?
Introduction
The nature of mountain geomorphology: what is
known about slope process in densely populated mountain terrain?
Estimation of denudation rates in the Himalaya
Geomorphic processes and human interventions:
summing up
The previous two chapters dealt with questions and assumptions about processes of change in the area of forest cover, the causes and effects of deforestation, and the relationship between soil erosion and the transformation of forested hillslopes to other forms of land use. The conventional assumptions (myths) concerning the linkages between these dynamic mountain processes were heavily challenged and the argument was made that a much longer-term perspective and much more reliable data are needed.
Another group of linkages that hold together several components of the Theory of Himalayan Environmental Degradation relate deforestation on steep slopes and construction of agricultural terraces to a rapid acceleration in gullying and landslide incidence and increased soil erosion. It is claimed that these in turn produce serious deleterious downstream impacts. We ourselves (Ives, 1981; Ives and Messerli, 1981) held this assumption on the initiation of the Nepal Mountain Hazards Mapping Project in 1978; we have subsequently reversed our position.
As somewhat representative of'expatriate experts" making short visits, we were able to observe the Middle Mountains of Nepal from brief road traverses out from Kathmandu in 1978 and 1979. Like many other visitors, we timed our presence to coincide with good weather - March/April and October - and were duly impressed with the large number of landslide scars (strictly - debris flows) and gullies that obviously had engulfed significant amounts of agricultural terrace land during the preceding summer monsoon periods. Within this context the 'Kakani Phase' of the Nepal Mountain Hazards Mapping Project was set in motion. This included studies of stream channel morphology and landslide and gully dynamics (Caine and Mool, 1981, 1982), investigations concerning hazard perception of indigenous subsistence farmers and their coping strategies (Johnson et al., 1982; Gurung, 1988), and systematic mapping of land use, geomorphic features, and mountain hazards on a scale of 1:10,000 (Kienholz et al., 1983, 1984). Since the project necessitated repeated visits of considerable duration to the field area and fieldwork throughout the agricultural cycle over a five-year period, we were afforded the hitherto uncommon perspective of time. We were also able to develop communication with the subsistence farmers. This gave us a greater understanding of the landscape changes that have been occurring over several generations and an appreciation of local attitudes to dynamic slope processes and local responses to them.
It became apparent that many landslide scars are eventually re-terraced and stabilized and that irrigation systems are repaired: this is the most important stabilizing process (Figure 5.1). In certain instances the local people perceive a landslide to be a beneficial occurrence because the more easily worked earth of the landslide scar actually facilitates terrace construction. In other instances landslides are deliberately triggered by water diversions in order to facilitate new terrace construction (Kienholz et al., 1984; Sumitra M. Gurung, personal communiction, May 1985). We began to appreciate the complicated balance between slope stability, the particular stage of the agricultural cycle, rainfall incidence, availability of emergency labour, and type of terrace that is threatened. Thus our early predictions on rates of progressive land loss and estimated population growth (Caine and Mool, 1981, 1982; Ives and Messerli, 1981) had to be adjusted. We also observed that the indigenous farmers had evolved an intricate set of coping strategies that, in addition to subsequent reterracing of collapsed slopes, included changes in land use to match changes in slope stability; Johnson et al. (1982) introduced the concept of agricultural deintensification, as an adjustment to the threat of slope instability. Moreover, in strictly physical terms, it has been demonstrated that many of the bedrock types in the Middle Mountains, and specifically in the Kakani test area, undergo very rapid weathering and a high incidence of soil formation; thus they can withstand a relatively high rate of soil loss (Peters and Mool, 1983).
Over the slightly longer period (1978 87) frequent visits to Kathmandu facilitated repeat photography of original landslide scars and slope segments at different times of the year. Figures 5.2 and 5.3 illustrate only a single example. Nevertheless, it is thought-provoking to see that an inherently unstable landslide scar photographed in 1978 is almost invisible in 1987.
Furthermore, the new terraces that have been cut into the original landslide scar were supporting vigorous crops of maize and rice in August 1986 (Ives, 1987). The kind of data contained in Figures 5.2 and 5.3 are not adequate for regional extrapolation. They are introduced here to explain how experts, who depend on short-term visits, can be led to believe that the Middle Mountains are on the point of collapsing into the Ganges River when their observations are confined to a single tourist season (which is frequently the case) and lack a longer-term perspective and a close communication with the local people, particularly in the local language.
