Melamchi Flood Disaster in Nepal Damage and Risk Quantification with Drone Survey, Satellite-Based Land Displacement Analysis, and 2D Flood Modeling EUROPEAN UNION M ELA MCHI FLOOD DI SASTE R I N NE PAL This work is a product of the staff of The World Bank and the Global Facility for Disaster Reduction and Recovery (GFDRR) with external contributions. The findings, analysis and conclusions expressed in this document do not necessarily reflect the views of any individual partner organization of The World Bank, its Board of Directors, or the governments they represent. Although the World Bank and GFDRR make reasonable efforts to ensure all the information presented in this document is correct, its accuracy and integrity cannot be guaranteed. Use of any data or information from this document is at the ‘user’s own risk and under no circumstances shall the World Bank, GFDRR or any of its partners be liable for any loss, damage, liability or expense incurred or suffered which is claimed to result from reliance on the data contained in this document. The boundaries, colors, denomination, and other information shown in any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. RIGHTS AND PERMISSIONS The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Cover photo: The affected area in Melamchi Bazar by the June and August 2021 flood events. Photo taken by Dr. Ranjan Kumar Dahal in December 2021. 2 M ELA MCHI FLOOD DI SASTE R I N NE PAL Table of Contents Executive Summary 7 Introduction 9 Drone Survey 11 Flood Modeling 17 Satellite-Based Land Displacement Analysis 22 Field Observations 26 Conclussions and Lessons Learned 31 3 M ELA MCHI FLOOD DI SASTE R I N NE PAL Acronyms ADB Asian Development Bank DTM Digital Terrain Model DSM Digital Surface Model GSD Ground Sampling Distance GIS Geographic Information System GPS Global Positioning System DEM Digital Elevation Model MCM Million Cubic Meter AOI Area of Interest NSSDA National Standard for Spatial Data Accuracy HEC-RAS Hydrologic Engineering Center’s River Analysis System INSAR Interferometric Synthetic Aperture Radar MRWDS Melamchi River Water Diversion Subproject NDRRMA National Disaster Risk Reduction and Management Authority UAV Unmanned Aerial Vehicles WB World Bank 4 M ELA MCHI FLOOD DI SASTE R I N NE PAL Acknowledgments Production of the report was led by Masatsugu Takamatsu, Disaster Risk Management (DRM) Specialist and Hemang Karelia, Senior DRM Specialist, in South Asia Region Climate Change and Disaster Risk Management (SSACD). The report was prepared by Professor Thomas Oommen from the Department of Geological & Mining Engineering & Science, Michigan Technological University and Dr. Ranjan Kumar Dahal from the Central Department of Geology, Tribhuvan University with significant contributions from the drone team led by Vikas Bhusal from the Tri Max and the satellite team lead by Pamungkas Bujed from the Synspective. The team extends huge appreciation to the following peer reviewers from the World Bank for providing insightful and useful comments: Rikard Liden, Lead Energy Specialist; Ximing Zhang, Senior Dams Specialist; and Lixin Gu, Rubika Shrestha, and Feriha Mugisha from the Nepal Water Team. The report was greatly benefited from editorial work by Mehsum Basharat, and professional graphic design by Juan Velasco from 5W Infographic. The team would like to extend sincere appreciation to Mr. Anil Pokhrel, the CEO, and Mr. Rajendra Sharma, Under Secretary, of the National Disaster Risk Reduction Authority under Ministry of Home Affairs, Nepal for supporting the entire engagement and Mr. Saugata Dasgupta and Mr. Hans G. Enggrob from the Asian Development Bank for close coordination, information share, technical discussions, and report reviews. Special acknowledgments go to the following World Bank staff: Atishay Abbhi, Sulochana Nepali, Sujit Maharajan (Consultant), and Dr. Manita Timilsina from the Geotech Solutions International, Nepal (Consultant). Finally, the team would like to appreciate support from Abhas Jha, Practice Manager of SSACD, and Niels Holm-Nielsen, Practice Manager of the Global Facility for Disaster Reduction and Recovery (GFDRR). The study was conducted under the “BUILDING RESILIENCE TO LANDSLIDE AND GEO- HAZARD RISK IN THE SOUTH ASIA REGION” project, in the frame of the European Union - South Asia Capacity Building for Disaster Risk Management Program funded by the European Union. The Task Team would like to express sincere appreciation to the European Union for its generous financial support to the South Asia Geohazard project. This technical report contents are sole responsibility of the Task Team. The European Union is not responsible for any use that may be made of the information contained therein. 5 M ELA MCHI FLOOD DI SASTE R I N NE PAL Foreword Nepal is highly vulnerable to climate change and is prone The NDRRMA quickly responded to the disaster, arranging to disasters caused by natural hazards. With its unique timely site visits and sourced inputs from experts, which geography, the country remains exposed to several climate were crucial for effective hazard response and recovery risks such as glacial lake outburst floods (GLOF), floods, planning. In close collaboration with the NDRMMA, the World and other water-related hazards triggered by torrential rain Bank urgently provided support to the disaster with a series enhanced by snowmelt during monsoon season. Millions of activities including a drone survey to identify all damaged of Nepali population and infrastructure are estimated to houses, map out inundation areas, and record the changes be at risk from climate change. The National Disaster Risk in topography caused by massive erosion and deposition. Reduction and Management Authority (NDRRMA) was These quick and focused data collection and analysis in the established in 2019, with a mandate to coordinate and area with difficult accessibility allowed for prioritization of implement activities related to disaster risk reduction and resource allocation and understanding of remaining flood management in Nepal. risks, as highlighted in the report. On June 15, 2021, the Melamchi River experienced massive Based on the data and information gathered and shared, the flooding and caused disastrous damage downstream. Over NDRRMA was able to focus on efforts needed for hazard six consecutive days, the region reported around 200 mm of awareness, dissemination, and disaster risk reduction