1. Introduction
Climate change has increased the threat of global flooding hazards, as indicated by the rising frequency and greater spatial coverage during the last decade compared with those observed in previous decades [
1]. Global warming is generally expected to increase the magnitude and frequency of extreme rainfall events, thus potentially impacting riverine hazards [
2]. Furthermore, increases in air temperature and heavy rainfall on a regional scale also influence flooding frequency and intensity [
3]. Floods often cause enormous damage to lives, property, crops, and infrastructure, and the number of casualties by floods can increase in the future [
4,
5,
6].
Monsoonal floods are a common hazard in Nepal, where glacial lake outburst floods (GLOFs) also threaten human safety and wellbeing [
5]. GLOFs can cause catastrophic flooding in downstream areas, severely damaging property in addition to causing loss of life [
7]. Melting glaciers contribute to the development of glacial lakes, which may cause lake outbursts. One-fifth of the world’s population depends on water derived from the Hindu Kush Himalayan region [
8]. Minimizing loss of life and property damage is crucial and requires a better understanding of GLOFs and their downstream effects [
9]. However, studies on GLOF events, moraine dam failure, peak discharge measurement, and their downstream flooding are limited.
Unmanned aerial vehicles (UAVs) are a helpful tool for monitoring and investigating disasters since they can source data from otherwise inaccessible areas. Many recent studies are based on UAV applications for support and assessment of disaster management [
10,
11,
12,
13,
14,
15,
16,
17]. Furthermore, UAV platforms are currently considered valuable sources of data for inspection, surveillance, mapping, 3D modeling, and the creation of evacuation routes [
14,
18,
19]. Most UAV applications in hydrological research are used for stream and riverscape flood mapping, all of which require continuous monitoring [
11,
14]. In particular, UAVs can assist with monitoring the state of dams, disaster tracking [
20], provide evacuation guidance, and facilitate damage investigations, bank erosion and volcano observation, flood-hazard modeling, and industrial disasters [
10,
14,
16,
21,
22,
23,
24]. Accordingly, UAVs can generate an evacuation map over a relatively short period. However, no research has been conducted on GLOFs and flash floods in the Himalayan region and downstream communities.
As unpreventable urban flood events have been increasing with the advancement of technology, flood-vulnerable areas can be identified through 1D/2D modeling to minimize losses of human life and property [
25,
26]. The Hydrologic Engineering Center River Analysis System (HEC-RAS) model developed by the U.S. Army Corps of Engineers has been widely employed in studies of risk assessment, hazard risk management, and emergency response guidelines in flood-affected areas [
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38,
39]. HEC-RAS creates one-dimensional (1D) and two-dimensional (2D) steady and unsteady flow simulations [
32], and its joint use with GIS is essential to risk reduction programs [
33].
Flood impacts are further exacerbated by increasing urbanization [
40,
41]. This is especially prevalent in developing countries where populations, economic activities, and housing are often concentrated in less desirable flood-prone areas [
36]. Various inaccuracies hamper public understanding; for instance, if a major flood disaster is known to generally occur once in a particular period, this leads to a false sense of security that no significant floods will occur for some time, which then carries over to construction planning for dams and other structures in the surrounding area. Moreover, the low-risk awareness of residents living in flood-prone areas is usually considered one of the main causes of low preparedness, which in turn generates inadequate responses [
35]. Thus, understanding disaster risk reduction, awareness, preparedness, and adaptation is a challenge [
42,
43,
44,
45].
Notably, a flash flood occurred on the Seti River in Pokhara, central Nepal, on 5 May 2012. The flood killed 32 people, left 40 missing (presumed dead), and displaced community buildings and vital infrastructures such as suspension bridges, electric poles, and potable water pipelines [
46]. The discharge rate of the flood at a nearby dam site was estimated to be 1450 m
3·s
−1 [
47]. Since then, many studies have examined the actual causal mechanisms of the flood, which have often been attributed to a fall of ~15 million m
3 of rock onto glacier ice at the slope bottom of the Annapurna IV peak [
48]. Other researchers suggested that a rockfall blockage of the Seti River Gorge occurred a few weeks before the disaster [
46,
47,
48,
49,
50,
51,
52,
53,
54,
55] and that a rock/ice avalanche from the northern part of Annapurna IV (~7500 m) dislodged the previous rockfall dam when being swept into the reservoir; however, studies focused on downstream communities are absent.