The experiences illustrated above are not considered as proof that there are no problems of landsliding, gullying and soil erosion in the Nepal Middle Mountains; most emphatically, there are. Rather, our intent is to argue that the worst scenarios that have been used to permeate the conservationist, development/ aid, and scientific literature may be exaggerations and, possibly, gross exaggerations that are emotionally or intuitively based. As human population numbers have increased over the past hundred years or so, leading to conversion of forested land to agricultural terraces on steeper and more marginal slopes, presumably more energy per unit of land is required to maintain a balance between stability and instability. The indigenous population may be losing ground' but not nearly so rapidly as has been assumed (Eckholm, 1976). Furthermore, it would be inappropriate to create the impression that subsistence farmers enjoy the landslide activity. Houses, human lives, and livestock are lost to landslides, and the terror of sleepless nights in small houses on steep slopes during periods of torrential rain is not to be dismissed lightly. Our aim here is to establish a better sense of proportion. It is proposed, for instance, that some of the most densely populated and extensively terraced land in the Middle Mountains probably experiences some of the lowest rates of soil erosion and land loss. A very real danger of soil erosion and slope collapse would arise if such areas were abandoned, a situation well known in the European Alps. Moreover, poor location, design, and maintenance of roads in this type of terrain has led to landslide regimes that are an intermittent or continual problem and source of stream sediment. If one looks closely at most photographs that show landslides, purportedly the result of deforestation and poor farming practices, one can almost invariably see that they have been initiated by a road or heavily used trail (Hamilton, personal communication, April 1987).
These more recent observations prompted a modified approach to geomorphic process studies during the 'Khumbu Phase' of the Mountain Hazards Mapping Project. Thus, in association with the production of hazards maps at a scale of I :50,000 (Zimmermann et al., 1986), Byers, Thorn, and Ives (1985), and Byers (1986, 1987c) planned for detailed soil erosion process studies at more than thirty plots through a range of altitude of over 1,000 m during the entire 1984 summer monsoon. The field experience in the Kakani area also prompted our close questioning of all the linkages relating deforestation and increased agricultural terracing to soil erosion, gullying and landslide activity.
Before a reasonable perspective on erosion and sediment transfer in the Himalaya can be developed, we should consider the broader issue of mountain geomorphology itself and its evolution over the past thirty or forty years. This will also facilitate the introduction and definition of a few key concepts and terms, the misuse of which has exacerbated the confused interpretation of processes affecting the Himalaya (and other mountain areas) today.
Geomorphology evolved steadily as a semi-independent discipline on the boundary between geology and geography from about 1850 to 1950. The broad concepts of landscape evolution were developed during this period. These included the concepts of base level, the geographical (or Davisian) cycle, and interruptions in the cycle, either by changes in climate or changes in base level brought about by tectonic adjustments, but progress was constrained by an overall reliance on a descriptive approach. Precise measurements of actual geomorphic processes remained a rarity. There was also an intellectual barrier in that the earth's geological history was perceived as being characterized by relatively short periods of instability, the periods of mountain building, separated by relatively long periods of quiescence during which the mountains were eroded until they were almost reduced to plains (i.e., peneplains = almost plains). This basic concept of descriptive geomorphology argued that, following a major mountain-building episode (orogenesis) sufficient time usually elapsed for streams and rivers, and the many other agents of erosion, to wear down the higher elevations until they were graded to a relatively stable sea level (base level). Thus most of the steeper slopes were eliminated and a large proportion of the terrain was reduced almost to a plain. The next cycle of orogenesis then uplifted the landmass and reactivated (or rejuvenated) the processes of erosion as the increased vertical difference between the height-of-land and sea level made available so much more additional energy. The peneplains were recognized in the accordance of summit levels in many of the most recent mountain ranges stemming from the last or Tertiary (Alpine) episode of mountain building, of which the Himalaya are amongst the most recent.
Dissatisfaction with the descriptive approach led to the 'quantitative revolution' heralded by the seminal study of Anders Rapp in a glaciated valley (Karkevagge) in northern Sweden (Rapp, 1960). The objective of this study was to rank, both relatively and absolutely, the slope-modifying processes responsible for landscape evolution in this maritime arctic area. Rapp was able to rank the dominant geomorphic processes as follows:
1. transport of salts in running water
2. earthslides and mudflows
3. dirty avalanches
4. rockfalls
5. solifluction
6. talus creep.
