The authors would like to thank University of Paris, UMR 8586 PRODIG, France, Kadoorie Agricultural Aid Association, Nepal and the Earth System Sciences Program (ESS) of the Austrian Academy of Science for their financial assistance for this study. This paper is a contribution to the SarDYN Project “Natural dynamics, infrastructure construction and tourism economy in the context of a developing country: Nepal (Central Himalaya)” granted by the LABEX DYNAMITE and, to the TouRES (Resilience of tourism systems to natural hazards in the Himalayas) project. The authors are also grateful to Mr. Sushil Babu Dhakal, Mr. Manish Kumar Shah and Mr. Krishna Raj Adhikari at Department of Roads, Government of Nepal, for providing drawing of GK motor bridge, and Strategic Road Network Map. We acknowledge with thanks local people of Ghatte Khola, Dana, Myagdi for their help during this study. Comments by two anonymous reviewers helped improving the clarity of the manuscript.

1. Introduction

1Nepal is prone to various types of natural hazards (earthquake, flood, landslides, debris flow, glacier/landslide lake outburst flood, etc.) due to its rugged and fragile physical environment: very high peaks, high angle of slopes, variable climatic conditions, complex geology and active tectonic processes. In addition, unplanned settlement, haphazard construction, increasing population, weak economic condition, and low literacy rate (Rankin et al., 2017), make Nepalese population vulnerable to geo-hazards and hydro-hazards. Other factors are responsible for natural disasters in Nepal: lack of coordination among agencies related to disaster management, no clear-cut job description of those agencies, resource constraint, lack of technical manpower, lack of public awareness, very remote, rural communities, and absence of modern technology (McAdoo et al., 2018). Natural hazards such as landslides and floods are regular phenomenon on the Earth but they turn into disasters when these affect human beings or destroy their properties and most natural disasters are happening due to anthropogenic activities because we decide where and how to build our structures/houses (Gurung, 2019).

2According to Rankin et al. (2017) and Sudmeier-Rieux et al. (2019), in the past fifteen years, road and road-bridge construction in Nepal has topped community and government priorities, a trend likely to continue for many years to come. In 1998, Nepal’s road network was one of the smallest one in the world with a road density estimated at 3.22 km per 100 km² (SSRN, 2018). By 2018, it was increased to 9.14 km per 100 km² and continues to increase at a very rapid pace (SSRN, 2018). The national roads expanded rapidly from 4,740 km (blacktop, gravel and earthen) in 1998 to 13,447 km in 2018 (SSRN, 2018). These roads are causing landslides, thereby increasing geo-disasters in Nepal (Petley et al., 2007). The built, under-construction and future road network in Nepal as at fiscal year 2017-2018 is shown in the Strategic Road Network (SRN) map (fig. 1).

3About 250 to 300 road bridges are built annually across Nepal (Department of Roads, Government of Nepal, 2019). However, during every monsoon season (June to September), a considerable number of them are damaged or washed away by geo-hazards (debris flows, landslides, rock falls) and hydro-hazards (floods), as shown in (fig. 2). This has significant consequences not only for travellers but also for local economy, development, and tourism industry. There is no official data available yet about how many road bridges get damaged or washed away by geo- and hydro-hazards annually across Nepal but, on the basis of information from online national newspapers, a total of seventeen bridges have been damaged in the year 2019 alone, with complete destruction of thirteen road bridges (tab. 1). Yet a number of bridges, though destroyed or damaged by geo- and hydro-hazards, may not have been reported or published on newspapers. Bridges can be the most expensive structure along a road, so it is important to get them right the first time.

4The main aim of this study is to analyse the vulnerability of the Ghatte Khola (GK) road bridge, – a key area along the Kali Gandaki (KG) corridor – in respect to geo-hazards (rock-falls, landslides, debris flows) and hydro-hazards (floods) of the GK sub-catchment, and whether these have been accounted for correctly during the design and construction of the bridge. Actually, our initial fieldwork was carried out in January 2019, but had to be completed after the May 25, 2019 event, a severe hyper-concentrated flood that destroyed the nearly built GK bridge. Further, we assess and discuss the impacts of floods and landslides along GK catchment on surrounding settlement, KG road, and KG River. Eventually, adapted designs and suitable remedial measures are proposed to make them as safely as possible, while optimizing the costs and limiting the impacts to property and the environment.

