Glacial Lake Outburst Floods Simulation in the Yi’ong Zangbo River Basin Based on One-Dimensional Hydrodynamic Model

Glacial lake outburst oods (GLOFs) are a serious potential threat to the safety of life and property in downstream areas. In this study, moraine-dammed glacial lakes in the Yi’ong Zangbo River basin were recognized based on Landsat ETM+/TM/OLI images in 2000 and 2019. And the GLOFs for the Jionglaco, the largest glacial lake in this basin, was simulated using the one-dimensional hydrodynamic model. The results show that the total number and area of moraine-dammed glacial lakes in this basin increased by 10 (10.52%) and 5.49 km 2 (48.24%) from 2000 to 2019, in which the area of the Jionglaco increased by 3.22 km 2 . The peak discharge at the breach outlet for ve scenarios with different combinations of breach width (80 m and 120 m), depth (2.5 m and 5 m) and ood time (1.5 h and 3 h) are 489 m 3 /s, 1327.43 m 3 /s, 444.32 m 3 /s, 617.47 m 3 /s and 1570.61 m 3 /s. With the addition of baseow in river, the peak discharge at bridge site 15 138.93 km from Jionglaco are 1040.89 m 3 /s, 1724.00 m 3 /s, 1024.85 m 3 /s, 1162.25 m 3 /s and 1990.52 m 3 /s. The combination of baseow in river and the GLOFs discharge results in the increasing peak discharge in the further downstream region. However, the arrival of peak discharge in downstream areas is delayed, which increases the chances of people escaping. This study aims to provide some references for the prevention of GLOFs in this region.


Introduction
). Some of glacial lakes are extremely dangerous because of the unstable geomorphology around them (Maskey et al. 2020). As the dams of these glacial lakes tend to be loose and unstable, the combination of external factors, such as the entry of ice avalanches and rock avalanches can lead to the collapse of dam and the release of water, forming glacial lake outburst oods (GLOFs). Consisting of a mixture of water and sediment, the GLOFs can travel at speeds more than tens of kilometers per hour for more than 100 km (Worni et al. 2014). Due to the high volume of emissions and long operational distances, lives and properties in downstream areas would generally suffer heavy losses (Maskey et al. 2020).
According to Zhang et al. (2015), there are a total of 5701 glacial lakes (> 0.003 km 2 ) in Qinghai-Tibet Plateau in 2015. Wang et al. (2020) synthetically identi ed 246 potential dangerous glacial lakes with a total area of 78.38 km 2 in the Qinghai-Tibet Plateau, in which glacial lakes with very high and high integrated risk of GLOFs disasters are concentrated on the central Himalayas, the central-eastern Nyainqentanglha range, and the southern Tanggula mountains. Both the Jionglaco and the Jiongpuco were included in these potential dangerous glacial lakes. The Jionglaco is a large glacial lake in the eastern part of the Nyainqentanglha range, located in the Yi'ong Zangbo River basin, where three historical GLOFs have occurred, which all caused signi cant damage to the downstream area (Yao et al. 2014; Sun et al. 2014;Liu et al. 2021). The Jionglaco is larger than all three glacial lakes that have been breached and has a rapid area expansion in recent years. The Jiongpuco is a moraine-dammed glacial lake in the southwest of the Jionglaco with area expansion in the past two decades. Moreover, there are many settlements, roads and bridges in the downstream area, so the GLOFs simulation for the Jionglaco is urgent, which can provide data support and theoretical basis for GLOFs prevention and mitigation in this region.

Study Area
Yi'ong Zangbo River basin (30°05'-31°03' N, 92°52'-95°19' E), situated in the southern aspect of the Nyainqêntanglha range, is fed by Yi'ong Zangbo River and covers an area of 13,533 km 2 ( Fig. 1(a)). The terrain is high in the west and low in the east, with an average altitude of over 4,000 m ( Fig. 1(b)). The average annual precipitation is 958 mm, and the peak precipitation occurs in the Indian monsoon season from May to September, which accounts for 74.9% of the annual total precipitation according to records from the nearest meteorological station in Bomi. The mean annual temperature is 8.8 ℃, with average temperatures of 16.7 ℃ in July and 0.28℃ in January (Ke et al. 2013(Ke et al. , 2014. The massive topographic landforms together with the Indian monsoons leads to the development of distinct temperate valley glaciers and a substantial number of glacial lakes (Fig. 1(c)). The Jionglaco (30˚39'44" N, 94˚29'01" E), located in the Jinling Township, Bianba County, which is a typical moraine-dammed glacial lake fed by the huge maritime glacier (Fig. 1(c)). The glacier melted dramatically, and the glacial lake expanded rapidly towards the glacier terminus with the area of 5.68 km 2 in 2020. Meanwhile, there are many villages and bridges located in the downstream area of the glacial lake ( Fig. 1(c)).