These are all processes characteristic of high mountain regions, regardless of latitude, including the Himalaya. Of course, any detailed consideration would emphasize an array of differences between the mountains of northern Sweden and the Himalaya, the major ones being climatic regime, vegetation cover, altitude, and rock type and structure. Human activities are also vastly more important in the Himalaya. It must also be borne in mind that there are considerable differences from one part of the Himalaya to another. A major question concerns the extent to which these processes and their ranking vary from one mountain area to another. Moreover, Rapp, for his analysis of the Karkevagge area, excluded consideration of glacial erosion (probably the most effective erosion process where glacier systems occur) and it was not possible to rank slope wash because of practical difficulties in collecting adequate data. Frost-bursting, the prying-off of rock particles from the valley side cliffs by freezing and thawing of water in joint planes and crevices, was not included, partly because it is not considered as a transporting process and partly because it is difficult to measure directly. The annual production of rock-waste by frostbursting on rock walls, however, was calculated at 100400 tons/ km² of wall surface, which would probably give it top ranking over the processes listed above. Rapp also concluded that, without many more data from other valleys, with other types of slope, these results could not be extrapolated to the immediate local region; the ranking of the six processes could be significantly different in adjacent valleys with somewhat different slope combinations even though they were subject to the same type of climate.
Rapp's work set the stage for process studies in many parts of the world. These ranged from Alaska to Tasmania, from New Zealand to the Canadian and Colorado Rockies, from the Alps and the Tatra to the Khumbu Himal. But already in Rapp's conclusions the problems were recognized that still beset any full understanding of what is often referred to as climatic geomorphology. To complete his treatise, Rapp and assistants had collected data during eight years, yet the results, as indicated above, cannot be extrapolated with confidence to Karkevagge's neighbouring valleys, let alone to the Dudh Kosi of Nepal; nor can they be extrapolated backward in time, nor used as a basis for prediction. In addition, despite Rapp's top ranking of transport by salts in running water, most subsequent research has concentrated on mechanical weathering and transport of coarse debris, with a very heavy emphasis on talus slopes and solifluction, the two lowest-ranked processes. Moreover, heavy rains in October 1959, Rapp's final year of data collection, set in motion processes, principally mudflows and debris flows, representing by far the largest geomorphic event to occur during the entire period of fieldwork. Thus the concept of the large event with a long recurrence interval was introduced very early in the 'quantitative revolution'; it remains a major dilemma for any assessment of the development of mountain slopes in the Himalaya and raises the problem of representativeness in time as well as in space.
During the period 5-7 October 1959 the Riksgransen weather station on the Narvik-Kiruna railway recorded 107 mm of rainfall in 24 hours and 175 mm in 72 hours. Total precipitation for the four months, July to October 1959, was 794 mm compared with the 1901-30 annual average of 308 mm. The October rainstorm was the heaviest since the Riksgransen station was established in 1904; the recurrence interval of such a downpour may exceed a hundred years. (Rainfall intensities such as this are not uncommon in the Himalaya.)
The effects of the hundred-year climatic event, setting in motion catastrophic erosive activity, also may be dwarfed by even more spectacular occurrences. For example, Heuberger et al. (1984) have documented a giant landslide in the Langtang Himal, Nepal, which occurred about 25,000 years ago. This landslide, which generated fused rock (frictionite) along its sliding surface, displaced approximately 10 km³ of debris through a vertical distance of up to 2,000 m.
One of the most spectacular geomorphic events to have occurred in historic time in the Nepal Himalaya was the outburst of a moraine-dammed lake behind the mountain Machapuchare between 600 and 800 years ago. This caused a flood surge down the Seti Khola which deposited 5.5 km³ of debris in the Pokhara Valley, damming Lake Phewa (Fort and Freytet, 1982). On a somewhat smaller but still catastrophic scale, in October 1968, rainfall, varying in amount between 600 and 1,200 mm, fell on the Darjeeling area, West Bengal Himalaya, during a three-day period at the end of the summer monsoon when the ground was already saturated. It is estimated that some 20,000 debris flows were released; the 50-km road between Siliguri on the plains and Darjeeling at 2,200 m was cut in 92 places and approximately 20,000 were killed, injured, or displaced (Ives, 1970). While there is some disagreement concerning the estimate of the recurrence interval of this event, Starkel (1972a and b) concluded that occurrences of this magnitude are the primary slope-forming processes and calculated an average denudation rate for the Darjeeling area of the order of 0.5-5 mm/yr and up to 20 mm/yr for the individual years when such catastrophes occur. These are amongst the highest denudation rates ever proposed and the implications are disussed in more detail below.
