Reconstruction of a glacial lake outburst flood (GLOF) in the Engaño Valley, Chilean Patagonia: Lessons for GLOF risk management
Introduction
Moraine-dammed lake failures have been documented in glaciated areas worldwide. GLOFs can release millions of cubic metres of water in a short time (minutes to hours) generating deep, high-velocity flows with significant erosive and transport capacity (Costa and Schuster, 1988, Breien et al., 2008). Thus, GLOFs can pose a severe hazard to mountain communities. The damaging capacity of GLOFs was dramatically demonstrated in the Peruvian Andes in 1941, when the Palcacocha Lake failed flooding the city of Huaraz killing ~ 6000 inhabitants (Lliboutry et al., 1977). GLOFs can be triggered by the impact of mass movements (Harrison et al., 2006) or ice avalanches (Vuichard and Zimmerman., 1987) into the lakes, by waves generated by calving or the floating of dead ice (Richardson and Reynolds, 2000), or by overtopping as a consequence of intense or prolonged precipitation or increased ice/snow melting (Korup and Tweed, 2007). Floods from upstream lake failures can also cause outburst floods (Lliboutry et al., 1977).
The number and size of moraine-dammed lakes has increased worldwide in the last 40 years as a consequence of glacier retreat. This tendency has been observed for example in the Himalayas (Gardelle et al., 2011), tropical Andes (Ames, 1998) and Patagonia (Loriaux and Casassa, 2013). In Patagonia, at least 16 moraine-dammed lakes have failed in historic time (Iribarren Anacona et al., 2015) and one of these events is one of the largest GLOFs, in terms of flood volume, reported worldwide (Hauser, 2000, Clague and Evans, 2000). However, GLOFs in Patagonia have affected mostly uninhabited valleys and thus have been underreported. The study of past GLOFs can shed light on flow dynamics and aid in anticipating future GLOF behaviour (e.g. through the evaluation of numerical models or the development of empirical relationships). Hence, knowledge about past floods has both scientific and societal value (Baker et al., 2002). The study of past GLOF socioeconomic consequences may also aid in unravelling planning and political issues affecting GLOF risk management (Carey et al., 2012).
Several studies have reconstructed GLOFs using stratigraphic and geomorphic evidence, numerical models or gauging data (e.g. Kershaw et al., 2005, Schneider et al., 2014). However, few studies have included eye-witnesses accounts, and when included, people's experiences are only recorded to provide data about flood timing, extension or damage. Memories of GLOF events can also shed light on the perception of GLOF risk that ultimately may affect the community response during GLOFs (Gyawali and Dixit, 1997, Carey et al., 2012). They also can inform about territorial planning practices and issues. Our main objective is to reconstruct the GLOF that affected Bahía Murta Village in March 1977 in order to provide insights into the GLOF dynamics. We also analyse the human response to the GLOF, and planning problems that led to the village's flooding and ultimate relocation. This information is used to delineate a GLOF risk management strategy. We suggest that an outburst susceptibility assessment and territorial planning could have prevented the GLOF damage.
Bahía Murta is located at the north shore of the General Carrera Lake in Chilean Patagonia (Fig. 1). The flooded village (known as “Bahía Murta Viejo”) is located on a delta formed between the Engaño and Murta river outlets, about 26 km from the failed moraine-dammed lake. The village was settled from the 1930s onwards, supported by Chilean colonisation laws. Public facilities (i.e. school and a chapel) were built in the 1950s, and by 1977 Bahía Murta Viejo had about 130 inhabitants (El Diario de Aysén, 1977a).
The Engaño Valley is flanked by the Andean mountain range with higher peaks in the Engaño Basin reaching 1950 m.a.s.l. The highest areas are generally covered by glaciers which descend to the lowest altitude of 750 m.a.s.l in the Engaño Basin. Glaciers in Patagonia generally have steep mass-balance gradients, are fast flowing and have high ablation rates (López et al., 2010). Most of the glaciers in Patagonia are receding (Davies and Glasser., 2012) as a consequence of twentieth century warming (Rosenblüth et al., 1995) as well as decreasing precipitation in the last few decades in North-Central Patagonia (Garreaud et al., 2013).