Socioeconomic vulnerability assessment is another critical component of comprehensive flood risk assessment, which aims to estimate the loss of life and economic damage [
56]. Social vulnerability highlights differences in the capacity to prepare for, respond to, and recover from disasters related to time, social groups, and demographic characteristics [
57]. The International Centre for Integrated Mountain Development (ICIMOD) researched the socioeconomic vulnerability of communities potentially affected by GLOFs [
58]. In particular, Nepal has experienced multi-hazard risks and socioeconomic losses in the recent decade [
59]. Furthermore, riverine floods are one of the major weather-related disasters in Nepal, affecting agriculture production and physical infrastructure [
60].
Recently, some researchers have focused on either modeling or social approaches to disaster vulnerability. For example, social scientists have identified many ways in which people respond to rapid onset threats from earthquakes, flash floods, tornadoes, tsunamis, and volcanic eruptions (e.g., [
61,
62,
63,
64]). A community-based early alarming system would be more beneficial to be introduced because an effective response to the early warnings greatly reduces casualties and loss of life [
65]. Some researchers have focused on tsunami caused flash floods and investigated pre-impact hazard communication, immediate post impact expectation, household’s evacuation, public risk perception, and behavioral intentions, [
66,
67,
68].
However, thus far, the two have not been interconnected [
69,
70,
71,
72,
73,
74]. Researchers have examined socioeconomic issues using community-based early warning systems and adaptive capacity-building strategies [
75]. Other studies focused on social vulnerability, disaster risk reduction knowledge, preparedness, and adaptation and awareness programs for downstream communities [
44,
76,
77,
78]. Research has also focused on the policy-making process and challenges of implementation [
79,
80]. While several studies focus on dealing with natural disasters, modeling, and social aspects separately, no study has combined these three aspects. This research identifies vulnerable houses, addresses social issues, and determines safe evacuation routes. We, therefore, believe that our suggested route can save lives in an emergency. This type of combined research is very important for a country such as Nepal. The findings of this research can be used as a decision-making tool to support the local government in long-term planning programs. This research is, therefore, different from previous studies.
The present study aimed to: (1) estimate the probability of flood inundation using HEC-RAS to identify vulnerable locations and houses; (2) create an evacuation route map for emergency use; and (3) analyze the relationship between the socioeconomic status of residents and potential flooding in their area. Three major settlements from Masinabagar to KI-sing were examined (
Figure 1) for their close proximity to the river and high population density.
2. Materials and Methods
2.1. Study Area
Pokhara (
Figure 1) is situated approximately 200 km west of the capital city of Kathmandu and is the second-largest city in Nepal with ~265,000 inhabitants. It is located at an elevational range of 812 m to 1511 m. The Seti River originates from the southern flank of Annapurna Himal. It flows down to the Lesser Himalaya along the steep and narrow Pokhara Valley, which was formed by at least two giant debris-flow events in the past [
81,
82].
The study area is in the northern part of Pokhara City (Pokhara Lekhnath Metropolis), which is categorized as a high flood risk zone. It is a marginal area close to the populated city center. Recently, the local government introduced a city planning act in Pokhara that strictly prohibited the construction of new houses within 10 m of the riverside; however, no plan exists for those already living in this area. Furthermore, there has been a rapid influx of migrants and the impoverished, thereby pushing legal boundaries to their limits in settlements such as Masinabagar, Laltinbazar, and KI-sing (
Figure 1). Thus, vital urban planning and management policies are needed to prevent further damage. Accordingly, three methods have been employed in this study: (1) inundation route mapping, (2) evacuation route mapping, and (3) a socioeconomic survey (
Figure 2), which will be detailed in the following sections.
2.2. UAV (Drone) Photogrammetry
In the first stage of the drone-based aerial survey (DJI Phantom 4 Pro), the ground control point (GCP) locations were finalized. The flight route was planned using Google Earth (v.2.3) to help determine the overall flight number, height, estimated time, and image percentage overlap. GCPs must be visible from the nadir view of the drone and evenly distributed to achieve a uniformly high accuracy throughout the study area. After the 21 GCP locations were identified and finalized, they were marked with a 60 × 60-cm red flag and white enamel cross to assist with drone image identification. The center of the white cross was marked as the GCP, and the kml file of the site was imported into the Drone Deploy software. The flight plan was set for a front overlap of 75% and a side overlap of 70% at the height of 55 m. A total of 25 flights were flown.
A differential GPS survey (Stonex DGPS) of the established GCPs was carried out to identify their proper position. The easting, northing, and elevation of the GCPs were also measured, and 21 points were utilized for 3D georeferencing. After the geotagged images were captured onsite, they were processed to obtain a digital surface model (DSM), contour lines, a digital terrain model (DTM), and orthophoto maps. A DEM is a bare-earth raster where non-ground (man-fabricated) features such as roads and buildings are not included.
