An Analysis of Social Vulnerability to Natural Hazards in Nepal Using a Modified Social Vulnerability Index

The social, economic, and demographic attributes of Nepal, which intersect with geophysical and hydrometeorological processes, reflect directly on the social vulnerability of the country. The status of these attributes is rooted in a historically sluggish development process. In the past decades, development activities have been centered in the capital city of Kathmandu and a few other urban areas, marginalizing rural areas in the Mountain, Hill, and Tarai regions from the process. The socioeconomic situation is further aggravated by a lack of infrastructure, energy, government accountability, and financial resources. As a result, Nepal is struggling to escape the poverty trap and remains one of the least developed countries, ranking 144th in a recent Human Development Index (HDI) report (UNDP 2016). According to the latest census of 2011, the population has been growing steadily at more than 2% per annum and has close to tripled from the 1961 population of 9.5 million (CBS 2012). The total adult literacy rate is 59.6%, with a female literacy rate of only 48.8%. Nepal is predominantly an agrarian society with roughly 60% of the population dependent on subsistence agriculture. The majority of land is not arable due to the steep slopes and rocky soils of the mountainous region, and agriculture is highly monsoon-dependent. Poverty has compelled many to emigrate to neighboring countries and the Middle East in search of employment, which provides much needed remittances. The recent 2011 census reports that one in every four households (25.4%) has at least one member living outside of Nepal, and the greatest proportion of the absentee population (44.8%) is among the economically productive 15–24-year cohort (CBS 2012). Most migrants are male, and female-headed households have increased from 14.9% in 2001 to 25.7% in 2011.

Nepal exhibits great demographic diversity, which influences social vulnerability. Hinduism is followed by more than 80% of the population (Buddhism and Islam are other common religions), but is comprised of many castes and subcastes. The 2011 census also accounts for 125 ethnic groups and 123 languages across this relatively small country (CBS 2012), and these demographics are geographically clustered. Only a few ethnic and caste groups are dispersed throughout the country—modern-day Nepal was formed in the 18th century by unifying numerous small principalities that were primarily based on caste, ethnicity, and language. Some of these social markers, specifically the caste system, lead to discrimination and social exclusion, especially in rural areas. Spatial attributes of these demographic features may be revealed through spatial analyses of social vulnerability scores.

Adding to the complicated socioeconomic and demographic contexts is the risk of and exposure to a multitude of hazards—such as earthquakes, floods, landslides, thunderstorms, debris flows, avalanches, glacial lake outburst floods (GLOFs), and forest fires—and these hazards are distributed unevenly throughout the country (Aksha et al. 2018). The 2015 Gorkha Earthquake alone killed about 9000 people and damaged or destroyed more than 750,000 buildings. A majority of the losses occurred in rural areas where low-quality, traditional masonry is the predominant housing material. Due to competition for resources such as jobs and land that is accessible, arable, and easy to develop, people are pushed into high-risk areas (Aryal 2014).

Nepal is situated in the Himalayas between India and China. Elevations range from just 57 m above sea level to Mt. Everest, at 8848 m the world’s highest peak, within a relatively narrow latitudinal spread. Nepal is divided into three ecological regions: Tarai, Hill, and Mountain (Fig. 1). The Tarai region is the southernmost part of Nepal. This “grain basket” of the country has relatively low, flat, and fertile land comprised of alluvial soils. The Tarai region covers roughly 24% of the total area of Nepal, but is home to roughly 50% of the population. The Hill region rises from 1000 to 4000 m above sea level and comprises about 42% of Nepal’s total area. The Hill region is the first orographic barrier to monsoon winds, which produce heavy precipitation on its southern flanks. The increasingly urbanized Kathmandu and Pokhara valleys lie in this region. North of the Hill region lies the Mountain region, which covers about 34% of the total land area of Nepal. The Mountain region has subalpine to alpine climates and is characterized by river valleys, tectonic basins, snow-capped peaks, glaciers, rocky slopes, and colluvial deposits. This region is sparsely populated.