2. Study area

5The GK torrential stream appears along the KG road corridor at Dana (Myagdi) (fig. 3). It is an intermittent, right-bank tributary of the KG River. The catchment area, 8.4 km², has a west-east elongated shape and extends from the upper ridges at 3,420 m to the confluence with the KG River at 1,400 m (fig. 4). The average stream gradient is of 30°, yet there are sharp contrasts between the upper and lower parts of the catchment, with a maximum inclination of approximately 65° up to the highest hill, where the channel is at its narrowest. Further down, the channel inclination decreases to approximately 13° at the exit of the gorges where the GK torrential stream suddenly widens across its fan, with an average slope angle of 7° to 8°. This large, distally confined debris fan (~1.5 km wide and ~2 km long), where the Dana village and infrastructure are located, encroaches more than 350 m over the 135 m wide KG floodplain (Fort et al., 2010). Not dated yet, it is probably not older than Holocene, according to former regional studies (Monecke et al., 2001; Fort et al., 2010).

Fig. 3 – Location map of Ghatte Khola (GK) torrential stream, Dana (study area), at the transition between the Lesser and Higher Himalaya (modified after Fort et al., 2010).

Fig. 3 – Carte de localisation de la rivière torrentielle Ghatte Khola (GK) et du village de Dana (zone d’étude), situé à la transition entre le Bas et le Haut Himalaya (modifié de Fort et al., 2010)

6The GK catchment morphology is controlled by the geology (Upreti and Yoshida, 2005), opposing the biotite-rich, landslide-prone, mica-schists of the Lesser Himalaya (right bank) to the steep quartzites and gneiss rocks of the Higher Himalaya (left bank), with a general N35°-40° dip (Fort, 2014). This asymmetric morphology influences the vegetal pattern on the lower, north-exposed hillslopes, in contrast to the south-exposed, steep slope where bedrock outcrops. Yet, the vegetal cover does not prevent the instability of hillslopes.

7Actually, the GK catchment behaves as a “landslide catchment” (Starkel, 1976), with pulsating, short debris-flows events initiated in the upper to medium GK catchment (fig. 5), as initially reported by Fort (1974). The north facing, dip slope (right bank), probably a deep-seated gravitational slope according to Chigira et al. (2019), is affected by several planar landslides, reactivated by pre-monsoon storms and during monsoon heavy rainfall (fig. 4). The GK catchment receives an average annual precipitation of 1,500 mm (fig. 6A) while it receives a total monsoon rainfall of 700 to 1,600 mm (fig. 6B). It is often subject to either debris- or hyper-concentrated flows affecting the village of Dana and the KG road. Rainfall data for the hydraulic analysis has been taken from Tatopani rain-gauge station, which is located ~5 km south of GK torrential stream (fig. 3). From time to time, during heavy cloudbursts, the landslide masses clog the river flow, where the channel bed is at its narrowest (fig. 5). As a result, the water is ephemerally blocked by the slide mass, until the water retention induces a sudden landslide outburst flood (LOF), causing bank erosion and riverbed widening downstream, all the most efficient along the low alluvial fan (Fort et al., 2010; Fort, 2014). These debris- or hyper-concentrated flows of the GK torrential stream are abrupt and short: in 1974, it lasted about one hour (Fort, 1974), and in May 2019, it lasted about half an hour (based on observations from local people). Yet such a behaviour of GK, occurring in average about once or twice a year, has a high potential to erode the stream banks, affecting the cultivated areas on the left bank and threatening the water intake feeding the mills (the “Ghatta”, Nepali) on the right riverbank. Since the last 12 years however, this behaviour may endanger the new KG road, interrupt the traffic and destroy the nearby infrastructure and settlements.

8A crucial point is the crossing of the GK channel. When the KG road was opened in 2008, an undersized bridge (wet section 42 m² vs 360 m² for the natural wet section) was constructed, but it was destroyed by a debris flow of moderate magnitude during the first monsoon season following road completion (Fort et al., 2010). Since the last three years, the road is progressively being upgraded to a 2-way tarred road and is going to be a major highway, as an annex of the new “Silk Road” that will cross Mustang linking Korala Pass (China border) down to Lumbini (Indian border). A 40-m long road bridge is under construction over the GK torrential stream where its channel is narrower, near the confluence with the KG River (fig. 7).