rainfall. The massive downpour coupled with rapid snowmelt led to the erosion of glacial deposits in the far upstream of the planning in the region. The findings of the World Bank Melamchi watershed. This phenomenon caused the formation activities were also presented in two-hybrid workshops with of a landslide dam and its eventual collapse in Bhemathan. policymakers and partners including ADB and ICIMOD. The flood event caused at least 17 casualties and damaged more than 540 houses and critical infrastructure including The NDRRMA appreciates the timely efforts and resources bridges, roads, and the headworks of the Kathmandu Valley mobilized by the World Bank to contribute to the response Water Supply Project financed by the ADB, and it may take efforts and in helping understand disaster risks in a complex up to several years to recover from the economic impacts. mountain setting. Anil Pokhrel Chief Executive National Disaster Risk Reduction and Management Authority (NDRRMA), Ministry of Home Affairs Government of Nepal 6 M ELA MCHI FLOOD DI SASTE R I N NE PAL Executive Summary This is a technical case study report demonstrating monitor, Risk Reduction and Management Authority (NDRRMA) and assess, and quantify high-mountain geohazard risks for other governmental agencies to plan for future actions. government agencies based on the flooding and landslides These activities include: i) an Unmanned Aerial Vehicle in Melamchi Watershed in Nepal. On June 15, 2021, the (UAV)/drone-based survey to document the damage from Melamchi River experienced massive flooding and caused the event, ii) flood modeling to understand the conditions disastrous damages in the downstream. The event led generated by the event and to estimate the future risk to the to several casualties and destroyed houses and critical community, and iii) satellite-based Synthetic Aperture Radar infrastructure, including the headworks of the Melamchi (SAR) analysis to identify location vulnerable to landslides in River Water Diversion Subproject (MRWDS) funded by the the region. Asian Development Bank (ADB). Further investigations identified that a landslide formed a natural dam in the The drone survey was carried out from July 6th to August proximity of the confluence between Pemdan Khola and 30th, 2021 and covered a river stretch of approximately Melamchi River. The natural dam blocked the Melamchi river 100 km. The survey provided timely documentation of the and pooled water upstream. The natural dam eventually damages from the event that could be used to calibrate and breached and abruptly released water, causing an outburst validate the flood modeling. The products derived from the of flooding. The overall situation was intensified by the drone survey include georeferenced orthorectified images, heavy rainfall and possible glacial lake outburst. A second 3d-point cloud, Digital Terrain Model (DTM), and approximate flooding event equally severe as the first one was observed estimates of erosion and deposition along the channel. The on August 1st. Following the flooding, the World Bank flood was modeled using Hydrologic Engineering Center’s initiated three activities to support the National Disaster River Analysis System (HEC-RAS) as a two-dimensional Figure 1: Diagram illustrating the flow of post-flood disaster risk management engagement. 7 M ELA MCHI FLOOD DI SASTE R I N NE PAL non-Newtonian flow with the rainfall data collected from risk management practice. the Global Precipitation Measurement (NASA), and a high-density mix of water and debris (10% debris in water) The area of interest and study focuses were decided based was selected for the final non-Newtonian dam breach on close coordination with NDRRMA who responded to model. The modeled flood compared well with the flood the disaster immediately after the June flood and needed extends mapped from the drone survey. Comparing the accurate damage and risk information. The World Bank modeled flood velocity with the erosion and deposition team also worked closely with the ADB team who focused regions identified from the drone imagery showed a strong on the detailed damage assessment and risk mitigation correlation. The average flood velocity was ~3 times more planning of the headworks of the MRWDS through data for the erosion region than the deposition. share, technical discussions, and mutual report reviews. The field investigation conducted by the ADB team was useful Sentinel-1 satellite Synthetic Aperture Radar (SAR) data was for the World Bank team in validating the flood modeling and then used to perform Interferometric Synthetic Aperture satellite-based SAR observations. The close coordination Radar (InSAR) analysis to map land displacement. InSAR and data share with partners equipped the government analysis was carried out using the data from April 2017 and other stakeholders with useful quantitative information to June 12th, 2021. The results from the analysis assisted to better understand the complex and disastrous flood in identifying several hotspots within the area of interest event that occurred in the Melamchi River and provided a that are vulnerable to ground movement. The SAR work basis for future risk reduction and management. Figure highlights the benefit of cutting-edge research to disaster 1 conceptualizes how the studies will be helpful in the risk managers and relevant authorities as a new geohazard resilience planning effort in the Melamchi Watershed. 8 M ELA MCHI FLOOD DI SASTE R I N NE PAL Introduction On June 15, 2021 the Indrawati basin consisting of the for the Government of Nepal to alleviate the acute water Melamchi, Yangri, and Larke rivers was struck with heavy shortage in Kathmandu Valley and reduce the incidence of rainfall causing disastrous flooding in the downstream disease caused by poor water quality. Figure 2 shows the area of the Melamchi River. Based on the National Disaster area of interest. Risk Reduction and Management Authority (NDRRMA), 17 causalities and at least 23 were reported missing due to According to the Department of Hydrology and Meteorology landslides and floods caused by this event. It also led to (DHM), Melamchi and Indrawati basins started receiving the destruction of houses and infrastructure in several rural rainfall starting June 9th, 2021. The highest hourly municipalities, including the damage to the headworks of the precipitation on June 10th was recorded at 22 millimeters MRWDS Project