Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

6. The Himalayan-lowland inleractive system: do land-use changes in the mountains affect the plains?

Regional assessment of watershed degradation
Downstream effects of watershed degradation



The previous three chapters have discussed the linkages between those segments of the Theory of Himalayan Environmental Degradation that relate primarily to the physical processes occurring, or that are assumed to be occurring, within the mountains. This chapter focuses on the larger question: assuming that accelerated erosion due to land-use changes is taking place on a vast scale in the mountains, what is the evidence in support of claims for dramatic downstream impacts? We will now attempt to estimate the regional distribution of watershed degradation in the central Himalaya and its hydrological and geomorphic effects downstream.

Much of this chapter depends upon a major data collection and preliminary analysis by Andreas Lautherburg of the University of Berne (Lauterburg and Messerli, 1986, unpublished). During several visits to the region he collected as many data as possible on soil erosion and watershed degradation in Nepal, India, and Pakistan. In addition he collected all available data on stream flow (runoff) and sediment load of the main Himalayan rivers. Much of the critical information was found in publications such as Laban (1978, 1979), Zollinger (1978, 1979), and Carson (1984, 1985), and in the files of four government and international organizations: the Department of Soil Conservation and Watershed Management, HMG, Kathmandu; the Remote Sensing Centre, HMG, Kathmandu; the Central Soil and Water Conservation Research and Training Centre, Dehra Dun, Uttar Pradesh; and the International Centre for Integrated Mountain Development (ICIMOD), Kathmandu.

The data-acquisition phase of the study is by no means complete. This is partly due to time and funding limitations, but also to the fact that many data on hydrology and sediment transfer were not accessible.

The available data can be divided into two groups: direct and indirect. An example of the former is the direct measurement of soil loss from controlled test plots. This kind of information, as indicated in the preceding chapters, is both rare and unreliable (unreliable partly in the sense of lack of representativeness). Nor is it a major contribution for estimating the spatial distribution of degradation on a regional scale. Thus, of more value is the indirect information; examples of this are time series on the suspended sediment load in rivers and the density of landslides per unit area.

Much of Lauterburg's study of erosion records of test plots and small catchments parallels the discussion in Chapter 5. He concludes that, at the local (micro) level, soil erosion is highly influenced by human impact and that corrective measures could reduce this dramatically. We also wish to reemphasize the positive aspects of certain forms of human intervention. Lauterburg also supports our earlier contention that the conversion of mountain lands under natural vegetation to an agricultural landscape does not automatically result in an increase in soil erosion (accelerated erosion) since soil loss is not dependent upon natural versus domesticated soil cover but on a conservation factor; an extreme case, for instance, is the positive influence of carefully tended agricultural terraces. This part of Lauterburg's analysis is taken no further here and we will discuss now the indirect data and its assessment.

Regional assessment of watershed degradation

Lauterburg analysed the work in Nepal by Nelson (1980), Laban (1978, 1979), and others. Nelson, for instance, attempted to determine the general status of degradation of all Nepalese watersheds by visually estimating a 'watershed condition index.' 'This 'relates the current state of soil erosion in an area in comparison with the soil erosion estimated for that area under natural or well managed conditions' (Nelson, 1980:2). This study shows a significant concentration of heavily degraded areas (accelerated erosion) in the central and lower Siwaliks and in the Kathmandu area.

Lauterburg reassesses Laban's landslide count from a light aircraft (Laban, 1979, and see above, pp. 106-7). This provides us with a Natural Landslide Index, which is obtained by dividing the number of landslides occurring within forested areas by the total number of landslides. This approach suffers from the uncertainty of being able to distinguish between 'natural' forests and degraded forests (lopped, partly grazed, etc.).