DEMs are useful for hydrological modeling, surface analysis, and soil mapping. A DTM is a 3D model visualizing surface elevation data; its structure is based on a TIN composed of vector data. A Triangulated Irregular Network (TIN) represents the terrain surface as a set of interconnected triangular facets. The TIN structure is a vector-based alternative to the traditional raster representation of the terrain surface Digital Elevation Model (DEM). A DTM reinforces a DEM by including man-fabricated features of the bare-earth terrain. Irrelevant images were removed before processing with Agisoft Photoscan (v.1.4.0). The overlapped and geotagged images were processed using image matching algorithms, i.e., the Scale Invariant Feature Transform (SIFT) algorithm. The SIFT algorithm is used to detect and describe local features in digital images. The output data from the initial image processing represent the tie points that were initially matched by the algorithm. These tie points were generated by matching the same features within the overlapped images. Subsequently, point cloud and mesh processing were carried out. The dense stereo matching algorithm generates 3D-textured point clouds to replicate the ground surface. These points were triangulated to obtain the 3D mesh of the project area. The output accuracy of the DSM was modified according to x·Ground Sampling Distance (GSD), where GSD is the distance between the centers of two adjacent pixels in the photo and x is a multiplier. The DTM was generated as per requirement. It is a mathematical model of the ground surface, most often in the form of a regular grid, in which a unique elevation value is assigned to each pixel. The contours derived from the DTM were used to prepare the topographic map. Major and minor contours of 5 and 1 m were used for this research, respectively. A georeferenced orthophoto was also generated (
Figure 1). Houses, roads, rivers, and other features were digitized and integrated with the contours to prepare the final topographic map of the project area using ArcGIS (v.10.2).
2.3. Inundation Mapping
2.3.1. HEC-RAS Model
The HEC-RAS model used to calculate the inundation area has a corresponding extension in ArcGIS (HEC-RAS v.4.1.0). Google Earth (GE) was used for the geographic coordinate system with the WGS84 datum. One of the main inputs for the HEC-RAS was a Triangulated Irregular Network (TIN), a type of DTM. A flood inundation map was prepared using both the topographic maps from the drone survey and the DGPS survey at 1-m contour intervals. A DEM map was also prepared at a 1-m contour interval to delineate watershed boundaries. Manning’s Roughness Coefficient (MRC) was employed for different land use types, which can be expressed according to Equation (1):
where
Q is the discharge (m
3·s
−1),
A is the cross-sectional area (m
2),
R is the hydraulic radius (m),
S is the friction slope (°), and
n is the MRC [
83].
2.3.2. Calibration and Model Validation
The channel model was calibrated by adjusting the MRC until the differences between the simulated and observed water levels at channel stations were plotted [
39,
71,
84,
85,
86]. The observed standard discharge data collected by the Department of Hydrology and Meteorology of Nepal (unpublished) from May 2011 were used as the input. The MRC was iteratively changed within the minimum and maximum values described in [
83]. The water surface results in every cross-section from each model were compared with the observed water surface measurements. The MRC value that led to the least discrepancy between simulated and observed values was taken as the calibrated value for the channel (
Figure 3). The simulated high flood level was checked against the discharge volumes observed at peak flood levels during the 2012 event for model validation. The selected MRC values were 0.02 (adjusted) for overbanks and 0.035 for the main channel [
83]. For inundation mapping, the flood level (peak discharge) was assumed to be the same as that observed on 5 May 2012, i.e., 1450 m
3·s
−1 [
49]. Notably, the measured MRC (n = 0.02) was identical to the modeled value derived by [
83].
2.4. Evacuation Route Mapping
2.4.1. Extracting an Existing Road Network
This study extracted the road network (for pedestrians) of three settlements using data from an orthophoto digital map using ArcGIS 10. First, the length and width of each road were obtained and measured from a digital map. Next, the road width and length at some sites were measured again in the field to confirm the data from the digital map. Afterward, the number of people able to use the existing road at once during a flood hazard was determined. Finally, a new evacuation route was suggested, depending on the existing topography and population.
2.4.2. Suggesting a New Route
The evacuation route for pedestrians was based on UAV photogrammetry and the shortest route with respect to the existing road network. During the field survey, local people were asked about the nearest safe site during flood occurrence. At first, existing road conditions were inspected and measured manually in all three settlements (Masinabagar, Laltinbazar, and KI-sing) to identify all possible routes for assisting with the traffic flow from the origin (disaster area) to the destination (safe area). The inundation risk area was initially identified. Afterward, the disaster vulnerability areas that needed evacuation plans were identified. Then, the safe place and exit route were located using the drone photos obtained via ArcGIS. The shortest and easiest route from each house (origin) to a safe location was determined.