that supplies drinking water to the people in (mm); by June 11th, it had increased to 37 mm per hour. Kathmandu Valley. The Project is a National priority project On 14 and 15 June, some rainfall was recorded at around 10 mm per hour. On June 11th, Sermathang recorded more than 100 mm of daily rainfall. Collectively, during the 6-day interval, the station had received more than 200 mm of rainfall. The intense rainfall and rapid snowmelt resulted in the erosion of glacial deposits in the headwaters of the Pemdan Khola (meaning creek in Nepali), Yangri River, and Larche River. Early investigations documented the formation of a landslide dam and its subsequent collapse in Bhemathan (situated in Langtang National Park). A past landslide formed the natural dam in the proximity of the confluence between Pemdan Khola and Melamchi Khola on Melamchi Watershed, as shown in Figure 3. The natural dam blocked the Melamchi River and pooled water upstream temporarily and was abruptly released, causing an outburst flood. The large volume of water was released downstream, destroying the riverbank settlements, bridges, and road on its way to Melamchi Bazar and causing heavy coverage of a thick deposit of mud and debris to the town downstream. The overall situation was intensified further by heavy rainfall runoff with snowmelt, possible glacial lake outburst, and moraine erosion in the Pemdan Khola region, as noted by ICIMOD and the Nepal Engineers’ Association. The high- speed flow of water hyper-concentrated with debris (glacial deposits in the headwaters of the Pemdan Khola and fan, talus, and lake deposits in the Bhemathan area) likely eroded the channel and riverbank throughout the stretches, causing numerous riverbank collapses and landslides. A second flood event occurred on 1st August 2021, possibly due to heavy Figure 2: Location of the Melamchi Watershed in the rainfall and erosion of the sediment deposited as a results of Sindhupalchowk District 60 km Northeast of Kathmandu the first Landslide Dam Outburst Flood (LDOF) event. 9 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 3: Comparison of the pre-event satellite image with the post-event drone image at Bhemathan Goal and Scope of the Studies flood zoning in the Melamchi Bazar area Following the June 15th flood, referred to as the “Melamchi Support the GoN for preparing preliminary recovery plans Flood”, the World Bank initiated the drone survey as per To address these goals three activities were conducted i) the request from the NDRRMA to support the Gov in an Unmanned Aerial Vehicle (UAV)/drone-based survey to understanding the cause and damage of the disaster. While document the damage from the event, ii) flood modeling to reviewing the progress of the drone survey with NDRRMA, understand the conditions generated by the event and to we identified further needs to understand the cause of estimate the future risk to the community, and iii) satellite- the complex flood, and evaluate remaining geohazard based Synthetic Aperture Radar (SAR) analysis to identify risks in the watershed. Therefore, the SAR analysis and location vulnerable to landslides in the region. These three flood modeling were initiated with an agreement with the activities complement each other as the drone survey helps NDRRMA. The eventual goal of the initiative was to help to document the post-flood site condition, flood modeling Gov plan the reconstruction and mitigation actions in this utilizes drone data for validation and provides a valuable tool region by providing: for flood risk reduction and planning, and the SAR analysis helps to evaluate historical and future landslides and its Urgent support to NDRRMA and key stakeholders risk to the area.In addition, this report includes a section on Coordinate, collaborate, share data, and discuss findings field observations conducted by the NDRRMA. This report with interested parties, including the ADB expert team summarizes these activities and includes important lessons tasked to carry out a disaster damage assessment of the for future events similar to Melamchi’s that hazard planners headworks of the MRWDS and the University team on and policymakers can use. 10 M ELA MCHI FLOOD DI SASTE R I N NE PAL Drone Survey The purpose of the drone survey was to document the The schematic showing the method followed by the drone damages from the floods and landslides and develop a high- team is presented in Figure 4. resolution ortho-rectified mosaic imagery and Digital Terrain The drone survey was undertaken by Trimax IT Infrastructure & Model (DTM) at Sindhupalchowk, Melamchi, Helambu, and Services Pvt Ltd with DJI drones between July 6th and August Panchpokhari flooding area along the river basin. Due to 14th. After prior data collection, the drone team used several their lower operational risks, and ability to remotely capture drones to capture the photographs from the area (Figure 5). high-resolution imagery, drone-based surveys are effective The team collected ground control points to geo-reference the for post-disaster documentation in remote areas such as photos accurately when possible. The spatial extent and flight Nepal where access is a critical challenge. period of the drone surveys are presented in Figure 6. Figure 4: Schematic of the method for the drone survey 11 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 5: Team capturing data using the drone, damage at Melamchi Bazar from the drone image, and the steep hilly terrain where the drone was used to document the damages and to collect post-flood topographic data Figure 6: The spatial extent and flight period of the drone surveys 12 M ELA MCHI FLOOD DI SASTE R I N NE PAL The data collected using the drone was analyzed using also useful in identifying several landslides along the river remote sensing and photogrammetry software. Results from triggered by the toe undercut by the debris flow. The one the drone footage indicated a significant change in the river observed at Bhemathan was the largest and had dammed pattern at Bhemathan (Figure 7). The drone imagery was the upstream portion of the river. Figure 7: Approximate dimensions of the landslide deposits obtained from the drone imagery 13 M ELA MCHI FLOOD DI SASTE R I N NE PAL 7a) longitudinal section showing the river profile before and after the flood 7b) cross-section showing the river profile before and after the flood. Before-flood DEM source is WDRF (5m spatial resolution), and the