A complementary approach is the use of data on suspended sediment load in rivers. By assuming that the sediment load of several rivers is measured accurately over a significant number of years, the total (average) amount of sediment yield per annum divided into the total area of the watershed will provide the so-called Specific Suspended Sediment Delivery (SSD) of a catchment. This, as discussed in Chapter 5, does not include all material eroded within a watershed, since much is left in temporary storage and does not enter the fluvial system. In addition, there were no available data for bedload or solutes. However, as long as the watersheds under comparison are of approximately the same size and form (that is, their Sediment Delivery Ratios are similar) the SSD does provide a useful comparative unit. Table 5.1 gave the SSD rates (actually denudation rates) calculated in this way for Nepal. Figure 6.1 shows the Suspended Sediment Delivery for most of the Himalayan Region. However, this map has been compiled from very different sources which cover different time periods. Thus the SSD data for the Indus watershed was collected in the late-1950s (Ahmad, 1960) and may be higher today. The somewhat higher denudation rates than those given in Table 5.1 are due to the use of estimated suspension data for Figure 6.1, especially for eastern Nepal.

Figure 6.1 Suspended Sediment Delivery of some Himalayan rivers. Prepared by Andreas Lauterburg, Geographical Institute, University of Berne, with data from Ahmad (1960 - Indus), Goswami (1983, 1985 - Brahmaputra). and many sources (1974-85).

Despite these obvious limitations, together with actual gaps in data availability, a broad-brush comparison may still be useful. Examination of the three macro-regions, the Indus, Ganges, and Brahmaputra watersheds, demonstrates immediately that SSD rates in the Nepalese Himalaya are much higher than the other two regions. Such a map (Figure 6.1), if updated with dependable information, would be extremely valuable. Unfortunately, this cannot be done at the present time. Another problem with this approach is that, whatever we learn about the current erosion rates, there is no time perspective. If an attempt is to be made to assess the relative, or absolute, proportion of total SSD due to human rather than natural (geophysical) processes, we would need to know the change in sediment load over the past hundred years or so. Moreover, no attempt has been made to estimate the stabilizing input by man through terrace construction and their effective maintenance.

As indicated in Chapter 1, the conservationist and scientific literature is replete with qualitative estimates of landscape degradation in specific small areas - this is largely what set in motion the Theory of Himalayan Environmental Degradation - but no attempt has been made to produce quantitative data nor, especially, to link them with the large-scale processes occurring on the flood plains and deltas of the major trunk streams.

In contrast, Figure 6.2 shows the lithological erodibility (specifically, the susceptibility to weathering and erosion of the bedrock) of the Karnali watershed in Nepal divided simply into low, medium, and high erodibility. The close relation between Ethology and susceptibility to erosion is apparent. Of particular significance, the largest area of 'high erodibility' is situated in the high Himalayan zone which has a very low population density. However, this map shows susceptibility to erosion and not actual sediment yield. Rambabu et al.'s (1978) map of annual erosivity in northern India and Nepal (Figure 6.3) is more useful. This is based on a regression equation which calculates erosivity from monthly or annual rainfall data, incorporating 30-minute maximum rainfall intensities and total kinetic energy.

The construction of iso-erosivity maps normally depends on availability of an extensive network of stations that record short-term rainfall intensities as well as the catastrophic climatic events with very long recurrence intervals. As mentioned above there is a great shortage of these kinds of data for the Himalaya and the problems associated with this approach have been discussed in Chapter 5. Nevertheless, we concur with Lauterburg's conclusion that, compared with the other great mountain systems of the world, the Himalaya experience very intense erosivity and high probability of catastrophic highintensity rains, and are at a very severe risk of climatic erosion. Even in making this very general statement, however, we must once again emphasize the gaps in available data, as well as poor data quality.

The discussion of the ratio of natural (geophysical) erosion to that caused by human intervention (accelerated erosion) is taken a step further by Lauterburg. He compares maps of natural erosion risk (Ethology and climatic factors) with maps indicating the state of watershed degradation and landslide frequency in Nepal (Figures 6.4, 6.5 and 6.6). Figure 6.4 indicates susceptibility to erosion according to a combination of Ethology and climatic factors with data assembled on a grid. Figure 6.5 demonstrates watershed condition, incorporating Nelson's (1980) and Laban's (1979) determinations of landslide incidence induced by human intervention. Finally, Figure 6.6, which combines data from Figures 6.4 and 6.5, identifies those areas in Nepal where human activities have had very high, high, and moderate impacts on watershed degradation.

The data sets used in the compilation of Figures 6.4-6.6 include quantitative and semi-quantitative information as well as subjective estimates. There are also significant gaps in the data base. Figure 6.6 is principally a qualitative estimation of risk of watershed degradation. As anticipated in the previous discussion, there is a heavy concentration of 'very high impact,¹ which is defined as low natural risk and high actual watershed degradation, in the central Siwaliks and the Kathmandu Valley and adjacent Middle Mountains. The much-discussed heavy damage in the Tamur and Arun watersheds in eastern Nepal does not appear and the Middle Mountains west of Pokhara are largely blank. The implication is, of course, that according to this approach, human impact (negative) must be quite low.

Figure 6.2 Lithological erodibility of the Karnali watershed, Nepal. Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Shrestha. 1980).

Figure 6.3 Iso-erosivity map for northern India and Nepal. *Annual erosivity (joules per square metre per millimetre per hour). Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Narayana and Rambabu, 1983; Mohns, 1981).

Figure 6.4 Climatic and lithological erosion hazard in Nepal. Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from many sources including Shrestha, 1980; Mohns, 1981).

Figure 6.5 Watershed condition and human-induced landslide increase in Nepal. Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Nelson, 1978; Laban, 1979).

The approach of Lauterburg is interesting and, if more data can be accumulated, future construction of such maps could provide a valuable guide to land reclamation and watershed protection policy development. The present attempt, however, while giving some useful indicators, is not only limited because of data availability and accuracy, but cannot take into account the indigenous land-reclamation efforts of the farmers. Nevertheless, and regardless of gaps in data availability, Figure 6.6 creates the impression that the areas of high human impact in Nepal are very restricted. We are still left with the impression, however, that considerable areas of Nepal are in a condition of potential instability whereby heightened subsistence-farming pressures, or reduced maintenance of agricultural terraces, could lead to a rapid and dramatic increase in watershed degradation.

Figure 6.6 Human impact on watershed degradation in Nepal. Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Nelson, 1978; Laban, 1979; Shrestha, 1980; Mohns, 1981).

Downstream effects of watershed degradation

Natural erosion in the Himalaya has been shown to be an important phenomenon and is probably higher than in most other major mountain systems. This implies that the Himalaya as a region is experiencing some of the highest, if not the highest, denudation rates in the world. This is due to the monsoonal character of the climate with high annual total precipitation concentrated in a three- to fourmonth period on an area of very high relief, susceptible lithologies and structures, and high seismic incidence. Further more, there is great annual variability in rainfall totals and occasional dangerously high rainfall intensities. And it is clear that this condition has existed for several million years.

It has been widely claimed in the literature that the devastating annual floods in the Ganges and Brahmaputra lowlands are influenced by extensive deforestation and intensified land use in the mountains. The human component of the total streamflow cannot be identified from any available data. Nor do the existing publications demonstrate any significant recent increase, either in sediment load of the larger rivers and tributaries, or in the magnitude of the annual flooding and levels of river discharge. Nor has any attempt been made to determine quantitatively the human impact on sedimentation and flooding on a large scale. Despite this it has been shown that, in very small watersheds, erosion and streamflow are highly influenced by man (Tejwani, 1984b, 1987). This prompts us to differentiate three scales, or sizes, of watershed to further our assessment of the downstream impacts of watershed degradation: the microscale; the meso-scale; and the macro-scale.

The somewhat arbitrary differentiation into three scales is illustrated schematically in Figure 6.7. The entire watersheds of the Brahmaputra, Ganges, and Indus, each represent the macro-scale. The meso-scale is illustrated by major Ganges or Brahmaputra tributaries, such as the Teesta, the Kali Gandaki, and the Karnali. The micro-scale can be illustrated by the Thulo Khola and Ghatte Khola in the UNU Kakani test area close to Kathmandu (Caine and Mool, 1981). Thus an approximate size class is as follows, bearing in mind that we will fudge this by including the (Sapta) Kosi in the macro-scale class for reasons explained below:

Figure 6.7 Schematic representation of the relationship between human and natural processes at three different scales.

Micro-scale < 50 km²
Meso-scale 50-20,000 km²
Macro-scale > 20,000 km²

Figure 6.7 demonstrates hypothetically how successful soil-conservation measures, traditional or introduced, can be in a degraded micro-watershed. Conversely, it also indicates the dangers of inherent slope instability and implies that adverse human impact can produce high-magnitude degradation. The value of soil conservation and watershed management practices on this scale have been amply demonstrated for watersheds of a few hectares (Gary, 1971; Chatra Research Centre, 1976; Mathur, 1976; Kollmannspererger, 1978/9; Impat, 1981; Christiansen, 1982; Tejwani, 1982, 1987; Narayana and Rambabu, 1983; Rambabu, 1984; and CSWCRTI, Dehra Dun, Annual Reports).

The lower and upper limits of the different watershed size classes are arbitrary, partly because there are insufficient data available for the Himalayan region to warrant a more precise approach. However, we wish to emphasize that this differentiation into three rough size scales of watershed is useful because the relative importance of human interventions within a watershed, and the downstream effects, change with watershed size. Because of limited data, however, critical sizes cannot be determined exactly.

There are good reasons for the foregoing statement. In small watersheds, as we have demonstrated, streamflow is less influenced by human activities than is the actual sediment load of the rivers. It is also reasoned that exceptional climatic events (for example, very high intensity rainfalls) will reduce the relative importance of human activities (accelerated erosion) compared to overall natural processes. The larger the size of the watershed, the greater will be the probability for local heavy rainfalls and these will also more likely influence the meso-scale watersheds downstream of the rainfall locality. Furthermore, large watersheds tend to have a smaller proportion of agricultural land to total area than small watersheds. This characterization only applies, of course, to micro-watersheds within the intensely settled Middle Mountains and lower attitudinal belts in comparison with the larger watersheds of which they are a part. For instance, micro-watersheds at high altitudes will contain little or no agricultural land. We therefore wish to emphasize the condition of mesoscale watersheds which contain a range of altitudinal belts with different degrees of land-use intensity, in comparison with a typical micro-watershed, such as the upper Bagmati (Kathmandu Valley), or the Ghatte Khola, in Nepal. These are almost totally transformed into an agricultural landscape: the natural vegetation and even the original natural slopes, have been virtually eliminated (see Figure 5.4). A typical mesoscale watershed, such as that of the Trisuli, or Karnali (Nepal), embraces a high mountain belt with areas of steep rock, glaciers, and steep forested slopes where human subsistence activities are reduced to minute levels. Also, the larger watershed will possess a greater range of natural retention basins (lakes) that will reduce the downstream effects of processes occurring in the upper watershed (see also Vuichard and Zimmermann, 1987).

The Micro-watershed

Figure 6.8 is a graphic display of the rainfall, runoff, erosion potential, and actual soil loss for a small watershed near Chandigarh, Uttar Pradesh, that has been subjected to soil-conservation measures. Hydrologic response to treatment after 1965 as well as the reduction in soil loss are impressive. Comparison of the soil loss curve with the 'percent runoff to precipitation' (runoff = streamflow) curve leads to the conclusion that soil loss is much more effectively influenced by land-use practices (in this case, soil-conservation measures) than is streamflow.From this it follows that the high variability of rainfall (Parthasarathy and Mooley, 1978), especially in the western and lower ranges of the Himalaya, has a much more important influence on measured streamflow regime than on soil loss. An important exception is the type of catastrophic rainfall event, such as described by Starkel for the Darjeeling area (1972a and b), which influences both streamflow and sediment transfer. Nevertheless, we can conclude that soil-conservation measures are more useful in preventing, or reducing, soil loss than in modifying the hydrological characteristics of the treated areas.

Figure 6.8 Annual rainfall, runoff erosion, and actual soil loss from small treated watershed near Chandigarh, India. Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Central Soil, Water Conservation, Research and Training Institute, Dehra Dun: Annual Report 1976).

In terms of reservoir sedimentation it also follows from the foregoing discussion of the micro-watershed that soil-conservation measures can be extremely valuable in correcting a damaging situation for reservoirs in small catchments. However, the next sections will demonstrate that for meso-scale and macro-scale watersheds high reservoir sedimentation rates probably must be considered as an inevitable natural phenomenon.

The Meso-watershed

Lauterburg's search for data series over a long period illustrative of meso-scale watersheds was unsuccessful. But since we can assume that watershed degradation theoretically should influence streamflow characteristics an attempt will be made to interpret the few sets of streamflow data that are available for meso- and macro-scale rivers.

Figure 6.9 shows the annual peak discharge of the River Teesta from 1956 to 1975. The Teesta originates in Sikkim and is tributary to the Brahmaputra. The curve graphically illustrates the impact on peak discharge of the catastrophic rainstorms of 1968 and 1974 in the Darjeeling Himalaya. However, during the period for which streamflow data are available there is no demonstrable tendency for a trend toward higher or lower peak discharges. A full interpretation of the Teesta streamflow curve also requires access to watershed precipitation data as well as information on land-use changes over the same period. These are not available. In general, we do know that the Teesta watershed has been extensively modified by the spread of tea plantations and subsistence and market-gardening plots and that the area under these uses has increased since the early 1960s (Starker, 1972a). It is also significant that catastrophic floods occurred in 1950, 1968, and 1974. The eighteen years between the first two events is perhaps long enough for local people largely to forget the impacts of the first event. The interval between the second and third events, however, is quite short, which can create an impression of increasing flood frequency. Nevertheless, the streamflow curve clearly demonstrates that the average streamflow has not increased; consequently we must regard the 1974 flood as a purely natural phenomenon. Because of the lack of the ancillary land-use and precipitation data and the short period of river flow record (twenty years) this statement is made as a working hypothesis which should be tested as more data become available.

For the meso-scale watershed, therefore, downstream effects of landscape degradation are frequently stipulated but seldom have been demonstrated, and then only at the scale of a test plot (10 x 10 m²) or micro-scale watershed. Moreover, the dimension of anthropogenic downstream effects is not understood quantitatively. There is little information presently available with which to approach this problem. We conclude that the high intensity and high variability of natural events obliterate the effects of human interventions.

Figure 6.9 Annual peak discharge and recurrence interval of River Teesta, northern West Bengal, India (Teesta Bridge gauging station). Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Gosh, 1983, in Central Board for Irrigation and Power).

The Macro-scale Watershed

Macro-scale rivers in India and Nepal are much better documented than those of meso-scale watersheds. Nevertheless, because of the very recent formation of Nepal's Hydrologic Service (Shanker, 1983) only the big rivers in India have long-term data sets. Even here, however, many of the data collections of streamflow and sediment load are'classified' end not available for scientific analysis one of the especially unfortunate aspects of 'uncertainty on a Himalayan scale.' Some information is available for the Ganges, Brahmaputra, and (Sapta) Kosi and it will now be discussed.

Figure 6.10 provides data on streamflow, sediment load, and high- and lowflow hydrographs for the Brahmaputra between 1955 and 1979. This three-part figure provides an outstanding example of the risk of misinterpretation of relatively long-term data. The annual streamflow curve shows a significant increase in streamflow volume between 1969 and 1979. If we disregard the antecedent decrease between 1955 and 1969 we could interpret the increase over the 1969-79 period as evidence for the impact of human land-use change in the watershed. However, even for such a large river system as the Yarlungtsangpo-Brahmaputra, annual streamflow variability is extremely high (up to 100 percent). In addition the fluctuations in average streamflow over periods of approximately ten years are remarkable. Zollinger (1979), for instance, has indicated that major Himalayan rivers sometimes behave like mountain torrents. It follows, therefore, that long-term streamflow data series of less than about twenty years should not be used for calculating trends.

Figure 6.10 Brahmaputra annual runoff, sediment load, and high- and low-flow hydrographs, 1955-79 (Pandu gauging station). Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Goswami, 1983).

Figure 6.10 also shows variations in sediment load. This was extremely high in the late 1950s, decreased to less than 100,000 m³ per annum in the early 1970s, and increased very rapidly betwen 1972 and 1979. Goswami (1985) and Rogers (personal communication, 1984) have interpreted the very high sediment loads of the 1950s as the consequence of frequent earthquakes in Assam between 1951 and 1956. Rogers maintains that the level of annual floods along the Brahmaputra has actually decreased since 1975 at an annual rate of about 15 cm. He ascribes this to the river having completed adjustment to earthquake-induced disturbance of its channel (including a throw of about 4 m). The reason for the steep increase in both annual streamflow and sediment load after 1979, however, is not known. It seems that earthquake frequency has not increased again (Goswami, 1985) and we dispute the assumption that deforestation is the cause.

The very rapid increase in suspended load of the Brahmaputra after 197677 (more than 300 percent) parallels the streamflow curve. Thus both increases, streamflow and sediment load, would have to be interpreted as resulting from the same causes - that is, deforestation or other human activitity. This is not a realistic conclusion. For instance, the high-flow hydrograph shows a decrease while the low-flow hydrograph shows an increase during the late 1970s. If human influence is to be considered, theoretically this would be demonstrated by a reduction in the water retention capacity of the watershed and, therefore, a reduction in low-flow and an increase in high-flow river discharge. This argument is based upon the standard predictions of the Theory of Himalayan Environmental Degradation. Thus, if increased human pressures on the landscape are resulting in land degradation - through accelerated deforestation, poorly maintained agricultural terraces, soil erosion, and landsliding for instance the anticipated hydrological responses of these processes would be increased streamflow during the summer monsoon (that is, an increase in highflow discharge) and reduced availability of water during the subsequent dry winter and spring (that is, reduced low-flow discharge). On this basis, and in conjunction with the Brahmaputra streamflow data that are available, we must exclude changes in land use as a significant factor for explaining the streamflow and sediment load variations.

Having discussed the available data on the annual flow and sediment load of the Brahmaputra, let us now turn briefly to the Ganges and (Sapta) Kosi (Figures 6.11 and 6.12). We are including the (Sapta) Kosi in the macro-scale class along with the Brahmaputra, Ganges, and Indus, because of the large size of its watershed (30,000 km², and its international position (it flows through China, Nepal, and India); as a tributary of the Ganges, of course, its watershed is much smaller than those of the other three. The (Sapta) Kosi is regarded as an extremely problematic river in terms of land use in Bihar State, and the attempts to control the river rank high amongst Indian hydraulic engineering projects in terms of total expenditures and duration of the work.

Figure 6.11 (Sapta) Kosi at Tribeni: annual runoff 1948-76 (Tribeni gauging station). Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Zollinger, 1979).

Figure 6.12 Ganges annual runoff, 1934-74 (Hardinge Bridge gauging station). Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Indian Meteorological Department, 1971; Rieger, 1975).

Figures 6.11 and 6.12 respectively show the annual streamflow of the (Sapta) Kosi at Tribeni (1946-76) and of the Ganges at Hardinge Bridge, about a hundred kilometres from the Bangladesh frontier (1934-74: 1963-70 missing), together with the 10-year running means. Both hydrographs display very high annual variability, especially that of the Ganges. The 10-year running means also demonstrate an increase in streamflow, for the (Sapta) Kosi from about 1954 to 1976, and for the Ganges from about 1940 onward.

The (Sapta) Kosi hydrograph is particularly interesting because of the widely acclaimed degradation of its upper watershed. While the information is qualitative, it is generally accepted that the Tamur watershed and the lower Arun valley (both main tributaries of the (Sapta) Kosi) are locally heavily degraded.

Whether or not the increased flow in the (Sapta) Kosi was caused by the increase in watershed degradation cannot be evaluated. This is because much of the upper course of the river and its main tributaries are located in Nepal and China (Tibet) and no long-term precipitation data are available. In addition, as we have seen in Chapter 4, the Dudh Kosi and Sun Kosi (also tributaries of the (Sapta) Kosi) are subject to periodic catastrophic outbursts of morainedammed and ice-dammed lakes. These natural events not only temporarily augment streamflow but add vast amounts of sediment to the river channels. Similarly, the large number of glacial lakes on the Chinese side of the Arun watershed would suggest that the Arun is also subject to such catastrophic and abnormal disturbances. Similarly, as has been demonstrated in the glaciological literature (for example, Østrem, 1974; Young, 1985) for watersheds with a significant proportion of their area under permanent snow and ice, the twentiethcentury climatic warming has accounted for a significant increase in streamflow by accelerating glacier and high altitude snow melt. While no adequate supporting data are available for the entire (Sapta) Kosi watershed (but see Ikegami and Inoue, 1978; Fushimi and Ohata, 1980), that the glaciers in the upper reaches have been thinning and retreating since about AD 1900 is well known (Mayewski et al., 1980). Nevertheless, we must conclude that a proportion of the increase in (Sapta) Kosi discharge may be due to human interference in the watershed. However, long-term changes in precipitation patterns and rates of snow and ice ablation could account for most, if not all, of the increase. Despite this it has been claimed that the (Sapta) Kosi is responsible for a massive increase in sedimentation and that the upper watershed is contributing 172 million tonnes/yr to the formation of islands in the Bay of Bengal (Fleming, 1978). Rogers (personal communication, 1984), however, counters that the Kosi barrage, close to the Nepalese border, effectively checks much of the downstream sediment transfer from the mountains onto the plain; the major problem below the barrage appears to be one of entrenching of the river which has been relieved of much of its sediment load.

Figure 6.13 is a sketch map of the (Sapta) Kosi alluvial fan and shows the dynamic nature of the river's distributaries over the past 250 years. This indicates a 100-kilometre westward shift, and twelve distinct mainstream channels, during this period. This should be ample evidence to support the contention that the (Sapta) Kosi has been depositing vast amounts of sediment on its fan for a much longer period than that of recent (post-1950) human watershed intervention. This lends further weight to our supposition that the ratio of human (accelerated) erosion to natural erosion in the (Sapta) Kosi watershed could be very low indeed.

Figure 6.13 Channels of the (Sapta) Kosi River over the past 250 years showing the progressive westward shift of the river across northern Bihar State, India (after Carson, 1985).

Figure 6.14 Actual and filtered rainfall for India from 1866 to 1970. Ten-year running means and periods of higher than average rainfall are indicated. Prepared by Andreas Lauterburg, Geographical Institute, University of Berne (data from Parthasarathy and Mooley, 1978).

Figure 6.12 shows that most of the increase in streamflow Ganges (1940-74) is due to a decrease in the number of years with low discharge rather than the increase in high discharge years. The reduction in the number of low discharge years is certainly not due to anthropogenic impacts in the mountains, but must be the result of long-term precipitation fluctuations, with allowance for the impact of human intervention on the plains. To test this contention we have introduced Figure 6.14 which is an amalgamation of all the available monsoonal rainfall data for India for the period from 1866 to 1970 (Parthasarathy and Mooley, 1978). This long homogeneous series of rainfall data displays a remarkably high annual variability. Superimposed upon this, however, are the equally remarkable short-term fluctuations, particularly apparent between 1918 and 1970. This section of the curve can be broken into periods of ten to thirteen years. There are four periods each of about ten years' duration with higher rainfall totals than the long-term average. These are interrupted by periods of one to four years which are significantly drier than the long-term average (Parthasarathy and Mooley, 1978).

A comparison of Figures 6.12 and 6.14 shows a partial correspondence between variations in the streamflow of the Ganges and the aggregated annual monsoonal precipitation amounts. Lack of a strong correspondence may be due to the fact that the precipitation curve, constructed by Parthasarathy and Mooley (1978), is an integration of all Indian monsoon rainfall data. Despite this qualification, comparison of the two curves does suggest that human influence on changes in the Ganges discharge during the period 1944-72 is not demonstrable. It is emphasized, therefore, that streamflow curves of Himalayan rivers, especially meso-scale and macro-scale rivers, for periods of less than about twenty years should not be interpreted without taking rainfall data into account. Lauterburg's study failed to locate any macro-scale river that showed an increasing streamflow tendency over a long-term period. Nevertheless, there are short-term periods with increasing amounts or other features, such as variations in the high- and low-flow hydrographs, which give the overall impression that natural factors have dominated streamflow regime up to the present. From this it follows that, on the macro-scale, sedimentation rates are also dominated by natural processes.

The arguments presented above on variations in streamflow and precipitation of four major rivers of the Himalayan region do not lead to the inference that damage and loss of life from flooding has not increased during the course of the past fifty years or so. The scale and tragedy of these losses have been well documented. The large-scale flooding on the plains of Bihar, West Bengal, and Bangladesh, however, are most likely the result of the increasing number of human beings and livestock and the increasing intensity of agriculture in these areas in recent decades. Furthermore, monetary damage figures are often presented that have not allowed for inflation; thus there is the appearance of increase in damage even if the actual damage value had been the same in constant rupees or dollars. Nevertheless, the progressively increasing scale of loss is obvious, yet it cannot be demonstrated that this is due to human intervention in the mountain watersheds. Because long-term data are not available we are compelled, therefore, to draw the following conclusions:

1. The information network has improved rapidly since about 1940 and has promoted a much fuller awareness of the basically natural phenomenon of large-scale flooding the lower Ganges and Brahmaputra.

2. The rapidly increasing population of northeastern India and Bangladesh has led to both intensification of agriculture and its extension into areas that probably were always affected by flooding. However, these areas could not be used for permanent agriculture until recently, or were not required when population pressure was much lower.

3. Hydro-technical modifications to the main river channels of the North Indian Plain (including causeways, barrages, and canals, for irrigation purposes, hydroelectric power plants, and spillways for diversions out of the macrowatershed of the Ganges) may themselves play an important role in changing the water-sediment ratio. Consequently this would affect the downstream sediment-carrying capacity below major infrastructures, and therefore may lead to local increases in the level of the river bed and to outbreaks of water from the main channels.

4. The fact that water is withdrawn from the river for irrigation purposes, which will also result in overloading the rivers with sediment further downstream, has not been taken into account, again because of the difficulty of obtaining adequate data. Nevertheless, responsible Indian ministries (personal communication to Lauterburg, 1985) are well aware of this kind of problem.

The overall message of this chapter, therefore, is that the fluctuations in annual streamflow and high sediment loads of the macro-scale rivers are the consequences of natural (especially climatic) processes. Locally, flooding and excessive sedimentation on the plains may be due to human intervention on the plains, rather than to the activities of subsistence farmers in the mountains. As attention is focused on progressively smaller watersheds, down to a few hectares, the potential for human impact undoubtedly increases. Even at the micro-scale, however, periodic catastrophic rainfalls and Ethology, or specialized processes such as the outburst of glacier lakes, may heavily outweigh in importance the negative impacts of human activities.

We do not contend that soil-conservation practices are either unnecessary or not effective for the specific purposes for which they are undertaken. This is an entirely separate issue. It should not be confused by the macro-scale claims that a few million subsistence hill farmers are undermining the life support of several hundred million people on the plains (WRI, 1985). It follows that forestation of mountain watersheds, and extensive soil-conservation measures, are valuable for their own sake and, if appropriately carried out, are vital for the well-being of the hill farmer. It is potentially disastrous, however, for foreignaid agencies and national government authorities to undertake such activities with the conviction that they will solve problems on the plains. Incorrect identification of causes of large-scale problems and attempts to treat them would appear to constitute a major factor in the steepening of the perceived downward spiral into environmental and socio-economic supercrisis. At the very least, considering the enormous costs and energy requirements needed for forestation of large areas in the mountains, if such were undertaken with the major objective of modifying conditions on the plains, an expensive disappointment is a likely result.

Contents - Previous - Next