3. Available data and methods

9This study was conducted by several visits to the GK valley over the period from January to December 2019. The vulnerability of GK road bridge with respect to geo-hazards and hydro-hazards was analysed on the basis of (i) site selection, (ii) hydro-geomorphological mapping and analysis of historical sources, (iii) hydrological analysis, (iv) river cross-sections, (v) hydraulic data, and (vi) sedimentology of channel deposits.

10(i) The construction of a bridge at poor location can make the bridge more susceptible to failure. Good site selection involves many disciplines and includes preliminary engineering, hydrology and hydraulics, geomorphological concerns, roadway alignment, and environmental and geological concerns. All of these topics must be addressed to make sure that the structure is appropriate for the site. The GK catchment was chosen as study site since it represents a common problem that most road bridges in Nepal are experiencing. Further, GK torrential stream lies along the KG corridor that links Nepal with Tibet (China) under China’s BRI (fig. 3). Finally, the site of the bridge itself corresponds to the narrowest part of the channel of the GK.

11(ii) Hydro-geomorphological mapping of GK catchment was prepared to spatialise the active-channel dynamics, flooding zones, connectivity between the river, and the slopes (fig. 8). A detailed field investigation of the GK catchment was performed in January, March, June, October and December 2019, including a comprehensive geomorphological hazard study and a gathering of information on previous events based on available photo and video material, eyewitness reports from the local community and silent witnesses detected on-site (geomorphic features, traces of flood impacts, etc., as described in Kondolf and Piégay (2003), and Arnaud-Fassetta et al. (2009). More specifically, past-events were reconstructed from geomorphological mapping (fig. 8), deduced from field observations, analysis of natural cross-sections, and data provided by local people in order to have knowledge about past processes and events in GK catchment. Besides, past events (Fort, 1974) and interviews with three locals (68 – 70 years old, two from Ligma, one from Dana) were also used to prepare geomorphological mapping.

12(iii) Hydrological analysis is based on the annual rainfall recorded at Tatopani rainfall gauge station, the only rainfall gauge station at the proximity (5 km downstream) of the GK catchment. Rainfall data of fifty years period (1970-2019) have been analysed. The peak discharge of each year for fifty years period was computed based on the maximum rainfall intensity and catchment area by both Rational method and Water and Energy Commission Secretariat (WECS) method:

• In Rational method (Mulvaney, 1851; Kuichling, 1889; Shrestha, 2008; Rijal, 2014), discharge was computed by using empirical formula, i.e., Q = CIA, where Q is the design flow (in m³/s), I is the rainfall intensity (in mm/h), A is the catchment area (in km²), and C is the run-off coefficient. In the GK catchment, A = 8.4 km² = 8.4 x 10⁶ m², C = 0.3 (as the catchment is vegetated), and for the different rainfall intensity for each year, Q was computed for both peak and annual rainfall intensity.

• Water and Energy Commission Secretariat (WECS) method (WECS/DHM, 1990) was developed for estimating the peak flow of the ungauged catchments of Nepal. The regression equations, which have been proposed to estimate instantaneous flood of 2-year and 100-year return periods by WECS, have been modified by Sharma and Adhikari (2004) based on long-term data. They proposed the following equations for instantaneous peak-flow estimation. For 2-year return period, Q₂ = 2.29A^0.86 and for 100-year return period, Q₁₀₀ = 20.7A^0.72, where Q is the discharge (in m³/s), and A is the catchment area (in km²). A 100-year return period is often taken to design major bridges in Nepal, hence only Q₁₀₀ was computed by this method.

13It is common practise for the design of bridges on ungauged rivers in Nepal to run the calculations with both methods. Subsequently, the higher values provided by one of the methods are used for hydrological analysis. In this study, the higher values provided by Rational method were used, considering for the worst cases scenarios since these give extreme discharges in the river. The bridge should be able to pass all extreme flood discharges safely through it that are likely to occur in the river.

14(iv) River cross-sections were measured (with a margin of error of ± 10%) to perform hydrological analysis in HEC–RAS (2016) and to compute hydraulic data required for the design of a road bridge. A total of thirty cross-sections were taken along 1.5 km length at 50 m intervals and a longitudinal section of the GK torrential stream were also measured. Furthermore, six places (fig. 7) were chosen for the measurement of river cross-sections including one along GK road bridge to see any changes in river sections, particularly after flash flooding if there occurs any in the GK catchment in future.