The regularity, or irregularity, of occurrence of the extraordinarily large events, such as the Seti Khola torrent, the Langtang landslide, or even the Riksgransen-Karkevagge rainstorm, poses a serious problem for any attempt to rank geomorphic processes and to deduce long-term denudation rates. A primary difficulty is the problem of estimating the magnitude of very long recurrence intervals when there is no historical record to give adequate control; another is the actual identification of the enormous deposits that result from such large-scale events. From all of these considerations it will be appreciated that the task of determining the relative importance of catastrophic events to total landscape evolution in space and time faces severe difficulty.
Within the same context as variations in large-scale geomorphic activity through long periods of time, rates of slow mass-wasting that operate continuously in millimetres per year also appear to vary through time. For instance, Benedict (1970) demonstrated for the Colorado Front Range that rates of mass-wasting are lower now than at any other time during the Holocene (the last 10,000 years or so) and, within this period, have varied by an order of magnitude. Current rates of movement range from 4 to 43 mm/ yr in the uppermost 50 cm of surface weathering mantle, and up to an order of magnitude higher during the close of glacial episodes when higher soil moistures can be assumed (Benedict, 1970).
The foregoing discussion amply illustrates the difficulty of ranking slow mass-wasting processes, such as solifluction and soil and frost creep, with medium-scale events, such as the 1959 Riksgransen-Karkevagge and the 1968 Darjeeling rainstorms, and with the giant events such as the Langtang landslide and the Seti Khola torrent. These are problems that face geomorphologists in areas such as the Colorado Front Range, the Alps, and Karkevagge that are blessed with ease of access, a decade or more of accumulated data, excellent topographic maps and air photograph coverage, and even permanent mountain research laboratories (Ives, 1980). How much more difficult, therefore, is such ranking of geomorphic processes in the relatively inaccessible Himalaya?
Much of the present-day knowledge concerning the effectiveness of different geomorphic processes, especially mass-movement processes, has been gleaned from field observations in high mountains. By this we mean from areas above the upper timberline and more especially from sites picked because they are largely unvegetated and thereby characteristic of the mountain landscape where processes are operating most rapidly, where there are talus slopes, rock glaciers, block fields, and solifluction terraces on debris slopes with a broken vegetation cover. The reasons for this are pragmatic. If the field scientist is to obtain observations on the movement of rock particles that are significantly larger than the magnitude of error inherent in his instrumentation, and to complete this in a reasonable length of time - we should remember that Rapp needed eight years, and data sets collected over a decade are not unusual - then perforce his field site must experience high rates of mobility. While this overemphasis on high mountain, largely unvegetated sites can be offset by estimation of regional denudation rates from river sediment load at lower elevations, as Caine has demonstrated in his concept of the alpine cascade of sediment fluxes (Caine, 1974; Barsch and Caine, 1984), there is frequently a lack of coupling between alpine hillslope systems and the fluvial system. In other words, much of the geomorphic 'work' measured relates to mass movement of debris that remains within small alpine watersheds and does not pass through the system as fluvial sediments. It is perhaps understandable, therefore, that data sets and hypotheses acquired from field studies in high mountain environments must be applied to middle and low mountain belts with great care. This is especially important in an area such as the Himalayan Middle Mountains where frequently landscapes are totally transformed by subsistence agriculture (Figure 5.4).
Despite this qualification all available geomorphic studies have demonstrated a difference in degree, rather than in kind (the glacial system excepted) of erosion between high mountains and regions of less pronounced relief. In very simplified terms, this means that the same basic processes operate more rapidly (= effectively) on steep slopes than on less steep slopes. This difference in degree appears to be consistent and measurements, however sparse and unrepresentative they may be, range from five times to one or two orders of magnitude, based upon river sediment load estimates (that is, the denudation rate, or overall rate of landscape lowering), reservoir sedimentation, and geological data (the sediment record in the plains representing the longterm accumulation of material eroded from the neighbouring mountains).