2.5. Socioeconomic Survey
Primary and secondary information was used to prepare community profiles and assess vulnerability for the three settlements. Published and unpublished documents containing relevant information were collected and reviewed. The questionnaire survey covered all homes within 100 m of the river area in Masinabagar, Laltinbazar, and KI-sing. Basic income information, migration status, family size, education status, knowledge of hazard risks, evacuation sites, primary occupations of the head of households, and house type (permanent or temporary) were collected from the heads of the households. Migrant or non-migrant statuses were based on a 20-year residency period. Such a survey was necessary for Masinabagar and Laltinbazar, where many families had immigrated from different areas in recent years (
Figure 1a). Per capita incomes were categorized based partially on a United Nations criterion [
87,
88], in which low income (i.e., extreme poverty) was defined as an annual income of ≤1025 USD, lower-middle incomes were those with a per capita gross national income between 1026 USD and 3995 USD, upper-middle incomes were between 3996 USD and 12,375 USD, and high-incomes were those >12,375 USD.
Focus group discussions in two settlements of Masinabagar and Laltinbazar were conducted in November 2015 to collect information on community-level problems faced after the 2012 flood, infrastructure development and other related activities, institutions, hazards, risk evacuation sites, and refugees. Lecturers from Prithivi Narayen Campus, Pokhara, a local primary school teacher, and NGO staff helped as facilitators. There were 12 participants in Laltinbazar and 43 in Masinabagar. The discussions ranged from 1 to 3 h.
Furthermore, teachers, administrative officers, local political leaders, NGOs, international non-governmental organization (INGO) staff, and FM radio employees were interviewed to collect information on flood disaster challenges and future disaster mitigation efforts. These members actively participated in the 2012 relief distribution program. A checklist was prepared to record the information gained through these interviews.
For the statistical analysis, we used the interquartile range (IQR). The IQR is a measurement of the variability of the median. More specifically, the IQR tells us the range of the middle half of the data. Another method we used for this study is the Kruskal–Wallis test. It is a nonparametric approach to one-way ANOVA. The procedure is used to determine if there are statistically significant differences between two or more groups of an independent variable on a continuous or ordinal dependent variable. We used a chi-square test to compare observed results with expected results. A low value of chi-square implies that there is a high correlation between the two sets of data.
All statistical analyses were completed in SPSS v.17. Interquartile ranges (IQRs), Kruskal–Wallis tests, and chi-squared tests were used to compare the inundation values and interview data. Data are presented as medians and IQRs, and p < 0.05 (two-tailed) was considered significant in all analyses.
4. Discussion
4.1. Inundation and Evacuation Routes
The inundation maps (
Figure 4,
Figure 5 and
Figure 6) revealed that many houses are situated near the riverbank and in the estimated inundation zone. The situation was similar in all three settlements (
Table 5). In Masinabagar, Laltinbazar, and KI-sing, 268 (57.5%), 232 (70.5%), and 68 (32.4%) houses are located in the inundation zone. Notably, more houses fall under the at-risk category in Masinabagar and Laltinbazar.
Despite such high rates of at-risk houses in the three settlements, the current routes that can be used for evacuation are narrow, located on steep slopes, and their number was limited, as described earlier. Therefore, increasing the evacuation routes is strongly suggested as shown in
Figure 7,
Figure 9 and
Figure 10. The population density of these areas is relatively high, and thus, adequate evacuation routes must be prioritized for the safety of these >2000 residents. These residents in the three settlements represent the nighttime population. However, none of the evacuation routes have streetlights, which makes evacuation extremely difficult at night. Installing streetlights on the routes is strongly suggested.
Regarding daytime evacuation, all three areas contained primary schools. In Masinabagar and Laltinbazar, the schools are entirely in the inundation zone. In KI-sing, many temples and community centers containing children and the elderly are located in the inundation zone. If a flood were to occur, the current evacuation routes are insufficient for these at-risk age groups to use. Thus, for all three areas, establishing adequate evacuation routes and relocating these schools, temples, and community centers must be prioritized among other short-term and long-term strategies.
4.2. Socioeconomic Survey
Based on the results of hydraulic modeling and socioeconomic survey, the statistical relations between potentially inundated houses, safe houses, and residents’ socioeconomic factors were examined via correlation coefficient analyses (
Table 6). Results showed that highly impoverished and immigrant households were at the highest risk in terms of income and migration (
p < 0.001). Similarly, 455 laborers’ houses were also significantly correlated with inundation risk (
p < 0.001).