15(v) Hydraulic analysis has to be performed to determine the effect of the waterway area on upstream floodwater elevations of a bridge (fig. 9). Hydraulic data, mainly linear waterway and high flood level were computed for both extreme and average discharges of GK torrential stream (the margin of error for the measurements is plus or minus 10%). Linear water way was computed using Lacey’s empirical formula (1958), W = 4.75√Q, where W is waterway (in m) and Q is discharge (in m³/s). The freeboard was computed by HEC–RAS analysis, which integrates the river geometry (river cross-sections and L, river longitudinal profile) and hydraulic data (discharge). HEC–RAS is a system software, which is capable of performing one dimensional steady flow analysis (for gradually varied flow) among many other capabilities (HEC–RAS, 2016). River geometry, roughness coefficient, and flood discharge are three important parameters for the hydraulic modelling in HEC–RAS.

16Since detailed historical data on flood are not available in GK catchment, they were derived from available precipitation data. Water and Energy Commission Secretariat (WECS) method and Rational method were used to estimate the design flood. In the hydraulic modelling of the GK torrential stream, channel geometry parameters were measured directly from the field survey. Since GK is a mountainous steep stream with boulders, Manning’s roughness coefficient was assumed to be 0.05 based on the river-bed material and mixed flow regime (Arcement and Schneider, 1989).

17With the available data, a steady flow model was constructed with energy gradient as a boundary condition and the flow assumed to be as a mixed regime because the bed slope is variable and is expected to encounter both sub-critical and super-critical flow. For mixed flow regime, steady flow model requires boundary conditions at both ends of the model, and from the longitudinal river profile, slope of the river bed So at upstream and downstream ends was measured to be 0.052118 and 0.081339 respectively. After the construction of the steady flow model, it was run combining all the parameters (geometry, flow and boundary conditions). The model finally generated water surface elevations (free board) at different cross-sections including the GK bridge section, which was the main matter of concern for this study. Fr is the Froude number, which demarcates the flow regime, whether it is subcritical (Fr > 1) or supercritical (Fr < 1). The model was run with the flow regime to be as supercritical flow regime, boundary condition is required only at the upstream end, the value of which is equal to the slope of the energy grade line and that was computed to be Se = 0.052153.

18Highest water surface profiles were calculated with the Rational method at different magnitudes of extreme flood discharges. The HEC–RAS model was also run for average flood discharges to see whether the high-flood-water level rises above the GK road bridge level or not. Further, highest water surface profiles were calculated at different magnitudes of average flood discharges.

19The sufficient scouring depth under the GK bridge was estimated according to Indian Road Congress (2000), where average scouring depth was given by: D_sm = 1.34 (q²/f)^0.33 (Lacey, 1958) where q = (Q/W) = (design flow/waterway of bridge), f = 1.76√d_m (Lacey’s silt factor), and d_m = mean size of bed material.

20(vi) Sedimentology of channel deposits lead us to specify the origin, transport, and deposition of the materials that compose the sediment layers at GK bridge site with its stratigraphic sketch. The goal is to understand the sedimentology of deposited materials and how good these materials are for the bridge foundation. Debris samples were collected from different places along GK torrential stream (fig. 7). Lab tests (sieve analysis) were performed at the lab of Tribhuvan University, Paschimanchal Campus, Pokhara, to determine, by the percentage of fine matrix sediment (FMS), the type of hydro-morphological event (i.e., the transport mode: torrential: FMS < 40%; hyper-concentrated flow: 40% < FMS < 60%; debris flow: FMS > 60%).