Glacier and glacial lake mapping
As the primary source of medium spatial resolution Earth observations, the Landsat series images is an important data source for studying glacier and glacial lake changes (Wood et al. 2008;Chander et al. 2009). In this study, a total of six scenes of Landsat TM/ETM+/OLI images with little snow and cloud cover downloaded from the United States Geological Survey (USGS) (http://glovis.usgs.gov/) were used to delineate the outlines of glacial lakes in 2000 and 2019 ( Table 1). The spatial resolution of Landsat TM is 30 m. However, Landsat ETM+/OLI images are available in multi-spectral band 30 m and panchromatic band 15 m, fusing the two data results in multi-spectral data with spatial resolution of 15 m, which is bene cial for accurate extraction of the glacial lake outlines. And one scene of GaoFen-6 image downloaded from the Chinese High resolution Earth Observation System (CHEOS) (https://login.cheosgrid.org.cn/) was used to delineate the outlines for the Jionglaco and the mother glacier in 2020, the downstream river channel and the sections from left bank to right bank. The spatial resolution of the multi-spectral band and panchromatic band for GF-6 images are 4 m and 2 m, respectively. After the same processing as Landsat ETM+/OLI, the multi-spectral image with a spatial resolution of 2 m was obtained, which clearly shows the outlines of the Jionglaco, the mother glacier and the river channel. Because GF-6 image does not match well with Landsat image, it was orthorecti ed using ALOS PALSAR DEM data with a spatial resolution of 12.5 m.
Glacial lakes are generally small in area and scattered in the overall region, making them easily confused with the surrounding mountain shadows and snow and other feature information when automated extraction is undertaken. Ultimately, a lot of manual checking and revision work after automated extraction is essential (Yang et al. 2019). In contrast, manual visual interpretation can exibly adapt to the vectorization of glacial lakes of complex background information and large differences in image quality. Meanwhile, the expert empirical knowledge is good at controlling the impact of adverse factors such as shadow, snow, and cloud cover on glacial lake extraction (Yang et al. 2019).
Therefore, in this study, manual visual interpretation method was used for extraction of glacial lake outlines, and the area error was controlled within one mixed pixel.

River network, cross-sections, and boundary condition
River network is used to de ne the modeling ood path, which was represented by the manually extracted river centerline from GF-6 PMS images ( Fig. 2(a)). Cross-sections are used to re ect the topography of the river channel, which determine the state and calculation of the ood evolution process ( Fig. 2(b)). The thalweg re ects longitudinal speci c drop of the simulated river channel, which impacts the travel speed and potential energy of the ood ( Fig. 2(c)).
Cross-section is the data that has the greatest impact on the model and can usually be obtained from actual measurements and DEM data. The former is more accurate but time consuming and labor intensive. By contrast, the latter is easy to obtain but the accuracy is in uenced by the accuracy of the adopted DEM data. Both measured crosssections (in Yi'ong Zangbo River) and DEM-based cross-sections (by ALOS PULSAR Digital Elevation Model (DEM) (12.5 m)) (in Xiaqu River) are included in this study. The DEM-based cross-sections are extracted roughly every 500 ~ 1000 m along the Xiaqu River ( Fig. 1(b)(c)).
Boundary le de nes the interaction between the model and the external environment, which is divided into two types open boundary and additional boundary (MIKE11 User Manual 2003). In this study, upstream open boundary of the model is the glacial meltwater entering the glacial lake, because the discharge is relatively small and has slight in uence on the model, a steady ow 1 m 3 /s was taken. The downstream open boundary is the eld measured water level-discharge relationship at the bridge location downstream of the Bazhui village ( Fig. 1(c), Fig. 2(d)). In addition, ve lateral entry ows were set as internal boundary in this model representing the base ow of the river channel consisting of the con uent tributaries which is 637.5 m 3 /s during ood season (May to October) according to the monitoring data.