Table 7 shows the relationship between per capita income and horizontal distance to the riverbank (medians and IQRs). In Masinabagar, extremely impoverished and low-income households existed closer to the river (medians = 31.8 and 29.6, respectively) compared with lower-middle-income family houses (
p < 0.64). In Laltinbazar, the lower-middle group was the nearest to the river. The low-income and extremely poor groups were the second and third closest to the river, respectively (although all three groups lived in houses significantly close to the river bank;
p < 0.004). The situation in KI-sing was similar (
p < 0.001). Thus, it was concluded that low-income family houses exist significantly closer to the river across the entire study area (
p < 0.046).
The interview surveys and focus group discussion suggest that migrants came to Pokhara for a better future. However, owing to their relative lack of skills and money, they were forced into the less desirable riverside area, even though most knew that living in the area was illegal (public property). In Nepal, it is not permitted to build a house for personal use on public land. Public land is land allocated by the general public for paths, ponds, fountains, wells and their banks, grazing land, graveyards, public inns, temples, places for religious practice, memorials, courtyards, sewerage networks, market places, public entertainment, and sporting grounds [
89].
Table 8 shows the relationship between household income, migration status, and occupation of the people interviewed. In Masinabagar, 288 of the 295 highly impoverished families were migrants. In Laltinbazar, 236 families existed in extreme poverty, all of which were migrant laborers. In KI-sing, 63 of the 85 families were migrants in extreme poverty, and 71 were laborers. The residents of all five upper-middle-income families in KI-sing received remittances. In total, labor was the primary occupation of 592 extremely impoverished families. Moreover, of 616 families, 587 were migrants, and for 140 households, the major source of income was business. Correlation coefficients were computed among the three settlements based on income, migration status, and occupation (
Table 8). The results suggested that the correlation between income and migration was statistically significant for all households (
p < 0.001). Similarly, the correlation between income and occupation was also significant (
p < 0.001).
Focus group discussions showed that very few people were aware of local disaster risks or could identify safe sites near their houses after the occurrence of the 2012 flood. Indeed, the majority of residents were still not aware of any safe sites near their settlements. In total, 70% of the inhabitants of Masinabagar, 85% in Laltinbazar, and 56% in KI-sing did not know about safe sites near their settlements. As a result of the exposure to radio and other awareness-raising programs, some people from Masinabagar (29%), Laltinbazar (45%), and KI-sing (43%) could identify safe destinations and quickly travel to these places if evacuation routes are constructed. This research identified evacuation routes, and we believe that such routes can save lives when a disaster happens. Evacuation planning is an essential aspect of today’s growing initiative for early warning systems in the study area as well as in Nepal Himalaya. A long-term strategy must be established to increase evacuation preparedness during future flooding disasters [
75,
90].
4.3. Proposed Evacuation System
This study designed an evacuation system to focus on immediate relief (short-term) and considered the most pertinent factors for alleviating flooding vulnerability to reduce the risk to life and create safe and organized evacuation routes (long-term). Based on research, temporary safe places were identified in the proximity of settlements, even though they do not at present, contain any form of shelters. Land that is outside the inundation area and at a higher altitude was assumed a temporary safe place. The evacuation route was also designed based on that.
Evacuation planning is an essential part of any disaster response strategy in vulnerable areas [
91]. Initially, flooding risk awareness must be addressed, in addition to the actions required for safety. Local authorities and government agencies should provide the public with proper preparedness and prevention training. Residents must be aware and well-trained, including evacuation drill practice, for disaster preparedness. Abarquez and Murshed (2004) stated that preparedness is the ability of the community to avoid the negative impacts of an impending hazard [
92]. In the following sub-sections, we discuss the proposed flooding risk alleviation system in the study area (
Figure 11).
Proper planning for evacuations is an important part of keeping communities safe from the harm caused by disasters. For that purpose, proper evacuation strategies are essential. Planning for evacuations should occur early and involve local governments and communities. Evacuation planning should factor in key issues such as evacuation routes, shelter facilities, messaging to communities, and risks to safe evacuation. Evacuation plans should be regularly reviewed, where necessary, to ensure that they address all relevant issues. Therefore, we propose and categorize short- and long-term evacuation plans for the local government and people.
In the following sections, we discuss (a) short-term strategies for local people and government and (b) long-term strategies for local government.
4.3.1. Short-Term Strategies for Local People
Emergency Plan and Regular Rehearsal Drills
Evacuation planning is an integral part of emergency plans, particularly for regions vulnerable to disasters [
91,
93]. Evacuation is a complex process consisting of several consecutive phases [
91]. The evacuation decision during a flood primarily represents the household’s choice to evacuate or remain in the risk area [
94]. Although superficially, this may appear a simple decision, in practice, it involves a complex set of behavioral and external factors at a particularly critical time. Thus, when flash floods occur, people must make quick decisions to protect themselves and their families, which is most often achieved in a rapid evacuation to a safe place or shelter house, if available. Therefore, frequent regular drills are indispensable.