Annual precipitation in the Engaño Basin varies from ~ 1400 (at 200 m.a.s.l) to more than 2000 mm in the highest parts of the catchment. There are no gauging data of the Engaño River, however, the mean river discharge is probably less than 50 m3/s, with peaks in December and January associated with increased snow and ice melting. The Engaño valley has a low gradient along most of its length, however, the valley narrows and steepens in two bedrock gorges. The floodplain, as well as the valley sides, is forested. However, since the 1950s, large patches of forest have been cleared for pasturing or wood extraction.
The Engaño GLOF was reconstructed by a suite of sources including eye-witnesses accounts, newspaper reports, interpretation of aerial photographs and satellite images, and numerical modelling. Field analyses (e.g. identification of water marks or the description of GLOF deposits) were not undertaken due to difficult access, and because vegetation growth has concealed evidence of the GLOF. The data and procedures followed to reconstruct the GLOF dynamics are described below.
In January 2014 we held semi-structured interviews with twelve Bahía Murta inhabitants who recalled their memories of the flood. Commonly, data about the time, duration and flood stage were given. In other cases, we asked the interviewees to recall these and other data including: planning actions, evacuation procedures and information about the village's history, including past floods and settlement. The interviews were conducted in the people's homes or in the street and were audio recorded. The first interviewees helped to locate eyewitnesses of the 1977 GLOF. Three interviewees were in Bahía Murta Viejo on the day of the flood, three were living in the Engaño Valley about 16 km from the lake, and two were in Bahía Murta Nuevo. The rest of the interviewees were Bahía Murta Viejo inhabitants whose houses or land were damaged by the flood, but were elsewhere when the GLOF occurred. Ten interviewees were adults and two were children in 1977.
The geomorphic and topographic setting of the Engaño Lake and its surroundings prior to the GLOF were analysed from aerial photographs of 1955 and a topographical map of 1975. The GLOF extension and its geomorphic effects were examined in a Landsat MSS image from February 1979 (60 m spatial resolution) and Google Earth Images. The analysis allowed identifying GLOF pre-conditioning and triggering factors (e.g. potential source of ice avalanches or mass movements) as well as describing the GLOF path (patches of stripped vegetation or sediment deposition).
The Río Engaño GLOF was reconstructed (flow extension, arrival time, depth and velocity) using the 2D capabilities of HEC-RAS 5.0 Beta which solves the Full 2D Saint Venant equations or the 2D Diffusion Wave equations using an implicit finite volume algorithm. A description of the model can be found in (HEC RAS 5 Beta Manual). The model set up included the definition of upstream and downstream boundary conditions, the creation of a grid with elevation data, the selection of Manning roughness values, and the definition of the model spatial domain.
The GLOF was modelled as an unsteady flow using simulated dam-breach hydrographs as an upstream boundary condition and the normal depth or energy slope (i.e. 0.001 m/m) as a downstream boundary condition. A base flow of 30 m3/s was run for 72 h to pre-wet the Engaño River channel until the water reached a steady state along the entire spatial domain. Manning coefficients for the channel (0.04 and 0.05) and floodplain (0.08 and 0.1) were derived by comparing field photographs with roughness values in the literature (Arcement and Schneider, 1989). The analysis of field photographs was complemented with the interpretation of satellite images. These Manning values represent a realistic uncertainty range according to field observations. Only one Manning value was assigned to the floodplain and the channel in each simulation to simplify the model parameterisation.
Floods from failed dams commonly overtop river channels inundating floodplains and behaving as truly two-dimensional flows. Thus, 1D hydraulic models can be inappropriate for modelling such events (Hromadka et al., 1985). 2D models are well suited for reconstructing the complex hydraulics of high-magnitude flows such as simultaneous channelised and sheet flows and divergent flow around obstructions (Carrivick, 2006). Furthermore, 2D models usually have a better performance than 1D models in wide floodplains (Horritt and Bates, 2002) where detailed river cross sections are lacking. Thus a 2D model is warranted for simulating the Engaño Valley GLOF.
