Assessment of Social Vulnerability to Natural Hazards in the Mountain Kingdom of Lesotho

Abstract Lesotho's landlocked mountainous setting with thin regolith cover and poor economic situation make it vulnerable to hazardous events associated with climate such as drought, floods, strong winds, heavy snowfall, and severe frost. To date, no quantitatively based vulnerability assessment has been undertaken in this heavily populated, developing region in southern Africa. The primary aim of this study was to assess social vulnerability of communities to natural hazards by applying a place-based social vulnerability index developed for the United States, to the Lesotho context. The study used 2006 Lesotho census data, district government records, and household interviews to identify 27 indicators of social vulnerability in southern Lesotho, and then used principal components analysis to generate a social vulnerability index for the study region. Index scores were summed and then mapped to quantify spatial variability in social vulnerability. The study results show a clustering of highly vulnerable communities in the rural highlands as a result of underdevelopment, poverty, and inaccessibility.


Introduction
Considerable attention has been paid to the vulnerability of developing and landlocked countries to climate change and associated natural hazards (UN-OHRLLS 2009;Karki and Gurung 2012). Article 4 of the United Nations Framework Convention on Climate Change requires parties to give full consideration to meeting the needs and concerns of developing countries arising from the adverse effects of climate change. Of particular interest are landlocked countries with fragile and mountainous ecosystems (UNFCCC 2011;UNOHRLLS 2014).
Lesotho is a small (30,588 km 2 ), mountainous, and landlocked country and one of the least developed countries in the world; it is prone to natural hazards, its fragile ecosystems are regarded as highly vulnerable to climate change (Matarira 2008;Gwimbi et al 2012;Matarira et al 2013;UNDP 2014). Natural hazards in Lesotho, which include floods, drought, frost, strong winds, and heavy snowfall, have caused many human and livestock deaths as well as damage to property and loss of crops Matarira et al 2013;Grab and Linde 2014).
The impacts of natural hazards are exacerbated in many mountain areas; more than 70% of Lesotho's population lives in remote and ecologically fragile mountainous terrain. The importance of understanding the social dynamics of vulnerability in mitigating natural hazards has been acknowledged by the Lesotho government (DMA 2010;Gwimbi et al 2012). However, while some studies have focused on vulnerability to food insecurity (LVAC 2007(LVAC , 2011DMA 2012;Famine Early Warning Systems Network 2013), there has been no comprehensive institutionalized effort to assess social vulnerability to natural hazards. Although social vulnerability is considered a multidimensional construct, which is not easily defined by a single variable, it is generally viewed as a ''measure of both the sensitivity of a population to natural hazards and its ability to respond to and recover from the impacts of hazards'' (Cutter andFinch 2008: 2301). We focus on individual, family, and community vulnerability within various geographic subregions of southern Lesotho, but for ease of expression, henceforth we refer to such collective social vulnerability as that concerning the ''community.'' Such an assessment is an important initial step toward understanding natural hazard risks, improving response capabilities, and supporting natural disaster prevention, mitigation, preparedness, and recovery programs.
This study was an attempt to quantify social vulnerability to natural hazards in southern Lesotho using a social vulnerability index methodology to assess the social factors and processes that make communities vulnerable and the spatial dynamics of that vulnerability. We incorporated indicators specific to Lesotho's context, which may also be applicable to other developing mountainous regions where data availability is a challenge. Adjustments in the selection of data and input variables were required to fully capture the geography of Lesotho and the study area.

Environmental setting
Lesotho is a mountainous yet densely populated (ca 1.9 million inhabitants), landlocked country within South Africa (Figure 1). It has 4 agro-ecological zones: lowlands (1400 to 1800 m above sea level [masl]), foothills (1800 to 2000 masl), the Senqu River Valley (1400 to 1800 masl), and highlands (2000 to 3482 masl); the highlands make up 70% of the country (LMS 2001). Lesotho has varied geomorphology and topography, including microclimatological influences, which have a significant impact on its ecology, livelihoods, and economy. Rainfall, most of which falls between October and April, varies between 700-1000 mm per year depending on location. The northern and eastern highlands are wettest, while the study region in southwestern Lesotho is driest (Sene et al 1998;Hydé n 2002). Monthly mean minimum temperatures in winter range from about 26uC in the highlands to 5uC in the lowlands; subzero temperatures occur during most nights in winter (mid-May to mid-August). In summer (November to February), mean maximum temperatures vary from 29uC in the lowlands to 16.5uC in the highlands (Wilken 1978;LMS 2005). Severe frosts and snowfalls are possible from May to September Moeletsi and Walker 2013;Grab and Linde 2014).
Several reports and authors have argued that the country has experienced increasing numbers of severe weather events during recent years, as measured by an increasing frequency of natural disasters (eg LMS 2002(eg LMS , 2010Gwimbi et al 2012;Matarira et al 2013). Archival research has demonstrated that, between 1825 and 2012, human lives were occasionally lost, livestock was killed, crops were destroyed, and buildings and infrastructure were damaged or destroyed by severe weather (Eldredge 1987;Showers 2005;Grab and Nash 2010;Nash and Grab 2010). Heavy snowfall has been the most frequent hazard, followed by severe frost, drought, floods, and strong winds (Table 1).
Given Lesotho's highly varied topography and associated climate gradients, the spatial and temporal occurrence of hazards is also varied, and so is their impact. While much of the country is vulnerable to natural hazards, some hazards (eg heavy snowfall, severe frost, and drought) are particularly prevalent in the study area ( Figure 2). For instance, most villages located above 2700 masl are prone to heavy snowfall and prolonged snow cover (Grab and Linde 2014). Often, villages and settlements in the highlands are negatively impacted by snow, mainly because it isolates them from other parts of the country. The highlands have a weak agricultural and natural resource base (due to mountainous terrain and unproductive soils) and suffer from inaccessibility and scarce income earning opportunities (Turner 2005;Sechaba Consultants 2006;One World 2009, 2010. Reliance on progressively lower-yielding subsistence agriculture, deteriorating pastures, and decreasing livestock health leads to deteriorating household income, assets, social status, and social support.