4. Results

4.1. Before the event of May 25, 2019

4.1.1. A catchment prone to landslides and flooding

21The geomorphological map (fig. 8) shows that both GK road bridge and KG road have been built on the complex of GK fan terrace – KG fluvial terrace (i.e., the medium one), which is prone to extreme flooding in both rivers KG and GK. It means that extreme flooding either in GK or KG may have serious damage on GK road and bridge. Furthermore, dominating mass movements (planar to complex and rotational landslides) at their respective source areas were identified along the right bank of GK in the upper catchment of GK (fig. 8). In its lower part, the higher terrace of GK torrential stream consists mainly of alluvial, whitish (gneissic) material by KG River, as seen on the opposite Garpar village, overtopped by GK, mud-supported dark debris. The medium terrace of GK fan is cut into the KG gravels and filled with 1-to-2 m thick debris flow material deposited by GK flood events (fig. 10). The lower fan-terrace of GK is a cone built by recent (a few decades) GK debris flow events at the junction with KG. Indeed, the KG River has potential to block GK torrential stream, which has also equal potential to block KG River during floods, as in 1974 (Fort, 1974, 2014). In the worst scenario, flooding may take place at the same time in both GK torrential stream and KG River, which may partly inundate the lower fan-terrace and medium terrace of GK fan. Part of the GK road bridge lies on the medium terrace (fig. 8).

4.1.2. Historical frequency of flood-events

22From the interviews with elderly local people, it was learnt that flooding in GK catchment is a repeated event and is taking place since their memory (70 years). The largest flooding in GK catchment to their living memory, had taken place about 50 years ago (AD 1971), the debris of which had covered both lower and medium terraces of GK fan and even to some areas of higher terrace along GK bank (fig. 8) killing 45 goats, 10 buffaloes, 6 cows, and 7 people. That means the location where GK road and bridge are built, is vulnerable to future extreme floods, hence extreme precautions need to be taken for this for infrastructure projects like bridge or road.

4.2. The event of May 25, 2019

23Like previous flash-flood events (Fort, 1974; Fort et al., 2010), the flood event of May 25, 2019 in GK catchment was triggered by heavy cloudburst (>16 mm/h) and LOFs. According to Dahal and Hasegawa (2008), a rainfall intensity of 12 mm/h is sufficient to trigger landslides in Nepal for rainfall events of less than 10 h, while an average precipitation of less than 2 mm/h appears sufficient to cause landslides if continued for more than 4 days. Further, an eyewitness (a local from Ligma who was present on that time at the upper catchment of GK herding his goats) narrates that there was a breaching of landslide-dammed lakes at least at two places with a big sound at the upper catchment of GK on May 25. The May 25, 2019 flood was a hyper-concentrated flow since the deposit consisted of some 50% to 54% fine (less than 2 mm in size) and 40% to 46% coarser materials (fig. 11).

4.2.1. Change on river cross-sections due to the flooding event

24Results from the comparison of the six river cross-sections along the fan (fig. 7, 12) before and after the May 25, 2019 flooding event (Gurung et al., 2019a, 2b) are given in the following and in (tab. 2). Along section A, we observed deposition of debris on the left bank and up to the flow area, then erosion at flow area, whereas along section B, some deposition took place on the left bank and scouring at flow section, thus the river-bed level was lowered and, the channel width was increased at flow section. Along section C, we observed deposition on the left bank (up to 0.5 m) and scouring at flow section on the right bank, thus the river-bed level was lowered and, the channel width was increased at flow section. Along section D, scouring at flow section occurred, so that the river-bed level was lowered and, the channel width was increased at flow section. Similarly, along section E, scouring (approximately by 1 m) developed on the left bank including flow section, thus the river-bed level was lowered. Along the GK road bridge section F, we observed deposition on the left bank, with a water flow (max. 0.5 m height) and significant deposition on the right bank from the river flow (up to 3 m thick), so that the river-bed level was significantly upgraded.

25During the flash flood of May 25, 2019, both erosion and deposition were observed at upstream and middle sections (fig. 12A-C) of GK torrential stream. Erosion predominated upstream from the bridge (fig. 12D-E) and, as the hyper-concentrated flow reached the narrow section close to the confluence, more aggradation was evident (fig. 12F). Eventually, during this event, we estimate that a total of 16,050 m³ (± 1,605 m³) of debris were eroded while a volume of about 18,100 m³ (± 1,810 m³) of debris was deposited in the lower GK catchment.

26Just upstream of the bridge, the flow is sub-critical with lower velocity and less energy. However, at the downstream side of the bridge, the scenario is just opposite with wider available waterway for the flood water to pass. Consequently, the flow is super-critical, with high velocity of 11.24 m/s and energy grade line as high as 8.44 m. The deduced water surface from HEC – RAS led to check the design adequacy of the under-construction road bridge over the GK torrential stream.