after-flood DEM source is drone data (50 cm spatial resolution) as part of this project 14 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 8: Drone image showing the muddy and clear water from Pemdan Khola (west) and Melamchi River (east). The difference in the color of the water (muddy vs. clear) the damaged house inventory (Figure 10). from Melamchi River and Pemdan Khola at its confluence near Bhemathan indicated that most of the deposition Photos and videos taken by the drone survey team provided occurred from the Pemdan Khola (Figure 8). very clear visuals of the post-flood ground conditions along the Melamchi River, and the prepared topography based The drone survey data was used to estimate the erosion on the drone imageries were useful in understanding the and deposition volumes. It was made possible by generating massive river erosion and deposition caused by the flood. It longitudinal profiles along the river on the pre- and post-flood should be also noted that the access to the upstream areas DTMs. During this exercise, caution was implied to avoid was extremely difficult because of the high altitude and longitudinal profiles passing over wet areas as the post-flood limited road access. The team had to hike long distances DTM does not penetrate floodwaters. Figure 9 illustrates the with the survey equipment, safety gear, and other necessities longitudinal profile and the area of maximum flood extent. to complete the survey of most areas. NDRRMA provided a helicopter ride to the drone team to access the Bhemathan In addition, the drone data and site visit information were area. Based on the pre- and post-flood topographic data used to develop an inventory of the damaged houses. The comparison, major depositions occurred in Bhemathan inventory includes information on coordinates for each (16 MCM) and the stretch between Melamchi Ghyang and damaged house, owner and resident information, building Melamchi Bazar (10 MCM), while major erosion occurred story and types, and photos. In Melamchi Bazar and in the stretch between Bhemathan and Melamchi Ghyang Helambhu, a total of 291 and 252 houses were registered to (26 MCM). 15 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 9: This figure illustrates the area of maximum flood extent, wet area when the drone survey was conducted, profile path, and pre and post-flood elevations Figure 10: Inventory of the damaged houses generated from the drone survey and field visit 16 M ELA MCHI FLOOD DI SASTE R I N NE PAL Flood Modeling Developing a flood model for the area helps in evaluating maximum elevation of 6,192 m. For calibrating the flood future flood risks and policy planning regarding setback model, we focused on the lower reach (i.e., from Timbu and distances and mitigation measures. At the same time, the downstream to Melamchi Bazar). In contrast, the ADB team drone survey assisted in documenting the flood extent. has focused on the upper reaches (i.e., from Timbu and Besides, a flood model is also required to determine the upstream to Bhemathan). flood depth, velocity, and other parameters. The landslide dam breach mudflow at Melamchi was simulated in The Melamchi watershed was delineated using ArcSWAT Hydrologic Engineering Center’s River Analysis System (ArcSWAT is an ArcGIS extension and interface for SWAT), (HEC-RAS) as a two-dimensional non-Newtonian flow for a and the boundary of the study area was fixed. The land use rapid assessment to reproduce the flood conditions in the was classified using Sentinel-2 satellite images using the model. supervised classification technique. 8 Land use classes were identified - Snow, Built-up, Water, Dense Vegetation, The Melamchi Valley is typically a narrow, steep Himalayan Rock, Barren Land, Sediments, & Sparse Vegetation. The River-Valley. The lower valley has slopes with floodplains rainfall values were collected from the Global Precipitation and is more like a U-shaped valley and are the sites of Measurement (NASA) due to the unavailability of data from settlements. The upper mountain slopes are very steep (1:6), rain gauges. HEC RAS’s full 2D model was selected for the rocky, and pointed sharp ridgelines. Elevation differences study area due to the highly dynamic nature of the flood between the valley floor and surrounding ridges exceed wave. First, a rain-on-grid model was built for the watershed, over 1,000 m in the upper part, and the watershed has a and from this model, the inflow hydrograph for the final Figure 11: Schematic showing the flood modeling methodology. 17 M ELA MCHI FLOOD DI SASTE R I N NE PAL model was extracted. After analyzing the preliminary rain- stability conditions (courant number <=1). Flood hazard, on-grid model, a refined model was constructed for the depth, velocity, and flood extend maps were generated, study area. This refined model was subjected to sensitivity which were subsequently used to analyze the hazard to the analysis. Sensitivity analysis of the input parameters was community. A schematic showing the methodology followed conducted, and the results were compared with the field for the flood modeling is shown in Figure 11. data and drone imagery. After the sensitivity analysis, 40 mm rainfall of duration of 1.5 hr with a high-density mix of Hydrograph extracted from the rain-on-grid HEC RAS model water and debris (10% debris) was selected for the final was constructed using 40mm (GPM NASA), rainfall of non-Newtonian dam breach model. The terrain model was 1.5hr duration. Five inflow hydrographs for the main river created by stacking 5 m DEM with Shuttle Radar Topography tributaries downstream of the landslide dam and two inflow Mission (SRTM) 30 m DEM. When available, the 30 m DEM hydrographs for flow into the dam were used for the final was stacked using 5 m DEM to utilize the higher spatial model. Figure 12 shows an example hydrograph. resolution.The model was further refined to meet the Figure 12: Hydrograph extracted from the Rain-on-Grid model at Melamchi Bazar 18 M ELA MCHI FLOOD DI SASTE R I N NE PAL The results obtained from the HEC RAS modeling were validated against flood boundaries generated from drone surveys, and the photographs during the flood. The maximum depth and velocity map for the entire watershed are shown in Figure 13. Figure 13: Simulated flood depth and velocity using HEC-RAS overlaid on the satellite image The results from the modeled flood compared to the flood This location was used for the