Temporary houses or transitional shelter is the most important factor when no other alternatives are available. Shelters provide immediate and short-term stay for the victims, help them recover from the trauma of a disaster, and provide a base to start the process of rehabilitation [
95,
96].
Identifying Evacuation Routes, Preparing for Emergency Evacuation, and Essential Kits
Low-risk awareness among the residents living in flood-prone areas is usually considered one of the leading causes of low preparedness, which generates an inadequate response to disasters [
97]. This step of spreading awareness should be implemented well-before the onset of floods, and locals must be educated about preparedness. Focus group discussions and interview surveys suggested that knowledge on preparedness was nearly zero in the study area. Authorized persons should provide background information regarding risk mitigation and evacuation techniques to help minimize disasters [
98].
Timely identification of a safe evacuation route is essential for successful evacuation. The proposed evacuation maps (
Figure 7,
Figure 9 and
Figure 10) help identify building-specific evacuation routes available within the study area to ensure that people reach the nearest safe destination. It is vital to have broad and alternative routes in case of roadblocks. Prior knowledge of the most appropriate evacuation route can help increase the efficiency of evacuation. Authorities in charge of determining safe places or temporary/permanent shelter locations must aid in minimizing the total evacuation time by aiding evacuees to reach their choice of destination via the corresponding route, as pointed out by [
99].
Injuries can happen during disasters. Therefore, it is important to prepare a basic first-aid kit with items such as sterile gloves, gauze, soap, burn ointment, bandages in a variety of sizes, pain relievers, scissors, and tweezers.
4.3.2. Short-Term Strategies for Local Government
Installation of Weather and Hydrological Stations
Nepal is considered one of the most disaster-prone countries in the world. Moreover, apart from flooding, other natural hazards such as earthquakes and landslides pose a recurrent risk to large sections of the population [
75].
At present, there is one weather station available between Masinabagar and Laltinbazar, which was damaged by the 2012 flood. If water levels rise and flood hazards occur, Masinabagar will be inundated even before getting information. The installation of a weather station in a proper place near the settlements is required. In addition, a hydrological station for monitoring the water level of the river is needed at an appropriate location. These will be helpful to obtain scientific data and for early warning as well. Careful planning is needed to select the location for establishing weather stations.
Flood Forecasting and Early Warning System
An early warning system is vital for disaster risk reduction [
75], and it is important to allow sufficient time for evacuation [
100]. In particular, timely warnings, communication, and decision-making are critical [
101]. An effective community-based early warning system must use local resources and capacities to prepare effectively and respond to flooding events [
75]. In the study area, installing a low-cost warning system for sustainable use is recommended, and it should be connected to the residents’ mobile phones so that they can act quickly. It is essential to understand how people perceive flash flood risks and what influences their response to warning information [
74].
People must also understand the importance of flash flood forecasts and early alerts. Accordingly, local authorities should design and establish an upriver weather station with an interconnected early warning system as soon as possible. Appropriate hydrological forecasts can also be used to predict future water levels [
102]. The underlying spontaneity of extreme flooding risks creates an extremely dynamic challenge [
101]. Appropriate, accurate, and timely communication and an early warning system are essential for saving lives.
Transmission of Early Warning
Flood warning broadcasts from various information sources are very important for individuals to make evacuation decisions [
103]. Once a flood risk has been detected, the early warning system will provide the first information. However, this may not be enough in some cases. In crisis communication, trust in the source of information is crucial; therefore, disaster broadcasts, such as TV and radio, and their messages should be trusted by the community [
104]. Recently, social media (SNS) has been thought to be a more effective method of communication. In many developing countries, SNS has become the most important means of communication for flood warnings. Using SNS through a mobile phone would be the most effective and quick way for the people in the study area.
The Nepal Government has already introduced advanced early alarm systems, e.g., in the area of Imja and Tsho Rolpa Glacial lakes for future GLOF events. However, a simple and cost-effective way should be adopted to transmit information. The early warning system helps reduce economic losses and mitigate the number of injuries and deaths, allowing the downstream communities to protect their lives and properties. Therefore, we suggest the implementation of a mobile phone-based early alarming system, which is cost-effective and easy for all those areas connected to the network coverage.