4.2.2. Damage to the under-construction GK road bridge

27The flash flood destroyed the nearly completed GK road bridge (fig. 13). It was unfortunate to lose the bridge but this was not an unexpected event. There is no record of the highest flood level that reached that bridge but, based on the photo taken few minutes after the washing away of the GK bridge (fig. 14), the highest water level might not have reached the deck level of this under-construction bridge. The failure of the GK bridge does not seem to be due to the extreme flood discharges, but more likely to the restriction of river flow by the temporary scaffolding work (fig. 15).

4.3. After the event of May 25, 2019

4.3.1. The engineering works

28The GK bridge construction (fig. 9) started in March 2018. During the flood of May 25, 2019, the bridge was washed away when it was on the verge of completion. In fact, its destruction was due to the poor engineering works carried out at wrong time. The waterway was almost blocked by scaffolding work and only a small hume pipe of 70 cm in diameter was laid for the water flow (fig. 15). It was sustainable as long as there was no flood. Since the work was being done in May, the season of cloudbursts in Nepal, floods were not unexpected at that period and, in the specific case of GK, in the form of a destructive LOF. Without the landslide activity it might have been a smaller flood causing no serious damage to the bridge.

29The reconstruction work of GK bridge started from November 2019 but despite the huge loss of May 2019 flooding, the work was resumed without any change on the original design and any consideration of river-flow dynamics in GK catchment. (i) In spite of the fact that GK receives recurrent flash foods once or twice a year, the present GK bridge design did not include sufficient river-bank protection works. (ii) We also noted that no adequate flood protection countermeasures were constructed to safeguard the bridge structure. The gabion baskets used as bank protection along GK were poorly designed and installed, with a lack of enough preventive maintenance, so that most gabion basket structures collapsed during monsoon flood (fig. 16). Due to the low-bank height and loose-bank materials, there is a strong possibility that future floods in GK might again encroach over the fields and settlements upstream from the bridge (fig. 7A-E). (iii) Further the existing bridge site consists of loose KG River deposited materials overlain by GK debris flows (fig. 10), both prone to river erosion and yet, both bridge abutments of GK bridge are on shallow foundation.

30The newly built GK bridge is now on the finishing stage (fig. 15). Since it is built close to the confluence with KG River, more aggradation is likely to occur on the GK bridge site such as during the May 25, 2019 flooding (fig. 12F). It means that the present free-board (~3 m) of the GK bridge above the river bed is likely to decrease due to possible aggradation processes at GK bridge section. In addition, the present cavity (fig. 15) will be refilled when the bridge is completed.

4.3.2. Hydromorphological assessment of engineering works

31First, insufficient linear waterway is observed. From hydrological analysis, the linear water way (clear distance between bridge abutments) of the under construction GK road bridge was computed to be 36 to 48 m (tab. 3) for 100 years return period for extreme flood discharges of 50 years (1970-2019). The provided linear water way of GK road bridge is 38 m (fig. 9). From Table 3, the computed linear water way is less than 38 m for seven particular years out of total fifty years. That means the linear water way of GK road bridge does seem to be sufficient only for seven times out of fifty years’ extreme flood events (extreme flood discharges are computed based on the extreme rainfall data of all 12 months of a year). In another word, the provided linear waterway is insufficient to allow extreme flood discharges safely for forty-three times out of fifty (86%) extreme flood events, hence the under construction GK road bridge does seem vulnerable to future extreme floods. But the GK road bridge was found to be safe for average flood discharges (calculated based on the average rainfall data of all 12 months of a year) since the linear water way for average flood events was computed to be 33.63 to 37.81 m, which is less than the provided water way, i.e., 38 m (tab. 4). Hence the under construction GK road bridge does seem safe to future normal/average floods.

32Though this analysis shows that the linear water way of GK bridge was found to be insufficient in more than 80% cases, this may not be truly applicable on the field. Furthermore, this approach may underestimate the discharge of flash floods since hyper-concentred flows, which often develop with run-off discharge during flash floods in GK, has not been taken into consideration. Since the linear waterway is directly proportional to flood discharge, hence the probability of bridge damage may even be more than the computed here. Furthermore, since the linear water way is insufficient by 10 m at the max. (48 m vs 38 m) that is some 26% more than the provided water-way (38 m), it does not mean that the bridge gets damaged or will be washed way. This will lead to increase the bank-erosion process and rise on high-flood level at the bridge section and thus may have an adverse effect on the function of the bridge.