We used the 2D Diffusion Wave equations to simulate the Bahía Murta GLOF. HEC-RAS is more stable and requires less computational effort in solving the 2D Diffusion Wave than the Full 2D Saint Venant equations. Models that use the 2D Diffusion Wave equations may fail to reproduce local phenomena (e.g. run-up and bores) due to the omission of the inertial terms in the Saint Venant equations, however, they have previously proven adequate in simulating the overall dynamics of dam break floods over complex topography (Hromadka et al., 1985, Prestininzi, 2008).
The SRTM v4 Digital Elevation Model (DEM) (3 arc sec of spatial resolution equivalent to 80 m in the study area Rabus et al., 2003) and the ASTER GDEM2 (approximately 30 m of spatial resolution Tachikawa et al., 2011) were used to model the flood. An inspection of the DEMs revealed that sinks > 10 m and counterslopes exist in narrow reaches and gorges within the Engaño River channel. These features are likely the result of sensor difficulties in capturing the channel topography, when the channel width was narrower than the DEM resolution. Sinks and counterslopes can modify water flow behaviour, reducing the accuracy of modelling results (Zhu et al., 2013). Thus, we applied the approach of Tarboton et al. (1991) to “hydrologically correct” the DEMs by filling the DEM sinks. This prevents, for example, the accumulation of water in spurious pixels which artificially delay the flood progression.
As detailed topographic and geotechnical information is not available for the dam, and bathymetric data from Engaño Lake are lacking, we estimated the dam breach hydrograph and the breaching time using empirical formulae. The outburst volume was previously estimated to be ~ 7.3 × 106m3 using an empirical formula relating lake area and volume (Iribarren Anacona et al., 2014). Several simulations were run increasing and decreasing this reference value up to ~ 70%, which is the error margin of the formula.
The breaching time was estimated with Froehlich's (1995) equation which is based on data from 63 dam breaches and has a smaller uncertainty range than other empirical approaches (Wahl, 2004)
Here T corresponds to the breach time in minutes, Vw is the volume of water above the breach invert (and equal to the outburst volume) and hb is the height of the breach. The height of the breach was measured using the ASTER GDEM2 data. We assumed that the lake emptied by overtopping and the progressive erosion of the dam. We assumed as well that the GLOF had a triangular shaped hydrograph, lasting for the time calculated with the Froehlich (1995) formula. Resulting hydrographs have peak discharges surpassing 10,000 m3/s (Fig. 2). Although this value could be an overestimation, flow discharges tend to converge downstream, and the channel slope and flood volume dominate the flood behaviour (Ponce et al., 2003). Thus floods can still be simulated realistically in spite of the uncertainty of the peak discharge at the dam breach.
Several flood simulations were undertaken to obtain an optimal set of parameters to model the Engaño River GLOF. Trials were performed using the SRTM and ASTER GDEM2 DEMs and different hydrographs and Manning values. Manning values ranging from 0.04 to 0.1 were tested since these values represent the range of roughness values of the Engaño channel and floodplain observed in the field. A one second time step was used in each simulation and the downstream boundary condition (i.e. energy slope) was held constant. The simulation results were contrasted from eye-witness accounts and the interpretation of satellite images.
Section snippets
GLOF preconditioning and potential triggering factors
The characteristics of the Engaño Lake and its surroundings made it highly susceptible to outburst floods. The Engaño Lake had a surface area of 1.15 km2 two years before the 1977 GLOF. However, the lake was considerably smaller in the 1950s when a glacier occupied the lake basin. The glacier retreated about 1.5 km between 1955 and 1976 increasing the lake area (Fig. 3). The Engaño Lake is surrounded by steep bedrock and ice-covered slopes prone to mass movement and ice/snow avalanching. The lake
Lessons for GLOF risk management
The interviewees highlighted GLOF singularities, as well as technical and planning issues, that contributed to flood damage. Although this data is mostly anecdotal, it helps to point out measures that should be taken to prevent or mitigate negative outcomes in GLOF management.
Conclusions
We reconstructed the 1977 Engaño River GLOF, in the Chilean Patagonia, by means of eyewitness accounts, newspaper reports and a 2D hydraulic model. We show that the HEC RAS 5 Beta model, which is freely available, can realistically simulate GLOF dynamics using medium resolution DEMs. Thus, this sofware can be used in GLOF hazard assessments in similar geographic contexts. Furthermore, we illustrate that small-scale, short-distance migration can be a feasible approach to cope with GLOF hazards.