In Nepal, mobile coverage and internet connection are available even in remote areas. In 2016, approximately 92% of the population in Nepal had access to a 2G mobile-phone network at the minimum [
105]. The World Bank publishes mobile-phone subscription data. Based on that, there were 139.4 mobile-phone handsets available in Nepal in 2018 per 100 people [
106]. The research area is an adjoining part of Pokhara city. There is 100% mobile-phone network coverage. Based on the social survey, 98% of the people had a mobile-phone handset.
Results show that among the 15- to 64-year-olds, 98.61%, 98.80%, and 99% in Masinabagar, Laltinbazar, and KI-sing, respectively, have a mobile phone (
Table 9). Altogether 98.68% of the 15- to 64-year-old people have a mobile phone. Therefore, an early warning system through mobile phones would be the best solution. Children at school can receive assistance from teachers, and their parents can inform them at night. In Masinabagar, Laltinbazar, and KI-sing, all students go to the local community school. Only about one-fourth of the elderly people (>64-years of age) have direct access to the early warning through a mobile phone. They are most likely to receive assistance from their family members at home, especially at night. However, they need assistance from the community when they are alone during the daytime or live alone.
Having a mobile phone is a necessity rather than a luxury even among the poor recently. In Gandaki province (Kaski district lies in Gandaki province), 97.6% of the people in rural areas and 97.7% in city areas have access to mobile phones. Prepaid billing systems and the low-priced mobile phone headsets (less than USD 10) are also major factors that have increased its user base.
There is no postpaid system in Nepal. Most mobile companies offer free Subscriber Identity/Identification Module (SIM) cards, low-cost internet data packages, and low-cost voice mail services. The official Nepal Telecom (official service provider) offers the SIM card for NRS 90 (USD 0.80), which comes with NRS 50 (USD 0.45) credit. In total, 88.3% of the poorest people have mobile phones [
107,
108,
109]. Therefore, using mobile phones for early warning is strongly recommended not only in the study area but also in downstream communities in other parts of the country.
Making Safer and Wider Evacuation Routes
Evacuation is an important preparedness measure in disaster management, and it requires careful modeling and planning [
110,
111]. For safe evacuations, a proper evacuation route is needed. Evacuation route recommendation plays an important role in emergency safety management, especially for natural disasters [
112]. The evacuation route should be short, and routes to different shelters cannot present intersection points either in order to allow continuous traffic flow and reduce potential accidents [
76]. At present, there is a very narrow, extremely bent, and congested route available (
Figure 7,
Figure 8,
Figure 9 and
Figure 10). In the future, evacuation routes should be designed based not only on the shortest traveling distance, but also on the existing population, geomorphological structure, and wider and easy-to-use roads.
Flood Prevention Using River Embankment
Managing and monitoring natural and artificial river levees is crucial to reducing hydrological risks [
113]. In addition to these processes that enable quick evacuation, installing a river embankment is one of the important strategies to prevent and reduce flood risk. It would not be realistic to install the embankment at every place in terms of budget. However, if a river embankment is installed where the population is concentrated, it would lead to a longer evacuation time.
4.3.3. Long-Term Strategies
Evacuations must be planned for safe execution, and those plans should be resourced (anticipating all the resources required to complete a project), implemented, and reviewed. Local governments have an important role in evacuation planning, and they need to consider local capacity and capability to manage the evacuation process. This includes preparing a long-term disaster plan, shelter house, and relocation sites.
Long-Term Disaster Preparedness Plan and Policies
Flood disaster plans in developing countries are mostly reactive, responding to prevailing disaster situations (emergency response and recovery) [
43]. The reactive response should be changed to a proactive response to increase management effectiveness and reduce loss of life and properties. We suggest that the plan should address installing weather stations in the settlements.
Government and law-making agencies must establish new disaster plans for the downstream communities. The local authority should make long-term sustainable plans for vulnerable settlements such as Masinabagar, Laltinbazar, and KI-sing and implement strict rules and regulations to prevent river encroachment.
Prohibition of Construction of New Buildings
Riverside communities often face situations that require creative short and long-term housing plans. Effectively moving survivors into post-disaster housing is a critical step toward long-term recovery. However, most parts of the country do not have disaster housing plans. While some housing programs were federally introduced not at the national level planning for disaster, housing should occur at all levels of government before a disaster strikes. Recently, the local government introduced a city planning act in Pokhara that strictly prohibited the construction of new houses within 10 m of the riverside. However, a lack of enforcement means that these processes have not yet completely stopped. Furthermore, no planning exists for those already living in the riverside area.