modeled flood validation extend mapped from the drone survey is shown in Figure 14. due to the availability of pre- and post-flood photographs It is observed from Figure 14 that majority of the modeled (Figure 15). The modeled estimates of flood depth indicated flood extend matched well with the observation from drone. a flood depth of 11.8m with debris, which indicates a good At site-3, the large variation between the flood extent from correlation between the modeled flood and conditions drone mapping and the modeled flood is a result of the observed on the ground. The observed flood depth was used observed flood extent incorrectly identified the landslide as to back-calculate the peak discharge (which was estimated part of the flood extent. The Chanaute Bridge (in the city at 4,256 m3/s). And as this calculation was based on Chanaute also referred to as “”Red Bridge” “) which is in the pre-flood cross-sections without considering elevated bed lower reach some 10 km upstream of Melamchi Bazar. levels due to sediment deposition during the event itself, the ‘Red bridge’ was destroyed during the Melamchi flood. The estimated peak discharge may have some uncertainty and height of the bridge from terrain (Pre-flood) was about 12 m could be on the higher side. 19 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 14: HEC-RAS modeled flood compared to flood extend mapped from drone survey (red line: observed, blue: modeled) Figure 15: A comparison of the pre- and peak-flood photographs from the Chanaute (Red) Bridge location 20 M ELA MCHI FLOOD DI SASTE R I N NE PAL Based on the simulated result, the areas with sediment rapid topographic change within a flood event was not deposition had an average velocity of 6.5 m/s, whereas considered in the simulation. the areas that experienced erosion had an average velocity Post-flood simulation utilizing the topographic data from of 16.2 m/s. The simulated velocity profile along the river the drone survey was tested but not completed because of (Figure 13) was consistent with the observed erosion and uncertainty in the riverbed elevation underwater when the deposition pattern in the main channel (Figure 5). The results drone survey was conducted. To improve the accuracy of of the HEC-RAS, i.e., flood depth, velocity, and water surface the topographic data prepared by the survey, a trail of the elevation, are generated as a single kmz file that can be hydraulic simulation was conducted as well. For example, easily visualized on Google Earth. wet area during the drone survey was delineated and The flood event was more complex in reality than how it excluded from the profile and transported sediment volume was simulated in the study. The riverbed might have been calculations, and topographic data were closely reviewed gradually increasing during the flood due to the massive and revised based on the unreasonable flow pattern deposition to exacerbate the flooding. In other places, the observed in the simulations. riverbed might have been lowered by erosion, while such 21 M ELA MCHI FLOOD DI SASTE R I N NE PAL Satellite-Based Land Displacement Analysis The vast mountainous terrain and high-altitude conditions periodically monitoring slopes with susceptibility. The free- in Nepal make it difficult and costly to directly apply in-situ of-charge Sentinel-1 SAR satellite data from European Space land displacement monitoring techniques, like installation Agency (ESA) provides even higher cost-effective monitoring of measurement devices as well as frequent and regular opportunities. ground surveys, over the entire region of interest in the upstream area of Melamchi River. While landslides and other Some of the goals for satellite-based land displacement types of slope failures can develop under various conditions mapping in the Melamchi River upstream area were to and are hard to predict, under the right circumstances, the generate time-series land displacement analysis, identify most susceptible slopes can be identified and monitored displacement hotspots area to indicate slopes with through land displacement mapping. Synthetic Aperture susceptibility, and identify possible precursors of the recent Radar (SAR) data and time-series interferometric analysis- disaster event. This task was undertaken by Synspective. based remote sensing techniques offer an alternative to Time-series of Interferometric SAR satellite imageries were Figure 16: Distribution of the slope-projected displacement points categorized unstable in the AOI. 22 M ELA MCHI FLOOD DI SASTE R I N NE PAL used covering the period from 2017 until the last SAR data The southern region of the AOI has vegetation covers that acquisition date before the June 15th, 2021 disaster event challenge Sentinel-1 radar wavelength (5.5cm) to reach the (June 12th, 2021). Due to the influences of seasonal snow, ground. On the other hand, displacement data density in ice melting, vegetation covers, and other SAR decorrelation Pemdan Khola (area of Glacial Lake) and Bhemathan old factors, time-series InSAR analysis is split into several landslide dam can be deemed sufficient. datasets (4 datasets) and skipping certain periods of data to avoid the inclusion of SAR data pairs with severe To identify slopes susceptible to landslide displacement, decorrelation. hotspots were determined from the SAR data. Displacement hotspots are locations with local displacement velocity Time-series InSAR analysis from Interferometric Point Target higher than regional displacement velocity. Changes in Analysis (IPTA) generated a data file containing a large the displacement time-series trend that reflect sudden/ number (up to a hundred thousand) of land displacement unprecedented displacements were taken as a proxy for data points with uneven distribution depending on ground displacement precursor. These conceptions are chosen characteristics (vegetated ground, ground experiencing under the presumption that a trigger factor can cause large physical change, seasonal change, will have less deviation inland displacement’s temporal pattern, which displacement data points density). might lead to near-future slope failure risks. Time-series InSAR analysis also generated 375,912 Line Based on the hotspot analysis, 163 hotspot zones were of Sight (LOS) displacement data points over the 403 km2 identified in the AOI. An ID number was assigned to each extent of the Area of Interest (AOI). LOS displacement hotspot zone based on