LULC and Manning's N
Friction of the river channel to a given ow is determined by the Manning's roughness coe cient (Coon 1998), which is dependent on the land use and land cover (LULC) of the modeling river channel in the study area. In the study, the value of Manning's N of cross-sections were obtained from the GLC10 LULC product (http://data.ess.tsinghua.edu.cn/fromglc10_2017v01.html) with the spatial resolution of 10 m. The extent to which all transect widths covered along the river channel was extracted for the analysis of LULC types in this region. A total of nine types were identi ed in the potential ood area, of which forest holds the highest proportion of 51.43%, followed by grassland of 21

Volume and depth of glacial lake
Lake bathymetric data is one of the most crucial parameters in the dynamic modeling of GLOFs (Westoby et al. 2014). However, glacial lakes are usually distributed in relatively inaccessible environments, making eld studies be complicated. Hindered by turbidity, the derivation of re ectance-depth relationships of a lake from satellite sensor has not yet been reliably achieved (Box and Ski 2007). Consequently, the most of studies adopted an empirical approach to calculate the volume or depth of a glacial lake (Evans 1986 where V is volume (m 3 ) and D is mean depth of a glacial lake (m); A is the surface area of a glacial lake (m 2 ).

GLOFs modeling
Triggers such as ice/snow avalanches, rock fall or calving processes usually generate impulse waves capable of initiating an overtopping failure ultimately leading to a GLOF event (Ashim et al. 2019). In addition, extreme weather conditions with heavy precipitation or sudden warming can cause the water level of glacial lake to rapidly rise and even over ow. Based on the high-resolution imagery and 3D terrain rendering scene view of ArcGIS Earth software, the mean water level of the Jionglaco was measured as 3915.5 m, the width of the dam as about 394 m. The moraine dam is at, there is a water outlet of about 40 m width on the right side. The dam rises slowly from the outlet to the both ends, reaching a maximum height of approximately 3920.5 m. Assuming that the breach depth of dam is half (2.5 m) and full (5 m) of the elevation difference between the highest point of the dam and both the lake surface and the outlet river, the breach width of dam is twice (80 m) and triple (120 m) the width of the outlet, and the time of the peak ood is 1.5 h and 3 h respectively, thus ve breach scenarios were designed to simulate the ood process (Table 3).

Status and changes of glacial lake
There were 95 moraine-dammed glacial lakes with a total area of 11.38 km 2 in the Yi'ong Zangbo River basin in 2000 and 105 moraine-dammed glacial lakes with a total area of 16.87 km 2 in the basin in 2019, with increasing ratio in number and area of 10.52% and 48.24% (Table 4). In 2000, there were three glacial lakes with area larger than 1.00 km 2 , corresponding to the total area of 4.58 km 2 . In 2019, the number of glacial lakes of larger than 1.00 km 2 remained three, however the total area has increased to 8.38 km 2 with the ratio of 82.97%. The number and total area of glacial lakes in area intervals of 0.10 ~ 1.00 km 2 and 0.01 ~ 0.10 km 2 were both increased. However, the number and total area of glacial lakes smaller than 0.01 km 2 both decreased during the past decade, which does not indicate that the growth of glacial lakes has stagnated or regressed, but rather that the smaller lakes have been subsumed into other larger size intervals due to their expansion. As the largest moraine-dammed glacial lake, the Jionglaco's area has