Another perspective necessary for this review is that of the rates of mountain building and regional denudation within the context of geologic time. With the widespread acceptance of the theory of plate tectonics over the past twenty years our appreciation of the dynamism of continued mountain building has advanced considerably. Several estimates of present-day rates of Himalayan uplift have been published. Zeitler et al. (1982) indicate a rate of uplift for the Greater Himalaya of about 1 mm/yr; Low (1968) estimates 1-4 mm/yr since the close of the Lower Pleistocene; Iwata et al. (1984) about I mm/yr for the Nepalese Himalaya; and Zeitler et al. (1982) about 9 mm/yr for the Nanga Parbat region. Extensive geophysical work by the Chinese Academy of Sciences (Liu and Sun, 1981) on the Tibetan Plateau and the northern slope of the Himalaya has resulted in figures of 4-5 mm/ yr over the past 10,000 years, continuing today. Precise measurements in the vicinity of Garm, Tadjik SSR, over the past thirty years indicate that Peter the First Range, an outer range of the Pamir Mountains, is rising at a rate of 15 mm/yr. The very crude estimates of regional denudation rates, discussed below, barely match those of the uplift estimates. It is argued, therefore, that at present, uplift equals or even exceeds denudation in some areas, thus implying that the Himalaya-Ganges system over the past 10,000 years has continued to be an extremely dynamic section of the earth's crust.
Because the Himalaya and the Tibetan Plateau are being uplifted as the Indian plate thrusts beneath Central Asia, the enormous masses of eroded sediment are deposited in the foredeep to the south which, over the past several million years, has become, in effect, the great plains of the Indus, Ganges, and Brahmaputra. This point is nicely emphasized if we consider that drill holes have penetrated more than 5,000 m of alluvial sediments beneath the Ganges Plain (C. K. Sharma, 1983). On a more recent time frame, the (Sapta) Kosi River has shifted its channel across its great alluvial fan, which forms much of Bihar State, through a distance of more than 100 km in the past 250 years. There is, therefore, abundant evidence of massive erosion and regional denudation and equally massive sediment transfer and deposition that has occurred over the past million or more years. Present-day evidence and geophysical hypothesis would indicate that the height of the Himalaya is equal to, if not higher than, that of a million years ago, and that the relief energy between the crestline and the Ganges Plain remains undiminished over recent geological, as well as historical, time. Thus, without very convincing evidence to the contrary, it would seem reasonable to argue that the contribution of human interventions over the past three or four decades, or even centuries, has been insignificant when balanced against the natural processes at work. Before this point is examined further it will be helpful to bear in mind a few broad concepts and to reiterate some of the dilemmas facing geomorphic and hydrological research.
1. The Himalaya-Brahmaputra-Ganges-lndus system is one of the world's most dynamic mountain-building and sediment-transfer systems, processes that have continued unabated over recent geological time and will likely continue into the future.
2. These processes, the endogenous, tectonic/ isostatic activity and the exogenous, climatic/ weathering/ hydrological ones, have created an unstable landscape of the utmost complexity.
3. Given the massive scale of relief, from more than 5,000 m below sea level to nearly 9,000 m above sea level, and the enormous variations in climate, vegetation, and topography, and the variability of major geomorphic events in time and space, the present data base is completely inadequate for determination of actual rates of activity of the various processes affecting the land surface. Thus, determination of the impacts of human intervention, including deforestation, land-use changes, and manipulation of water flow, and their differentiation from the natural processes as a proportion of the total rate of change, is not possible.
Regardless of this apparent total obstacle to geomorphic evaluation, some important contributions are feasible provided they are set in the context of recent geological time. It is equally important to question the widespread tendency over the past thirty years or so to assume that problems of erosion affecting either the plains or the mountain slopes themselves are entirely, or largely, the result of human intervention, in terms of misuse of the Himalayan environment.
Some critical definitions are needed. First, soil erosion (or surface erosion), a widely used term, should be restricted to the secular loss of soil, especially the A-horizon in which most organic matter is concentrated. Soil erosion occurs in entirely natural environments as well as in environments transformed by human intervention; for the latter it is convenient to use the term accelerated erosion, which implies a combination of a natural process and a man-induced process. Soil erosion should be distinguished from mass-wasting, which is the down-slope movement of the mass of fractured and weathered bedrock, the weathering mantle, on which the topsoil forms as the end-product of that weathering process. Mass movement includes such almost imperceptible, continuously operating processes as soil creep, frost creep, and solifluction by which the weathering mantle moves downhill under the influence of gravity at rates of a few millimetres per annum. Mass movement also includes more dramatic, intermittent processes such as landslides, mudflows, rock falls, and rockslides with short, or long, indeterminate, recurrence intervals.