the size. A smaller ID number data were then projected to the slope direction resulting signifies a smaller size of the displacement hotspot zone. in 350,332 along-slope displacement data points. The smallest size of displacement hotspot zone is 50.47 m2 Displacement velocity of +/- 15.62mm/year was used as a identified in Yangri Watershed. The biggest displacement threshold for identifying instability (a positive value indicates hotspot zone is found in the Melamchi Watershed region, uplift such as due to gradual sediment deposition, upheaval with a size of 60,000 m2. processes, folding, debris accumulation downstream of alluvial fan, etc., and a negative value indicates downslope For the 163 hotspot zones, class 1-3 intensity of movement or erosion). Using this criterion, 11.4% (39,981) displacement velocity was identified. Table 1 summarizes of displacement data points were categorized as unstable. the count of displacement hotspot zones in each intensity Figure 16 describes the distribution of these data points. Out class of displacement velocity for the 3 watersheds of the of these unstable points, 23,431 points indicate downslope AOI. The largest number of hotspot zone with high-velocity moving displacement with an average velocity of -31.8mm/ displacement is located in Melamchi Watershed. Compared year, and 16,560 points indicate uplift displacement with an to other watersheds, data also indicates that the downslope average velocity of 31.27mm/year. moving displacement hotspot zone in Melamchi Watershed has, on average, higher velocity than in other watersheds. Both the active downslope or uplift moving displacement Plotting over a map, the distribution of the 163 hotspot mainly occurs at higher altitudes (average 4,200 – 4,300 m zones with their intensity class of displacement velocity is asl). In general, displacement data density in the northern described in Figure 17. The displacement data were also region of the AOI is higher than in the southern region. partially presented online. Table 1: Displacement hotspot distribution summary 23 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 17: Distribution of displacement hotspot velocity intensity class The displacement hotspot zones represent slope locations hotspot zones were identified. Subsequent analysis was with high susceptibility, indicating either an area with carried out by overlaying the location of these displacement downslope moving displacement, or deposition of debris, hotspot zones with the location of ground change during the or bulging slope toe zone. These high priority zones can Melamchi Flood disaster event. The result highlights that the be listed for strategic actions such as mitigation and displacement hotspot zones are co-located with locations of monitoring. ground change in the Glacial Lake area. (Figure 18). These results also point to the possible hypothesis that downslope Further analysis was carried out at the displacement hotspot moving displacement hotspots identified in this analysis zone around the Pemdan Khola region. At the perimeter were a slow-moving type of landslide zone triggered during of the glacial lake in the upper reach of Pemdan Khola, the the Melamchi Flood disaster and became the source of debris presence of both downslope moving and uplift displacement that was transported to the lower altitude. 24 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 18: Labels indicate displacement velocity (mm/yr) of hotspot zones; the red label indicates downward slope displacement and blue uplift. The red-purplish-colored area shows locations with ground change. 25 M ELA MCHI FLOOD DI SASTE R I N NE PAL Melamchi flood disaster field observations The Melamchi flood, is a combined effect of heavy rainfall, process in the upper catchment area where permafrost is temperature change in snow line, erosion in end moraine found in the deglaciated valley. As a result, on 15th of June of Pemdan lake, possible breach of the natural dam end moraine dam of Pemdan lake (4700 m) was possibly responsible for the lake, cascading effects of natural dam eroded (Figures 19a and 19b) and the lake began to be breach along with erosion and series of landslides along the emptied. The Pemdan Khola faced a flash flood, and a Melamchi River. The NDRRMA together with the ADB expert massive amount of boulders, gravel, and sand accumulated team conducted several field visits including one using in the Bhemathan area. Bhemathan is an old landslide a helicopter. The following section summarizes the field dam, and the reservoir was filled with sediment. It was well observations provided by Dr. Ranjan Kumar Dahal, a World forested also. On June 15, 2021, the trees of Bhemathan were Bank and Asian Development Bank expert who participated also mixed with the debris flow of Pemdan Khola. As a result, in the field visits. the old landslide area of Bhemathan was blocked, and flood water with debris started to flow as overtopping flow through Pemdan lake outwash and the old landslide dam. Again, on August 1, 2021, heavy rainfall occurred. The overtopping water flow at the Bhemathan area Bhemathan debris flow eroded the old landslide dam abruptly (Figure 19c). A massive The flood events of 2021 in the Melamchi River can be flash flood started to erode downstream where old glacial traced out from the upper catchment of the Melamchi River. deposits and river channel deposits were abundant in more On June 14-15, 2021, heavy rainfall enhanced the erosion than 4 km stretch of the river (19d). Figure 19: Events from Pemdan Lake to Bhemathan in June 15 and August 1, 2021. 26 M ELA MCHI FLOOD DI SASTE R I N NE PAL Temporal damming of Melamchi hours on June 15, 2021 (Figure 20). On June 15, 2021, hydrological data of Melamchi River at Nakote recorded this Ghyang landslide damming (lowering of water) and rising of water flow, and The Melamchi Ghyang landslide blocked the river for a few the gauge was washed away in this event. Figure 20: Damming of Melamchi River down to Melamchi Ghyang village. From field observation, it is understood that this landslide blocked both the overtopping flood of Bhemathan on June 15 and potentially the flood on August 1, 2021 as well. Extreme erosion in downstream from the Melamchi Ghyang landslide damming, the downstream of the Melamchi River was intensively eroded channel and deeply incised, and a bridge was washed away in Due to the sudden release of a massive amount of water Nakote (Figure 21). Figure 21: Erosion into bedrock level at Nakote, site of immediate downstream from Melamchi Ghyang landslide. Only 16 days old newly constructed bridge was also washed away during the flood. 