Acknowledgments
We thank Bahía Murta inhabitants for sharing their memories of the 1977 GLOF and the village's history. We also thank Rodrigo Iribarren for helping carrying out the interviews. Comments of anonymous referees improved the manuscript. This work was approved by the Human Ethics Committee of Victoria University of Wellington.
References (62)
- et al.
Calving processes and the dynamics of calving glaciers
Earth Sci. Rev.
(2007) Application of 2D hydrodynamic modelling to high-magnitude outburst floods: an example from Kverkfjöll, Iceland
J. Hydrol.
(2006)- et al.
A review of catastrophic drainage of moraine-dammed lakes in British Columbia
Quat. Sci. Rev.
(2000) - et al.
Contrasted evolution of glacial lake along the Hindu Kush Himalaya mountain range between 1990 and 2009
Glob. Planet. Chang.
(2011) - et al.
Evaluation of 1D and 2D numerical models for predicting river flood inundation
J. Hydrol.
(2002) - et al.
A two-dimensional dam-break flood-plain model
Adv. Water Resour.
(1985) - et al.
Ice, moraine and landslide dams in mountainous terrain
Quat. Sci. Rev.
(2007) - et al.
A regional view of fluctuations in glacier length in southern South America
Glob. Planet. Chang.
(2010) - et al.
Evolution of glacial lakes from the Northern Patagonian Icefield and terrestrial water storage in a sea-level rise context
Glob. Planet. Chang.
(2013) Suitability of the diffusive model for dam break simulation: application to a CADAM experiment
J. Hydrol.
(2008)
The shuttle radar topography mission—a new class of digital elevation models acquired by spaceborne radar
J. Photogramm. Remote Sens.
An overview of glacial hazards in the Himalayas
Quat. Int.
A documentation of glacier tongue variations and lake development in the Cordillera Blanca, Peru
Z. Gletscherk. Glazialgeol.
Guide for selecting Manning's roughness coefficients for natural channels and flood plains
U.S. Geol. Surv. Water Supply Pap.
The scientific and societal value of paleoflood hydrology
Erosion and morphology of a debris flow caused by a glacial lake outburst flood, Western Norway
Landslides
An integrated socio-environmental framework for glacial hazard management and climate change adaptation: lessons from Lake 513, Cordillera Blanca, Peru
Clim. Chang.
Flow hydraulics and geomorphic effects of glacial-lake outburst floods in the Mount Everest region, Nepal
Earth Surf. Process. Landf.
The formation and failure of natural dams
Geol. Soc. Am. Bull.
Accelerating shrinkage of Patagonian glaciers from the Little Ice Age (~ AD 1870) to the present day
J. Glaciol.
Calving processes and lake evolution at Miage Glacier (Mont Blanc, Italian Alps)
Ann. Glaciol.
Efectiva ayuda a damnificados
Auxilian a Pobladores en Murta
Fue arrasado Pueblo Viejo en General Carrera
Profesionales viajan llevando auxilio a pobladores damnificados
Listo proyecto Centinela que permitirá alertar a pobladores vaciamiento del lago Cachet 2
Flooding and the scale of migration
Nat. Clim. Chang.
Embankment dam breach parameters revisited
Large-scale control on the Patagonian Climate
J. Clim.
Geomorphometric comparison of ASTER GDEM and SRTM
How distant is Nepali science from Nepali society? Lessons from the 1997 Tsho Rolpa GLOF panic
Water Nepal
Cited by (57)
A glacial lake outburst floods hazard assessment in the Patagonian Andes combining inventory data and case-studies
2024, Science of the Total EnvironmentCombining geophysics, remote sensing and numerical simulation to assess GLOFs: Case study of the Namulacuo Lake in the Southeastern Tibetan Plateau
2023, Science of the Total EnvironmentAssessment of potential present and future glacial lake outburst flood hazard in the Hunza valley: A case study of Shisper and Mochowar glacier
2023, Science of the Total Environment