GLOFs simulation
A lot of bridges and villages are distributed in the downstream valley of the Jionglaco. Based on Google Earth imagery and ArcGIS Earth imagery, a total of 15 bridge sites and 20 villages were identi ed in the study river channel. We only make statistic for GLOFs hydrological processes at the 15 bridge sites, including peak ow, ooding of breach, water level and ood propagation time. Depending on the height of the dam and the width of the outlet of the Jionglaco, ve different scenarios with different combinations of breach width (80 m and 120 m), depth (2.5 m and 5 m) and ood peak time (1.5 h and 3 h) were simulated (Table 3). Each scenario produces the different magnitude of ood peak at breach outlet (Fig. 3). The results show that scenario 5 produces the maximum GLOF peak discharge The GLOFs peak discharge, ood propagation time, water depth and ow hydrograph at the downstream bridge sites were generated for ve scenarios (  (Table 5 and Fig. 4).
Bridge sites 3, 5, 9, 11, 14 and 15 are the locations with a high density of settlements ( Fig. 1(c)). In the scenario with  Table 5). The peak discharge decreases between bridge site 5 and bridge site 9, after which all show an increasing trend. The ood propagation times for above bridge sites are 23 minutes, 1 hour and 8 minutes, 2 hour and 8 minutes, 3 hour and 20 minutes, 4 hour and 11 minutes, 5 hour and 23 minutes, respectively (Table 5). Although the peak discharge in the upstream area is smaller than the downstream area, ood propagation time is short, leaving insu cient time for people's transfer. However, the longer ood propagation time in downstream areas provides su cient time for people to move to safe areas if timely warnings are received from upstream areas.
The modelled water depth is prone to uncertainty due to variations in datum between observed water level measured from mean sea level and modelled water level based on WGS-84 datum (Thakur et al. 2016). The DEM data used to extract the cross-sections also affect the accuracy of the water depth, and if the DEM data is of high accuracy, the error will be relatively small. The deepest water depth is at bridge site 15 and the shallowest is at bridge site 3, which is in uence both by the shape of cross-section and the depth of base ow. Deeper and narrower cross-section as well as more discharge can lead to deeper water depths. In this study, the 6.73 m increase of water level at bridge 15 in the extreme scenario (Table 5), whereas at bridge 3 the rise is only 2.55 m.

Potential GLOFs of the Jionglaco
The glacial lakes with high risk of breaching usually has the following characteristics: (1) being terminal morainedammed glacial lakes (Cui et al. 2003 andChen et al. 2004) with the loose material and ice cores comprised as well as unstable and steep structured dam (Xu and Feng 1989); (2) having the relatively large area, area change and complete lake basin topography (Richardson and Reynolds 2000); (3) being close to or directly connected to the termini of its mother glacier which has steep slopes in the tongue section and fast change rate (Chen et al. 2003(Chen et al. , 2008(Chen et al. , 2009); (4) having the small crest. The Jionglaco is a typical terminal moraine-dammed glacial lake with a huge area of 5.69 km 2 in 2020, which is close to the maximum size of glacial lake in the Qinghai-Tibet Plateau in 2015 (Yang et al. 2019). The Jionglaco is directly connect to its mother glacier, which has retreat rapidly in company with the lake expansion. It has an intact lake basin topography and the average height difference between water level and dam is only 5 m. Therefore, the Jionglaco was identi ed as a potentially hazardous glacial lake by Duan et al. (2020) and Wang et al. (2020). However, the dam of the Jionglaco is very at and a natural outlet of 40 m in width has formed, which is a favorable condition for maintaining the water balance of the glacial lake and reduces the impact of hydrostatic pressure on the dam. In addition, the dam has been reinforced by the local government to prevent a breach disaster. The presence or absence of ice cores within the moraine dams cannot be determined at present and further eldwork is required. The slope in the mother glacier snout is smaller than 10°, which lowers the possibility of ice avalanche into the lake.
The main direct causes of GLOFs on the Qinghai-Tibet Plateau includes that (1) ice body, glacier tongue and mountain boulders collapse into lake, causing surges that over ow the dam and lead to dam failure (Clague and Evans 2000); (2) a sudden large increase in glacial meltwater or a sudden precipitation event that causes the lake level to rise and overtop the dam, leading to the dam to break (Richardson and Reynolds 2000); (4) melting ice cores in the dam body, in ltration water erosion, leading to the expansion of the pipe surge and eventually the failure of the dam (Yang et al. 2019). The main causes of GLOFs in the south-eastern Qinghai-Tibet Plateau are ice avalanches and ice slides, followed by pipe surges (Pu 2004). In contrast, GLOFs in the Himalayas are mainly caused by ice avalanches and ice slides (Richardson and Reynolds 2000). For the Jionglaco, the possibility of ice avalanche and rockfall caused GLOFs is relatively low due to the gentle-sloped glacier tongue. Second, because the area of the glacial lake is very large, the larger the area has the stronger bearing capacity for the mass entering the lake. Therefore, the Jionglaco is more resistant to disturbances caused by substances entering it, reducing the pressure on the dam from water uctuations (Wang et al. 2016). However, the occurrence of GLOFs cannot be ruled out in extreme cases due to the injection of large volumes of water, which may be caused by the large increased meltwater from mother glacier due to abnormally high temperatures, extensive basin in ow due to extreme precipitation and in ow from other GLOFs. A high-risk glacial lake, Jiongpuco is located to the southwest of the Jionglaco, which has also undergone an area expansion in the past 20 years (Duan et al. 2020;Wang et al. 2020). Jiongpuco's out ow directly injects into the Jionglaco, therefore the hypothetical GLOF of Jiongpuco will probably produce a large disturbance to the Jionglaco, thus triggering the occurrence of GLOF in the Jionglaco. In addition, there are a lot of oating ice on the lake in the summer of some years, which may collect at the estuary, and then block the lake outlet and rise the water level of glacial lake, and eventually cause the over ow of lake.