27 M ELA MCHI FLOOD DI SASTE R I N NE PAL Debris flow through the narrow this area overall. Several massive landslides have also occurred in this area. The narrow gorge just upstream of the gorge and buried headworks of the headworks is the endpoint of the erosional channel of the Melamchi Water Supply Project Melamchi River in the 2021 flood. From the headworks to downstream, the channel experienced massive depositions. The headworks area of the Melamchi Water Supply Project, As a result, all the downstream settlements were partly with financial support from ADB, experienced severe buried. The effects were well observed downstream to the damage, and more than 16 m of debris accumulated in confluence of the Indrawati and Melamchi rivers (Figure 22). Figure 22: Landslide in headworks area, deep gorge and headworks area under cleaning. Damage in Gyalthum to The third location is the bridge at Phatte, which temporarily blocked the sediment. The thick sediment accumulated due Talamarang area to the temporary ponding. At this location, downstream to Massive sediment and debris deposition was observed the bridge area consists of mainly fine sediments, which at four (4) locations while traveling from Talamarag to buried cultivated lands and permanent buildings and Melamchi Bazar. The first location is the gorge area (Figure buildings under construction. 23) of Bhattar Phat, where the remnant of the old chain suspension bridge was buried during the 2021 flood event. Similarly, the fourth was at the confluence of the Indrawati In this narrow channel, the upstream area up to Gyalthum and Melamchi Pul Bazar areas. The sediments with huge of Helambu Rural Municipality, the sediments accumulated, boulders could not pass through a narrow channel below the huge boulders were deposited in the riverbed and buried 2nd bridge, due to which debris reservoir was formed, and fine level terraces (Figure 23). The second location was at the sediments deviated towards the Pul Bazar area (Figure 25). Thulophat area, where many huge boulders, eroded from the As a result, more than 16 m thick sediments were deposited narrow upstream channel, were accumulated (Figure 24). in the area (Figure 26). 28 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 23: Four locations (Talamarang to Melamchi Bazar section) of sediment blockage during the 2021 flood in the Melamchi River. These locations highly controlled the sedimentation process. Location 1 is Talamarang, location 2 is Thulophat, location 3 is Phatte, and location 4 is Melamchi River. Figure 24: Significant deposition of debris occurred at Phatte due to a slowdown of flow caused by a downstream bridge as a blockage. Only fine sediments were found deposited downstream of the bridge. 29 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 25: Before and after view of damaged Melamchi Bazar. The Pul Bazar was the major obstacle, and more than 16 m of sediment was accumulated in the Melamchi Bazar. Figure 26: Many buildings in Melamchi Town was covered by the debris. 30 M ELA MCHI FLOOD DI SASTE R I N NE PAL Conclusions and Lessons Learned After the Melamchi flood in June 2021, three studies were The sensitivity analysis using the HEC-RAS model identified implemented in parallel on a tight schedule based on the that the landslide dam breach caused a significant elevation request from and close discussions with the NDRRMA. in the flood depth and resulted in inundating more areas The implementation period of the three studies overlapped of the Melamchi Watershed and surrounding places. This each other with regular progress meetings with the observation further emphasizes the need to characterize and government stakeholders, and all the three teams were monitor the landslide hazard in the region using technologies engaged. These studies are crucial in understanding the such as the satellite-based land displacement analysis so flood more holistically and enhanced the quality of each that the risk of such future events can be avoided. Comparing study. For instance, utilizing the topographic data prepared the clear water simulation and the non-/Newtonian (mud/ by the drone team, the flood modeling team identified debris-flow) flows showed that utilizing non-Newtonian key issues and helped modify the data. Using photos and conditions to model such events involving large loads of videos captured by the drone team, coupled with field sediments better reproduced the flood event in the model. observations by the experts, enabled the satellite team to The presence of the debris reduced the flow velocity resulting orient their focus and improve analysis quality. In addition, in higher flood depths. The drone data was valuable for the the World Bank, together with the NDRRMA organized calibration and validation of the flood simulation. Particularly the flood extents mapped from the drone were used to verify two-hybrid workshops presenting the study results to policy the accuracy of the HEC-RAS model. The comparison of stakeholders. The final workshop was attended by more the modeled flood velocity with the erosion and deposition than 71 participants from various government agencies and regions identified from the drone imagery showed a good external stakeholders such as ICIMOD and NASA for sharing correlation. The deposition areas had an average flood the findings and promoting discussions. velocity of 6.5 m/s, whereas the erosion areas had an The three activities were critical to understanding the average velocity of 16.2m/s. The results indicate that the hazard and risks of the Melamchi Watershed. Notably, flood velocity simulation is critical for designing mitigation the drone-based survey provided timely documentation measures and ensuring they would sustain the erosion and of the damages from the event. Such documentation is deposition mechanisms in the region. The results of the HEC- valuable to calibrate and validate models and ensure that RAS simulation were distributed to the interested parties as the regional mitigation measures are sustainable. The drone Google Earth files for future risk reduction and visualization, survey provided a georeferenced ortho-mosaic and Digital making it readily accessible for researchers, policy planners, Terrain Model (DTM) of the post-disaster condition. The and any relevant stakeholders to plan setback distances and ortho-mosaic helped identify flood extent, landslides, and other relevant policies for community development. damages to houses and infrastructure. The DTM generated The application of SAR satellite data and land displacement from the drone survey can be used for approximate volume analysis emphasizes 1) the power of SAR technology and calculations of debris and landslide deposits and cross- its benefit to disaster risk managers and other authorities, section developments. The drone survey results indicate and 2) the adoption of the technology itself in disaster risk that cadastral maps can be produced quickly and easily in management practice. The role of SAR satellite data and complex and difficult-to-access environments. However, it analysis results is most suitable in circumstances in which it was observed that the drone imagery does not penetrate the is challenging to obtain the onsite measurement in a timely/ water surface. Hence, any deposition or erosion quantification regular manner due to the hindrance of cost and location below the water level will be inaccurate from the drone accessibility challenges. The adoption of SAR technology imagery. It is important to note that the accuracy of the DTM and analysis in disaster risk management practice can be can be limited (+/- 5m) in complex terrain when only limited considered following the conceptual scheme illustrated in ground control points can be established with high precision. Figure 27. 31 M ELA MCHI FLOOD DI SASTE R I N NE PAL Figure 27: Conceptual scheme for SAR technology adoption While this study highlights the capability of the SAR Contribution to detailed mitigation technology in providing valuable insights in slope instability- related risk management works, however, this process planning studies also has limitations that need to be acknowledged. All input and output data used for generated through the The analysis technique itself requires expertise in studies were shared with the NDRRMA and their partners, executing and interpreting the analysis generated from including the ADB expert team undertaking a detailed the data processing. In many cases, it cannot be done damage assessment of the Melamchi Headworks and straightforwardly, thus resulting in some difficulties. Hazard Mapping of the Catchment for the MRWDS project Especially in regions with geographic characteristics like and the University team on flood zoning in Melamchi. the AOI in this assignment because of thick forest and Melamchi Municipality has prepared preliminary plans of snow covers, the level of difficulty in both image analysis recovery and mitigation on the basis of Drone Survey data processing and interpretation can even be higher. Post- for this work. They utilized the topographic data prepared by processing additional analysis based on field observations, the drone team and reviewed the 2D flood model setting and like the ones demonstrated in this work, needs to be done to parameters to build their detailed hydraulic modeling. The improve insight drawing from the data and analysis results. ADB team also requested detailed InSAR land displacement With the advancement in the field of artificial intelligence data for the slope near the headworks to better understand (AI) and data science technology, complexity and hurdles in the remaining risk of a large landslide nearby. Such close the analysis can be minimized by automating the process collaboration on sharing data and findings with the Gov and delivering the result as user friendly as possible so and partners maximized the use of limited resources for that the information can be quickly digested and used understanding the disaster and remaining risks holistically, as supporting material in the strategic decision-making and helped improve the quality of deliverables. process. Considering the advancements in this area and the Overall, this urgent support to NDRRMA through various encouraging potential of this work seen in the initial results suggest the possible use of this technology for monitoring studies proved helpful for the government agencies, critical slopes. The development of the InSAR analysis in this partners, and stakeholders to understand the disaster better work gives broad situational awareness regarding instability and provide a basis to prepare risk mitigation plans in the in the Bhemathan and the other critical locations impacted Melamchi watershed. by the Melamchi flood and highlighted within this report 32 M ELA MCHI FLOOD DI SASTE R I N NE PAL Application of this effort to other areas data can be prepared using the drone for high-risk river channels as a critical input for hydraulic modeling. Besides, This effort demonstrated the value of drone-based the drone survey data can help develop a detailed inventory mapping, flood analysis, and InSAR analysis at a location of houses and infrastructure at different risk levels and significantly impacted by flooding and landslides. It shows quantify risk scenarios. the importance of drone-based data collection immediately The pre-event InSAR analysis can be invaluable in identifying after such an event in documenting the flood and developing vulnerable slopes and planning development measures. post-flood topography. The post-flood topography is Often these development measures could be to avoid critical to flood modeling, understanding the revised flood susceptible areas and route development through more hazard in the area, and assisting the InSAR-based ground stable terrain, which could lead to significant cost savings displacement analysis in identifying the site conditions in maintenance and improve the reliability and sustainability and validating the results. The methodology can be applied of the infrastructure. Even when the project doesn’t have the pre-event to other watersheds with high risk of such debris flexibility to develop through stable terrain, the InSAR-based flow for hazard characterization. risk reduction, hazard analysis can help identify the most vulnerable locations and management, and policy formulation. Pre-event drone build monitoring and mitigation measures. Such measures surveys can be invaluable in developing topographic data for can reduce human casualty and increase the reliability of the high-risk channels to use in hydraulic modeling. Topographic development. 33 EUROPEAN UNION