Nepal was divided into five development regions, 14 zones, 75 districts, 53 municipalities, and 3918 village development committees (VDCs). However, in 2015, a new constitution was adopted that created three levels of governance: the federation, the province, and local bodies. Now, Nepal is divided into seven provinces and 744 local bodies (four metropolitan cities, 13 sub-metropolitan cities, 246 municipalities, and 481 rural municipalities), although administration through these bodies has not yet commenced. This study was based on the previous administrative bodies, that is, the 3918 VDCs and 53 municipalities. The previous bodies were chosen because the scale is smaller (total of 3971 spatial units compared to 744), which allows statistical and spatial analyses to produce finer results. The new administrative bodies were formed by simply combining groups of the previous bodies, so the new administrative bodies are still captured in our analyses, but at a finer scale. Use of the previous bodies supports analyses across more observations and more spatial units while still being administratively relevant. Finer-scale analyses are especially important in the Nepali context, which exhibits great socioeconomic, cultural, and physiographic diversity across short expanses of space.

This study adopted the SoVI method developed by Cutter et al. (2003), which employed more than 200 variables to quantify and analyze social vulnerability in the United States. Conceptualizations of social vulnerability and the SoVI methodology have been refined over time and applied in many contexts (Boruff and Cutter 2007; Holand and Lujala 2013; Zhou et al. 2014; Hummell et al. 2016). Similarly, this study analyzed individual factors such as education, employment, ethnicity, and health as well as community characteristics such as level of urbanization, access to medical services, and built environment attributes to construct a modified SoVI for Nepal.

Since the original SoVI was applied to the United States context, modifications were necessary to adapt the model to Nepal’s distinct sociophysical context. To construct the modified SoVI, we extracted data from the full dataset of the most recent 2011 census, which was provided by the Central Bureau of Statistics (CBS), Government of Nepal (CBS 2012). The CBS conducted the census using two household level questionnaires. The first questionnaire collected basic demographic data, and the second questionnaire included more detailed questions on topics such as migration, fertility, and economic activity. The CBS adopted a systematic sampling method, and roughly 15% of the entire population of Nepal were sampled. Using all available data from the 2011 census, Table 1 provides a list of concepts and variables used to construct the SoVI.

Modifications to the original SoVI were made to capture fundamental demographic differences between Nepal and the United States. To capture “Ethnicity” in Nepal we included Dalit (the lowest strata of the Hindu caste system) and Minority Populations. Dalits are considered “untouchable” and encounter systemic social and economic oppression. About 15.6% of the total population in Nepal are Dalits, and there are 26 distinct, concentrated Dalit groups distributed heterogeneously throughout Nepal—each with a different language, culture, religious practices, and means of employment (Dahal et al. 2002). Minority Populations encompasses indigenous populations and other historically disadvantaged groups. Indigenous populations in Nepal are defined as people who possess distinct cultural traditions, languages, and religious faiths based on animism (Bhattachan 2008). Fifty-nine separate groups are considered Minority Populations, and they are further divided into five categories based on their status: endangered, highly marginalized, marginalized, disadvantaged, and advanced. The advanced category, which consists of the Newar and Thakali castes, actually have the highest HDI values (Sharma et al. 2014) and were thus excluded from Minority Population status in our SoVI. However, Muslims and Sikhs/Punjabis were included since they are considered disadvantaged groups who together account for less than 5% of Nepal’s total population.

We also included “Built Environment” variables to reflect physical attributes of housing and public services such as electricity, water, and sewer infrastructures, as similarly included in a SoVI in Brazil by Hummell et al. (2016). According to the 2011 census, 25% of houses have external walls constructed from low-quality materials such as bamboo and wood, and only 10% of houses have reinforced cement concrete (RCC) foundations. As the entirety of Nepal lies in a highly active seismic belt, housing foundations and wall materials significantly impact vulnerability. Many fatalities from the 2015 Gorkha Earthquake, for example, were among people residing in low-quality houses. Severe damages were sustained by traditional mud and stone houses, especially in rural areas. In addition to housing quality, only 52% of households are connected to a piped water supply, only 39% have a toilet or latrine, and 33% are without electricity as their primary source of energy.