Soil erosion, whether natural or accelerated, and mass movement, in practice, grade into each other, but to facilitate a clearer understanding of this section of the Theory of Himalayan Environmental Degradation, they should be retained as conceptually separate processes in landscape change. It should also be borne in mind that the weathering mantle and the topsoil are continually forming as the bedrock is broken down by a combination of mechanical and chemical processes. In certain instances in nature the topsoil and weathering mantle may be shed from a slope (for instance, during a cycle of landsliding) and the partially weathered bedrock exposed. This type of rapid mass movement will usually be followed by a long period during which the weathered mantle and its vegetation cover will be replenished. On steep mountain slopes climax, or zonal, soils may never develop because the slopes are too unstable to allow a mature soil cover to evolve. In these circumstances agricultural terraces actually reduce slope instability and the soils developed on them are at least partially, and in some cases largely, man-made. In other words, the immature azonal mountain soils receive much of their organic matter from the addition of crop residues and domestic animal fertilizer. Soil loss is characteristic of all natural and man-modified slopes; it becomes a problem for subsistence farmers only when the decline in soil productivity due to topsoil losses cannot be compensated for by addition of nutrients from organic matter, and continued accumulation of inorganic matter.
Effective erosion, if the soil or weathered material is to be moved out of the immediate field area (small watershed, or hillslope segment), requires the assistance of a transporting agent. The most effective agent of transport is running water (hence the formerly popular concept of the 'normal' cycle of erosion, that in which transport by running water predominates: the fluvial cycle of erosion). Thus information on the relationship between mass movement on slopes and water transport is important. The river itself is an agent of erosion as well as an agent of transport and, by cutting its channel and undermining its banks, the river is the principal force in maintaining local relief energy in an orogenically active mountain range. The river is also responsible for deposition of its transported load, and hence for the formation of the plains. Glaciers and wind are effective transporting agents but will receive little attention here because of their minimal spatial significance in the intensely used Himalayan belts - the Middle Mountains and the Siwaliks.
Denudation is a term used to describe the overall lowering of the landscape resulting from the erosive and transporting activities of all operating processes. In practical terms this is usually calculated as mm/yr in surface lowering averaged over entire regions, usually watersheds (drainage basins), despite the fact that actual surface lowering will be extremely variable in space, as it depends upon many factors, including the underlying bedrock lithologies and structures, and slope angle. Regional denudation estimates are derived in two ways: they are obtained from numerous observations on the principal erosion processes characteristic of a watershed multiplied by the total area (as exemplified by Rapp's (1960) study of Karkevagge), or they are extrapolated from measurements of sediment being moved out of the watershed through the main stream channel. In either approach the sediment delivery ratio must be taken into account. This is the ratio of sediment yield in a river - the actual volume of material transported out of the watershed - to the gross sediment production upstream, much of which goes into temporary storage within the watershed. Mass-movement data cannot be directly translated into denudation because much of the material moved remains within the watershed in storage, for instance in the form of lake sediments or as accumulations (talus, glacier moraines, landslide deposits, footslope colluvium) lower on the slopes or on the valley floors.
River transport is conveniently divided into the suspended load, the load carried as dissolved salts, and bedload. The sediment yield that is actually measured is often only the suspended sediment. Even this is difficult to measure accurately for rivers that experience enormous variations in volume over the course of the year and from year to year. This is especially characteristic of rivers of monsoonal climates where low flow in late winter and spring may be several orders of magnitude below peak rainy-season discharge. This problem of measurement is exacerbated when we consider that Himalayan river channels are frequently dammed by landslides; the ensuing ephemeral lake, when it breaks the dam, will produce a peak discharge and be capable of carrying much larger sediment loads, sometimes an order of magnitude or more, than that of normal summer monsoon peaks. Even these periods of high sediment yield may be totally eclipsed by peak surges resulting from the outbreak of ice-dammed and moraine-dammed lakes. The critical importance of such catastrophic floods to Himalayan water resource development have only been recognized in recent years (Hewitt, 1982; Xu, 1985; Galay, 1986; Ives, 1986; Vuichard and Zimmermann, 1986, 1987).
Measurements of bedload have not been recorded on any Himalayan river, even under conditions of 'normal' summer flow. An estimate is usually made for this component of the total sediment transfer in calculating the design life of reservoirs. Bedload is now regarded as having been grossly underestimated systematically throughout the Indian and Nepal Himalayan foreland (this is a particularly critical observation since hundreds of millions of dollars have been expended on dam and reservoir construction with the design life overestimated two-, three- and fourfold). Finally the dissolved load of rivers in the Himalayan region is completely unknown. The importance of this lack of knowledge can be understood if we refer back to Rapp's (1960) ranking of the transport of dissolved salts in running water in northern Sweden. In the Himalayan region, which at lower elevation is sub-tropical, this component of a river's sediment load will not be of less significance.
It is now appropriate to consider the data that are available on sediment yield, erosion, mass movement, and denudation.