33Secondly, insufficient free-board is observed. In HEC-RAS analysis, the highest flood level of 100 years’ return period in GK catchment for extreme flood discharges was found to have overtopped the bridge i.e. above the deck level of the under-construction GK road bridge, hence the GK road bridge is at high risk from potential flooding (fig. 17). The computed highest water-surface elevation for extreme flood discharge at the bridge location is 122.52 m, which is 1.52 m (122.52 m minus 121.00 m) above the bottom level of the deck. This means that the flow is obstructed at this location due to the insufficient flow area to safely pass the extreme flood discharge. The major cause of the damage of the bridge could be (i) the constriction in flow area required for the safe discharge of extreme flood, and (ii) the large amount of debris that may have come along with the flow at the time of flood event.

34Furthermore, for annual flood discharges in GK, the computed high-water-surface elevation at the bridge location of 100 years’ return period was found to be below the deck level of the under-construction GK road bridge. The computed highest water surface elevation for annual flood discharge is 119.89 m level, which is 1.11 m (121.00 m minus 119.89 m) below the bottom level of deck of the under-construction GK road bridge, therefore the GK road bridge could be safe for annual flood discharges (fig. 18). Department of Roads, Government of Nepal (2010) has recommended a minimum free-board of 1 m for the discharge up to 200 m³/s, hence the free-board of GK bridge for annual flood does seem to be sufficient. However, we have to qualify this as strictly hydraulic results. Floods, even those that occur every year, are hyper-concentred flows that progress downstream as hydro-sedimentary waves. These waves are able to block the bridge, especially as the GK channel width is reduced downstream due to anthropogenic forcing (hydraulic deflectors, dikes, walls, roads). We confirm here that the management of a mountain stream is not only to be concerned with the water heights but also with the sediments they carry which play an important role to change the 3D geometry of the riverbed.

35Finally, sufficient scouring depth is required. The most common cause of bridge failures is scouring of bridge piers and abutments by river flow, hence the bridges should be designed to withstand scouring from large floods and from stream instabilities expected over the life of a bridge. According to Indian Road Congress (2000), the average scour depth in a bridge (D_sm) is given by D_sm = 1.34(q²/f)^0.33. Here q = Q/W = design flow/waterway of bridge, f = 1.76 d_m, with d_m = mean size of bed material. For the extreme flood, Q = 100.74 m³/s (discharge of 2003, see Table 4), W = 38 m waterway, and d_m = 1.1 m; the average scour depth (D_sm) was computed to be 1.7 m. Further, the maximum scour depth for bridge abutments (Indian Road Congress, 2000) is = 1.27D_sm = 1.27 x 1.7 = 2.16 m. But the provided scour depth of the under-construction GK bridge is 2.82 m (fig. 9). Hence the under-construction GK bridge was found to have sufficient scouring depth.

5. Discussion

5.1. Flash floods and river-training works

36Flash floods are not unexpected events in GK catchment and the flash flood that occurred on May 25, 2019 is just an example of this type of event. In the future, many such floods are expected to occur in this catchment and these may have the same or more effects on river cross-sections, depending on their magnitude. Therefore, all engineering structures and settlements must be built according to the dynamic behaviour of GK torrential stream. The under-construction GK bridge is vulnerable to future extreme floods, though it still looks safe for average floods. The present conventional bank-erosion protection works (Steel wire Gabion box retaining wall) along GK seem to be inadequate to deal with the scale of erosion that eventually takes place during pre- or monsoon flash floods (fig. 16A). Every year, during monsoon, gabion walls are failed by high flows in GK and shall be re-built before the next monsoon in order to protect river bank erosion. But this is simply not making worth. Since GK banks consist of alluvial deposited loose boulders or clasts mixed sandy gravel matrix, they are easily erodible by river, hence a proper retaining structure with reno-mattress, i.e., engineering structures built at river bed level to prevent retaining walls from scouring, is recommended (fig. 16B).

5.2. Safe design of bridges facing flood hazard

37A study by Wardhana and Hadipriono (2003), which investigated about 500 cases of bridge destruction that occurred in the USA between 1989 and 2000, revealed that flood-associated hydraulic factors, such as insufficient waterway, freeboard and scouring were the most common causes of bridge destruction. In another study by Lebbe et al. (2014), majority of bridges in Queensland, Australia were damaged and failed due to the flood since the waterway and freeboard of bridges were not sufficient to safely pass the high flood. The situation is not different in Nepal too.

38At least seventeen bridges got damaged (thirteen completely destroyed) by floods and landslides in Nepal in 2019 (tab. 1). Three of them (18%) were under construction indicating that the potential onset of natural hazard events was not carefully considered in timing the construction of the bridges. Further investigations on destroyed under-construction bridges are necessary to better evaluate the extent of this issue. However, considering geo-hazards and river-flow dynamics is essential not only for individually designing safe and sustainable bridges but also for carefully scheduling the construction of the bridges. From GK bridge study, it is clear that the bridge is not designed and built up to the standard considering geo-hazards and river-flow dynamics in particular. As mentioned by Reed (2015) for whom historical floods are often overlooked in river-flood risk assessment, past flood events are rarely considered during the design of road bridges in Nepal and GK bridge is not an exception.

39Hearn (2002) had stated that the wider use of engineering geomorphology in Nepal has been disappointingly low and there are instances where a lack of engineering geological or geomorphological appreciation has led to recurrent problems, leading to modifications of design and even alignment of road and bridges. Even today after 18 years of Hearn’s statement, it still looks valid since bad news about the damage to roads and road bridges by floods and landslides, especially during monsoon season, remain common in Nepal. River hydrology and geomorphology are often overlooked in the design phase, which might lead to the failure of bridges. The GK road bridge is just a representative case where catchment geo-hazards and river-flow dynamics in particular are not sufficiently incorporated in the design. This does not necessary mean that the GK bridge would be washed away by floods in near future because though the GK bridge is vulnerable to extreme floods, the bridge yet looks safe for average flood discharges.

40Furthermore, a completely no-failure design could be exceedingly costly. The cost of under-construction GK road bridge is US$ 405,000. A question can be discussed whether such a costly bridge over GK River was necessary or not because it is true that GK catchment receives flash floods on few occasions of the year and most of the time, the stream remains dry. On this aspect, the under-construction GK bridge does not look economically viable, in contrast to inexpensive, low-water crossings such as concrete fords, as could have been planned and carefully designed along the present road, just upstream from the GK road bridge (fig. 19). In fact, concrete fords are easier and quicker to repair or build even if they get damaged or washed away in flash floods. Though ford-type structures imply some periodic or occasional traffic delays during periods of high flow, it takes no more than a week to repair or clean them off and the traffic can be resumed. But when a bridge is broken, it takes months to rebuild. Furthermore, since GK flow stops the traffic only for few occasions of the year, building a ford would have been a better option both economically and functionally than an expensive yet vulnerable road bridge. There are many fords built along KG corridor (between Kushma and Beni), which are serving well.

41A new strategy to think about could be to build fords at places where they can easily be rebuilt if damaged or destroyed. That way quite some money could be saved to be used to construct safe and well-designed bridges at dangerous locations where fast reparation or rebuilding of bridges is not possible (e.g., due to topographical, geological and geomorphological constraints).

6. Conclusions

42Many road bridges in Nepal are being built without consideration of geo-hazards and river-flow dynamics in particular. In this study, geological, hydrological and geo-hazards of GK catchment were examined in detail in order to assess whether the under-construction GK road bridge has incorporated these properly or not. From this study, the bridge over GK torrential stream does seem vulnerable to future extreme floods since the provided linear water way and freeboard for the highest flood were found to be insufficient to safely pass the extreme flood discharges of 100-years return period. Engineers (designers), therefore, should consider risk of possible geo-hazards of the river catchment and take account of geomorphological, hydrological and analyses prior to the construction of bridges so that bridges do not get washed away or damaged during extreme flash floods. The present design of the GK torrential bridge has not considered enough the river-flow dynamics and catchment geo-hazards, hence the bridge is still subject to future extreme flooding. A simple inexpensive crossing such as a concrete ford would have been a far better option in GK torrential stream than the expensive yet risky present bridge, as illustrated by the last May 25, 2019 event. Eventually, given the continuing impacts of hydro-geomorphological hazards on road and bridges in Nepal, it would seem that understanding mountain streams and slopes dynamics is a prerequisite for safe, sustainable design, and good-timing of the construction of these mountain infrastructures.

*Corresponding Author. Tel: +977 98460 29188
jyonus@hotmail.com (N. Gurung).