Populations who do not speak/understand Nepali were added to the “Special Needs Populations” indicator because inability to speak or understand Nepali substantially restricts access to critical and timely knowledge on disaster warnings as well as post-disaster recovery and rehabilitation programs. In the census questionnaire, individuals were asked to report their mother tongue and second language. Those who did not report Nepali were identified as individuals who do not speak/understand the Nepali language. Knowing the most widely spoken language is important because it is used extensively for basic communication and relaying information on disasters. Such populations also encounter linguistic issues (for example, completing and submitting official forms) when attempting to receive compensation and other support from the government. Although Nepali is no longer the official national language, it is the only language used for administrative and government communications, which constitutes a major disadvantage for nonspeakers, especially in emergency situations.

A total of 39 variables were selected for the construction of the modified SoVI (see Table 1). All variables were normalized using percentage, density, or per capita functions prior to statistical analysis. Principal component analysis (PCA) via Statistical Package for the Social Sciences (SPSS) version 21.0 software generated a set of independent factors. Principal component analysis is a factor reduction methodology that identifies a smaller number of components to explain the variance observed in a larger dataset (Abdi and Williams 2010). The purpose of performing PCA is to distill a broad explanation of the data by grouping like-variables into components that adequately account for covariation among the larger number of analysis variables. The analysis also facilitates the interpretation of component groups and provides insight when data are examined with further analyses. Following methods developed by Cutter et al. (2003), Kaiser normalization and Varimax rotation were employed as extraction methods for components. Only components with eigenvalues higher than 1.0 were extracted and named to indicate the latent variables. Each component was named and assigned cardinality (±). The SoVI was calculated by adding the scores of each component to arrive at total SoVI scores. An equal weighting and additive approach was used in the absence of empirical and justifiable evidence for weighting components differently, as has been exercised in similar studies (Cutter et al. 2003; Hummell et al. 2016). Our modified SoVI for Nepal was calculated for each spatial unit (the 3918 village development committees and 53 municipalities) by adding the principal component values:The SoVI scores of each spatial unit were mapped using ArcMap version 10.3 to visualize the most and the least vulnerable villages in Nepal based on standard deviation from the mean value. We used global and local Moran’s I tests of spatial autocorrelation to investigate whether social vulnerability is clustered or random across space in statistical terms. Global and local tests calculate autocorrelation among all villages and determine statistically significant patterns of similarity and dissimilarity in the spatial location and distribution of social vulnerability. The global Moran’s I test determines whether the geolocations of index results exhibit a spatial pattern. The interpretation is similar to that of correlation coefficients: values close to + 1 indicate strong positive spatial autocorrelation (that is, clustering of either high or low values of social vulnerability), while values close to − 1 indicate strong negative spatial autocorrelation (that is, alternation of high and low values for adjacent observations). Values near zero indicate the absence of spatial patterns, or randomness.

Local Moran’s I was deployed to classify identified clusters according to type of association. These results were used to map four typological gradations of social vulnerability at the village and municipal level: (1) High–High (HH), a village with a high SoVI score surrounded by villages with high SoVI scores; (2) Low–Low (LL), a village with a low SoVI score surrounded by villages with low SoVI scores; (3) High-Low (HL), a village with a high SoVI score surrounded by villages with low SoVI scores; and (4) Low–High (LH), a village with a low SoVI score surrounded by villages with high SoVI scores. Hot spots (HH) and cold spots (LL) denote positive spatial autocorrelation, indicating clusters of similar values, whereas LH and HL denote negative spatial autocorrelation, indicating clusters of dissimilar values. Together, global and local spatial analyses identify patterns of social vulnerability that are either not visible (that is, masked) or appear visually meaningful yet are not statistically significant (that is, false positives). The inclusion of spatial clustering advances the original research work using the SoVI method by Cutter et al. (2003) by adding new analytical abilities to the method.