Administrative divisions
Lesotho is organized as a constitutional monarchy with 10 districts subdivided into 80 constituencies consisting of 129 community councils (village-based public representative bodies; KoL 2010). The kingdom is regulated by a dual legal system consisting of customary and general law. The study area focused on 2 geographically varied districts in southwestern Lesotho: Mohale's Hoek and Quthing (30u099S; 27u289E to 30u249S; 27u429E). Mohale's Hoek totals 8 constituencies and 14 community councils, while Quthing has 5 constituencies and 10 community councils (Table 2). Each community council consists of a number of villages.  The study area's 2 districts and 24 community councils were used as the spatial units of analysis for this study. Community councils were the most appropriate unit of analysis because of the administrative structure in Lesotho, availability of demographic data, and the statistical requirements for the creation of the social vulnerability index (SoVI). Community councils constitute an entity whose residents share ideas, have similar livelihoods, values, and natural hazard risk culture. They are also meant to represent an area with a fairly homogenous population with respect to demographic characteristics, economic status, and living conditions. Additionally, many government functions and decisions begin at the community council level.

Methodology
The study used both quantitative and qualitative methods, which allowed us to assess social vulnerability from different perspectives and integrate the viewpoints of different stakeholders. It relied on 2006 census data and the 2010 district information handbooks at the community council level, key informant interviews with the district disaster management offices, as well as 314 household interviews in the 24 community councils of the 2 districts studied (BoS 2015). We adopted a multistage sampling technique (Agresti and Finlay 2008), which included stratified sampling of respondents based on their geographical location, from the district level to the village level. A stratified random sampling technique was then employed at the village level to select respondents for household interviews and to ensure that the sample was representative.
The 27 social vulnerability variables chosen for the study were identified either from those mentioned in previous studies (Cutter et al 2003; see also Wood et al 2010;Dunno 2011;Yoon 2012;Chen et al 2013;Cutter and Morath 2013) or by disaster management officials in Lesotho. Most variables correspond to widely studied dimensions of social vulnerability: socioeconomic status, gender, education, age, unemployment, rural or urban location, special needs, housing conditions, family structure, occupation, and social dependency (Dunno 2011; Cutter and Morath 2013). Others were chosen for their relevance to the Lesotho context: access to roads, electricity, radios, mobile phones, toilets, and piped water.
Thus, we adapted a social vulnerability index (SoVI) originally developed by Cutter et al (2003) for the United States and applied it at the community council level in the 2 study districts. The SoVI provides an empirically based comparative measure that facilitates the geographic examination of relative levels of social vulnerability across space (Cutter et al 2003;Cutter and Morath 2013). It focuses mainly on the capacity of a population to prepare for, mitigate, respond to, and recover from natural disasters.
All the variables were normalized using percentage, density, or per capita (per km 2 ) functions, then variables were standardized using z-scores standardization. This generates variables with a mean of 0 and standard deviation of 1 (Cutter et al 2003;Yoon 2012). The variables were then input into the Statistical Package for Social Sciences (SPSS) to carry out a principal components analysis, which reduced the 27 variables into 8 multidimensional components explaining 80.95% of the variance. The PCA produced a minimum number of components that adequately accounted for the covariation among the larger number of analyzed  variables. The Kaiser criterion was then used as an extraction method (eigenvalues greater than 1) and a Varimax rotation. The Varimax rotation minimizes the number of variables that load high on a single component, increasing the percentage of variation between each. Only components with eigenvalues greater than 1.0 were extracted and named to indicate the latent variable. The component scores were then summed with equal weights to create a final SoVI score. SoVI scores were standardized, then ranked in such a way that positive values indicated high vulnerability scores while negative values indicated lower ones. In the absence of any theoretical justification for the weighting of components, the equal weighting and additive approach seemed to be the most practical (Cutter et al 2003;Schmidtlein et al 2008). For more information on the SoVI, see Hazards and Vulnerability Research Institute (2013). The SoVI scores for each community council were then mapped in ArcGIS as standard deviations (SD) to represent variations in relative social vulnerability levels. The SoVI scores were grouped into 5 categories from most vulnerable to least vulnerable, ranging from SD .1.5 to SD #1.5.

Results
The 8 components extracted from the principal components analysis, their rank in importance, the explained variance, and their directionality (6) are presented in Table 3. The components include: access to resources, vulnerable population groups, population density, family structure, economic status, employment, access to services, and rurality. The individually adjusted variables and their directionality were made to reflect their known influences on social vulnerability as derived from the empirical literature ( . