Snow and Glacier Hydrology (Proceedings of the Kathmandu Symposium, November 1992). IAHSPubl. no. 218,1993. 331 Estimation of Glacier Lake Outburst Flood and its Impact on a Hydro Project in Nepal G. MEON & W. SCHWAHZ Hydropower Projects Division, Lahmeyer International, Consulting Engineers, Lyoner Strasse 22, D-60496 Frankfurt/Main, Germany Abstract Glacier lake dams consisting of unconsolidated material are prone to failure and may cause disastrous surges of water heavily charged with debris. Consideration of potential glacier lake outburst floods (GLOF) is, therefore, of essential importance for the design of river engineering structures located downstream of hazardous glacier lakes. As an example, a flood study carried out for a hydropower project on the Arun River in Nepal is taken. The results indicate, that, in addition to the "classical" flood flow analysis, GLOF analysis should be considered for derivation of design floods of projects located in glacier dominated watersheds. INTRODUCTION Sudden, large river flow caused by an outburst of a glacier lake is generally termed glacier lake outburst flood or GLOF. The outburst may be caused by the failure of the damming moraine due to its own instability or glacier and/or snow collapse into the lake and may lead to overtopping and eventually to failure of the damming barrier. Nearly a quarter of the Nepalese territory is covered by snow, mainly along its Northern boundary. There are numerous glaciers and glacier lakes in that part of the country. Virtually all of the perennial rivers of Nepal are snow fed and many of them originate at the glacier lakes either in Nepal or in Tibet. As such, many rivers, particularly the larger ones, are associated with GLOF. In fact, a number of GLOFs have caused serious damage in Nepal. Among the most publicized were Zhangzanbo Lake outburst of July 11, 1981, and Dig Cho Lake outburst of 4 August 1985. The former destroyed the Friendship Highway Bridge on the China-Nepal highway and the intake dam of Sunkosi hydroelectric station. The latter was caused by the burst of the terminal moraine of the Ding Cho Lake. It led to a surge in the Dudh Kosi River of approximately 2500 m 3 s" 1 that washed away the nearly completed Namche Hydel Plant. In the Pum Qu (Arun) basin, at least five glacier lakes have burst since 1940. However, only one of them, i.e. Gelhaipuco Lake outburst of 21 September 1964, has been recorded as having caused damage in Nepal.
332 G. Meon & W. Schwarz Consideration of potential glacier lake outburst floods is, therefore, of essential importance for the design of river engineering structures located downstream of hazardous glacier lakes. As an example for a GLOF estimation, a flood study carried out for a hydropower project at the Upper Arun River in Nepal was taken (Morrison- Knudsen (MK), Lahmeyer International (LI), and others, 1991). The Arun River originates from a glacier on the North slope of Mt. Xixabangma Feng in the Southern part of Tibetan highland. The Tibetan part of the catchment accounts for about 98% of the total watershed area upstream of the planned dam site. The study comprised an inventory of glacier lakes in Tibet, a selection of the most critical lake(s) for GLOF analysis, a simulation of the progressive breach of the damming lake barriers and of the outflow hydrograph, and an assessment of the attenuation of the GLOF up to the project site. Comparison of the GLOF with maximized synthetic floods caused by storm rainfall confirmed the necessity to integrate GLOF analysis into the procedure of deriving design floods for river engineering structures. GLACIER LAKES UPSTREAM OF UPPER ARUN PROJECT SITE Most of the Tibetan part of the Arun River (Pum Qu) basin is formed by highland above elevation 4500 and is surrounded by high mountains. The geomorphology of the Tibetan basin is characterized by glacial or periglacial landforms. It covers an area of about 25 300 km 2. A minor part of the Tibetan basin, close to the Southern border, is topographically similar to the mountainous drainage basin on the South side of the Himalayan range in Nepal. The Nepalese drainage area up to the dam site is about 400 km 2. The average bed slope is about 0.8% in the upper catchment of Tibet, and increases drastically to 3-4% towards the Tibetan-Nepalese border, and in Nepal. According to an investigation carried out by the Sino-Nepalese Joint Expedition Team (CAS, WECS, and NEA, 1988) a total of 229 glacier lakes exists in the Tibetan Arun Basin. The accumulated area is about 46.750 km 2, and the water reserve is about 1.2 x 10 9 m 3. The major lakes and the Upper Arun dam site are indicated in Fig. 1. In Fig. 1, the glacier area has been divided into nine regions <A> to <I> located in Tibet and eight regions (A) to (H) located in Nepal. Table 1 summarizes the glacier lake types occurring in the Arun River basin. SELECTION OF CANDIDATE LAKES FOR GLOF ANALYSIS The following criteria have been adopted for the selection of a candidate lake for the GLOF analysis:
Estimation of glacier outburst flood and a hydroproject in Nepal 333
334 G. Meon & W. Schwarz Table 1 Glacier lake types of the Arun (Pum Qu) River basin. Type Number No. % Area Area (km 2 ) % Water reserves Water % storage (km 3 ) Mean area (km 2 ) Potentially dangerous No. Area (km 2 ) Cirque lake End moraine dammed lake Trough valley lake Blocking lake Total 79 109 40 1 229 34.5 47.5 17.5 0.4 100 6.244 27.500 11.774 1.228 46.746 13.4 58.8 25.2 2.6 100 0.0624 0.8250 0.2944 0.0491 1.2309 5.1 67.0 23.9 4.0 100 0.079 0.252 0.294 1.228 0.204 24 13.331 The lake should have an actual potential for outburst during the project life. The lake should have the highest potential of damage to the project. In assessing the outburst potential of glacier lakes the following points were considered: - Whether the lake is end moraine-dammed, and whether the end moraine is formed before the Little Ice Age or later. Whether the lake is situated near the glacier that might fail and fall into the lake. - Whether there is a potential of a glacier outburst related to: (a) moraine dams; such as size, shape and side slopes of moraine dams, their arrangement, i.e. single or multiple moraine dams; (b) glaciers; such as shape, size and slope of the advancing glaciers, particularly of their tongues; (c) glacier lakes; such as bed slope, upstream and downstream slopes. In assessing the damage potential of GLOF the main factors considered were: glacier or lake water storage; - the average bed slope of the river carrying the GLOF flow; - the distance of the lake from the proposed dam site. In terms of the Upper Arun Project, the subbasins of Ganma Zangbo River < A> and Natang Qu River <I> are of primary concern, since each of them is located close to the project site. In addition to the information provided by the Sino-Nepalese expedition, glacier lakes were identified by SPOT analyses. The SPOT imagery was particularly focused on Gamma Zangbo and Natang Qu. No large lakes were found in Gamma Zangbo. Several small glacier lakes located in the Nepalese portion of the basin were considered to be stable. Their damming structures consist of bed rock, and their glaciers have receded far backup to the ridge. In Natang sub-basin, which is located closest to the project site within the Tibetan catchment, there are three major glacier lakes; Zonggya Co,
Estimation of glacier outburst flood and a hydropreject in Nepal 335 Abmachimai Co and Quangzonk Co. Among these, the latter two lakes are likely to constitute the highest risk. The first lake is considered stable, because of its distance from the back glacier. It would not undergo abrupt disturbance by ice mass falling into the lake. The latter two lakes have end moraine barriers and valley glaciers on their hillsides. The end moraine dam of the Abmachimai Co was formed before the Little Ice Age, whereas the Quangzonk Co Lake was formed during the Little Ice Age and was, therefore, assumed to be less consolidated. Quangzonk Co Lake was judged more dangerous and selected for simulation finally, because of such field findings as floating ice masses and rapid development of the lake, which indicates that the back glacier is active and may raise the water level in a relatively short period. Moreover, being the largest lake in volume, the Quangzonk Co was expected to have the greatest possible magnitude of GLOF. The dimensions of the latter two lakes are listed in Table 2. Table 2 Dimensions of hazardous glacier lakes located in Natang River basin. Name of Lake Abmachimai Co Quangzonk Co Area (km 2 ) 0.565 0.753 W X L (m) 500 x 1525 475 X 2150 Depth (m) 71.5 67.5 Volume (10 6 m 3 ) 19.4 21.1 Elevation (m asl) 5200 5004 SIMULATION OF GLOF Data base and model The GLOF analysis was performed by simulating a progressive breach of the Quangzonk Co Lake moraine dam and routing thefloodthrough the Natang Qu and Arun River down to the project site. The breaching process of a glacier lake moraine dam caused by overtopping is little known. It can be assumed that the breaching characteristics are, to some extent, similar to those observed at historical earth dam failures (ICOLD, 1992). In all these cases, an instantaneous total or partial failure did not occur. The size, shape and time required for the breach development depend on the erodability of the embankment material and on the flow forming the breach. The breaching process can take place fairly rapidly, or it can last several hours. The size of the break is often significantly less than the entire dam (McDonald & Langridge-Monopolis, 1984). Thus, the breach itself strongly influences the peak and shape of the outflow. Numerical simulation of the erosion process is feasible under certain assumptions and simplifications. Existing numerical dam breach models can be grouped as follows: Parametric models like the breach component of Fread's (1984) well-
336 G. Meon & W. Schwarz known model DAMBRK: This model uses empirical breaching parameters derived from observations of previous dam failures. In DAMBRK, the given parameters "shape, time of breach formation and depth of final breach" determine the breach development. The velocity of widening of the breach is assumed to be constant. Combined parametric and physically-based models: these models use parametric approaches for the breach shape and physically-based approaches for the erosion and outflow processes. Their application to historical dam failures have been reviewed to be good (Ponce & Tsivoglou, 1981; Fread, 1985). Depending on the complexity of the models various soil data and other parameters are needed, which are not always available. Problems of numerical stability may arise if unsteady flow equations of the breach channel have to be solved in combination with sediment transport equations. Combined parametric and physically-based models, extended by stochastic components: These models reflect the latest developments in dam breach modelling. They allow quantification of the uncertainties of the breaching process and their influence on the outflow (Meon, 1989). Considering the data base and the practical relevance of the study, Fread's (1984) model DAMBRK was chosen for simulation of the breach and outflow as well as for the propagation of the flood down to the project site. The river flood routing component is based on the solution of the one-dimensional equations of continuity and energy conservation. Based on a literature study, on discussions with geologists, glaciologists and soil mechanics experts, and on various sensitivity runs, the following assumptions were taken as decisive for the GLOF: breach shape: base width = 50 m, breach height = 30 m, side slopes = 1:1; breach time: one hour, considering the fact that a natural moraine dam is less stable and resistant against breach than a constructed earth-fill dam. Computations As shown in Fig. 2, the river channel was represented by 16 cross-sections for both Natang and Aran rivers. Manning's roughness coefficients were empirically estimated and ranged between 0.04 and 0.10. Preliminary computations showed that the high variation of roughness coefficients and river bed slope (see Fig. 2) repeatedly led to fluctuations from subcritical to supercritical flow along the total reach. The program DAMBRK is, however, not capable to reliably simulate such mixed flow conditions. In order to obtain plausible upper and lower limits of results it was decided to perform two analyses, one for subcritical and the other one for supercritical flow along the total reach. The corresponding roughness coefficients were determined by trial and error procedures. The results are summarized in Table 3.
Estimation of glacier outburst flood and a hydroproject in Nepal 337 Elevation (1000 masl) 6 5.Quangzonk Co 4 3 2 1 0 0 10 20 30 40 50 60 70 80 Distance (River-km) Fig. 2 Natang-Arun Rivers: river bed profile. Main purpose of these GLOF computations was a conservative assessment of the maximum discharge. For this, the density of the water-sediment-bedload mixture of the GLOF was kept at 1.0. In reality, the density is assumed to increase significantly. If compared with the computations, the higher density of a "real" GLOF would lead to an increase of the friction losses and, consequently, to a decrease of the maximum discharge at the dam site and an increase of the time to the maximum discharge. Figure 3 shows the lake outflow hydrograph and the hydrograph routed to the dam site, both for supercritical flow conditions under consideration of lateral inflow from tributaries. As the slope of the Natang Qu is often very steep, the flow will be supercritical along those steep reaches. Since the Upper Arun Project is located some 15 km downstream of the junction of the Natang Qu and Arun rivers the maximum discharge at Upper Arun dam site is expected to be closer to the supercritical flow than to the subcritical flow discharge. Considering the fact that the GLOF viscosity is somewhat higher than that of the clear water assumed in DAMBRK program, the maximum GLOF discharge adopted for the dam site was finally assumed to be in the range of 6300 m 3 s" 1. Table 3 Maximum discharge caused by GLOF, with At as computation time step. Discharge (m 3 s" 1 ) Location Lake Quangzonko Co Project Dam site Project Powerhouse Site River-km 0.0 57.7 77.3 Subcritical flow At = 45 s 6300 5400 4900 Supercritical flow At = 90 s 6500 6900 6800
338 G. Meon & W. Schwarz Flow (1000 cms) 8 7 6 - - -/ \- \- - U.A. Damsite km 57.7 1 1 Glacier Lake km 00.0 5 4 3 2 1 X^-Lateral inflows included ^ ^ ^ ^ 0 i i i i 0 0.5 1 1.5 2 2.5 3 3.5 4 Time (tirs) Fig. 3 Natang-Arun Rivers: glacier lake outburst flood. CONCLUSIONS Comparison of the GLOF with synthetic floods caused by extreme rainfall (mainly) on the Nepalese portion of the Aran River basin is given in Table 4. For another project located some 25 km further downstream on the Arun River, the flood calculations provided a different range of maximum discharges: there, the PMF exceeded the GLOF because of the increased Nepalese portion of the catchment, which is exposed to extreme rainfall (Lahmeyer International et al., 1989). The maximum GLOF discharge at the Upper Arun dam site is about three times the spillway design flood (1000-year flood) selected for the concrete structure of the run-of-river dam. The dam will have a height of 37 m only, and be constructed such that it can be overtopped without being destroyed by a PMF or a GLOF discharge. Damage is likely to be caused by large boulders as parts of the GLOF bed load. Hence safety of the dam against damage will depend on a well-timed complete opening of the gates to lower the reservoir Table 4 Results of flood flow analyses for Upper Arun dam site (Morrison Knudsen Engineers et at, 1991). Type of Flood GLOF PMF 1000-year flood 100-year flood Maximum discharge (m 3 s" 1 ) - 6300-4400 ~ 2100 ~ 1800
Estimation of glacier outburst flood and a hydroproject in Nepal 339 water level before the GLOF arrives at the dam and to pass major parts of a GLOF and its bed load through the open gates. Installation of a warning system was recommended to be indispensable. As a minimum, an automatic water level monitoring and transmission system should be installed at the Nepalese- Tibetan border. All GLOF potential lakes should be regularly surveyed. Based on both studies (Morrison Knudsen Engineers et al., 1991; Lahmeyer International et al., 1989) it is concluded that, in addition to the "classical" flood flow analyses, GLOF analysis should be considered for derivation of design floods of projects in glacier dominated mountainous watersheds. REFERENCES CAS, WECS & NEA (1988) Report on First Expedition to Glaciers and Glacier Lakes in the Pumqu (Arun) and Poiqu (Bhote-Sun Kosi) River Basins, Xizang (Tibet), China. Science Press, Beijing. Fread, D. L. (1985) BREACH: An Erosion Model for Earthen Dam Failures. NWS Report, National Oceanic and Atmospheric Administration, Silver Spring, Maryland. Fread, D. L. (1984) DAMBRK: The NWS Dam-Break Flood Forecasting Model. Hydrologie Research Laboratory, National Weather Service, Silver Spring, Maryland. ICOLD (CIGB) (1992) Selection of design flood current methods. Bulletin 82, International Commission of Large Dams, Paris. Lahmeyer International, Energy Engineering International & Electric Power Development Company (1989) Arun III Hydro-Electric Project Detailed Engineering Services, Project Formulation Report 1, Vol. 1: Main Report, Nepal Electricity Authority. McDonald, T. C. & Langridge-Monopolis, J. (1984) Breaching characteristics of dam failures. ASCEJ. Hydraul. Engng 110(5). Meon, G. (1989) Safety analysis of a dam under flood load. Dissertation (in German), Heft 35, Institute for Hydrology and Water Resources Planning, University of Karlsruhe, Germany. Morrison Knudsen Engineers, Lahmeyer International, Tokyo Electric Power Service Co., Nepecon (1991) Upper Arun Hydroelectric Project, Feasibility Study-Phase II, Nepal Electricity Authority. Ponce, V. M. & Tsivoglou, A. J. (1981) Modelling gradual dam breaches. ASCE J. Hydraul. Div. 107(7).