Limitation
The water level-surface area relationship is a critical input parameter of the simulation, which is the data representation of the breaching glacial lake and was calculated by empirical formula in this study. The empirical formula used to calculate the depth and volume of a glacial lake is a regression function of its area. Due to the huge surface area, the depth and volume of the Jionglaco are likely to be overestimated. However, the use of empirical formulas to calculate volume and depth is the only means when actual measurement data is lacking. In the future, the measured lake depth data will help to improve the accuracy of the simulation.
Currently, there is a lack of adequate grasp of the mechanism of glacial lake outburst ood and the formation of the breaching outlet. The simulation of GLOFs is an approximate physical phenomenon and carried out according to certain parameters, such as the width, depth of the outlet and duration of the breach discharge reaching the maximum value. Therefore, these parameters determine the magnitude of the peak discharge of the breach. In this study, the width of the breach is assumed twice and three times the current ow outlet, the depth of the breach is assumed half and full of the current average height of the dam and the outburst duration is assumed 1.5 h and 3 h referencing to an already GLOF event of 2.5 h in this basin.
In addition, the evolution of ood in the downstream area is in uenced by the shape of the cross-sections, which are the re ection of the downstream river topography. The accuracy of the cross-sections is determined by the accuracy of eld measurements or the accuracy of the DEM data. High-quality cross-section data is a guarantee of high-quality simulations. The measured cross-section data has a high degree of accuracy, but the DEM-based data can only re ect the topography above the water surface.
Despite the di culty in accessibility of data, we simulated ve GLOFs scenarios based on different breach width, depth and ood peak time using a one-dimensional model. Meanwhile, the ood characteristics at the breach outlet and the 15 bridge sites in the downstream area were analyzed. Generally, the study can provide some reference for the subsequent prevention, mitigation, and control of ood in the region.

Conclusion
The increase of number and area of glacial lakes caused by the glaciers retreat in context of global warming also increases the likelihood of GLOFs which pose a huge threat to lives and property in downstream areas. We delineated 95 moraine-dammed glacial lakes with a total area of 11.38 km 2 in 2000 and 105 moraine-dammed glacial lakes with a total area of 16.87 km 2 in 2019 in the Yi'ong Zangbo River basin. The largest glacial lake in the basin, the Jionglaco, has experienced a dramatic expansion over the past 20 years with an area increase of 3.22 km 2 . When the breach width, depth and ood peak time is set as 120 m, 5 m and 3 h, the peak discharge of 1570.61 m 3 /s at the breach outlet is the largest. And the smallest peak discharge at breach outlet occurs at the breach width of 80 m, breach depth of 2.5 m and the ood peak time of 3 hours.
In the scenario of the largest peak discharge (1570.61 m 3 /s) at the breach outlet, the peak discharges at bridge sites 3,  Table 5 Due to technical limitations, Table 5 is only available as a download in the Supplemental Files section.   GLOFs hydrograph of different scenarios at 15 bridge sites

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Table5GLOFspeakdischarge.docx