Shelter
When flooding occurs, people should be evacuated safely to designated shelters along the optimal routes to minimize serious damages to lives and properties [
114]. An adequate shelter has a significant impact on human survival in the initial stage of a disaster, and that requires more than just a roof for a space to be habitable [
95]. Temporary houses or transitional shelters are the most important when no other alternatives are available. Shelters provide immediate and short-term stay for the victims, help them recover from the trauma of a disaster, and provide a base to start the process of rehabilitation [
95,
96]. Making a shelter and meeting its needs in the pre- and post-disaster stages remain a major challenge for the government. A shelter location may be required for periods that extend for several months or even years after a disaster, but local authorities should prioritize this factor when planning and designing shelters. The government can categorize shelter types, such as emergency, temporary, or permanent shelters [
95]. The government should ensure that emergency shelters are accessible and far away from vulnerable slopes, and allow large numbers of people [
115]. Therefore, the emergency shelters should have adequate space, basic living requirements, and be located in safe areas [
76]. Moreover, there are no shelter houses present in the considered locations, and the local government should consider the design of shelter houses for the future.
Relocation
Relocation implies permanently moving to a new location. Relocation is a process that introduces a newly built area for the displaced community [
116]. Population resettlement is a complex process that can create serious social, economic, and cultural problems for the people involved if it is not implemented properly [
117].
As per our findings, the population in this area is very poor, and therefore, the relocation plan should be free of cost. To help the resettled households adapt to the new environments, the local government should make development plans, which should include various job opportunities [
118]. Therefore, the local government should make a long-term plan for those vulnerable settlements. This research identified many houses, schools, and livestock farms that are within high-risk areas (
Figure 4,
Figure 5 and
Figure 6). Therefore, the local government should take action to relocate these to safe places as soon as possible.
4.4. Significance of the Combined Approach: Modeling and Socioeconomic Survey
Nepal has experienced numerous GLOFs and other water-related disasters in recent years. Thus far, most correlated disaster studies have focused on modeling, while others have adopted a sociological approach. However, these studies have tended to be conducted separately. The present study showed the relationship between the locations of individual buildings and houses, with or without flood-hazard dangers, and the socioeconomic status of these residents. It was revealed that households with the most significant inundation risk did not have access to adequate evacuation routes, nor was there a safe shelter house if a disaster occurred. Based on these findings, it was concluded that the buildings within the inundation risk zone must be relocated; however, this requires the financial support of governmental agencies, NGOs, INGOs, and other related agencies, as the at-risk residents belong to lower-income categories. Furthermore, we identified and suggested detailed evacuation routes and a safe evacuation system, which only became possible with the combined approach applied here.
The focus group discussions revealed a lack of adaptive capacity-building strategies and risk reduction knowledge among the locals. Furthermore, locals must participate in disaster preparedness training and awareness programs to increase their safety. In addition, any relocation strategy put forward by the government would likely require an extended period, during which the construction of a river embankment may be one of the realistic alternatives for decreasing flood disaster risks in the shorter term. These were derived from the combined approach, which can be adapted in other parts of Himalayan countries.
5. Conclusions
The inundation maps created in this study show that the Masinabagar, Laltinbazar, and KI-sing areas are highly susceptible to future flooding. In Masinabagar, 268 of the 466 buildings, including 229 houses, churches, crematoriums, livestock farms, restaurants, warehouses, and schools, were situated in the inundation zone. In Laltinbazar, of 329 buildings, 232 buildings, including 225 houses, warehouses, and a primary school were situated in the estimated flood-prone zone. Similarly, in KI-sing, of the 210 buildings, 68 were located in the inundation zone, including 32 houses, a research center, community centers, and temples. Evacuation routes in the three settlements were either inadequate or absent. Moreover, lower-income residents were at much higher risks, as most houses near the riverside area belonged to impoverished migrants and laborers.
The results highlighted that immigrant households were at the highest risk (p < 0.001 for both factors). Similarly, 455 laborers’ houses by occupation were also significantly correlated with inundation risk (p < 0.001). The correlation between income and migration was statistically significant for all households (p < 0.001). Similarly, the correlation between income and occupation was also significant (p < 0.001).
Most riverside residents remain unaware of the preparation and risk associated with flood emergency evacuation; thus, adaptive capacity-building strategy programs, disaster preparedness training, and evacuation drills must be urgently incorporated. Establishing adequate evacuation routes and shelter houses is the most pressing need. It is suggested that governmental agencies, NGOs, INGOs, and other related agencies prepare an adequate plan that targets the impoverished residents of the inundation zone, as they cannot afford safe lands for relocation. This research contributes to reducing the impact of flood disasters, and the proposed evacuation system will help save lives. This study will be useful for planners with respect to preparedness, early warning systems, and safe evacuation and will serve as a guide for sustainable development in the region. Accordingly, this study recommends applying a combined approach that uses modeling and socioeconomic surveying in other parts of the Himalayan region and impoverished areas around the globe.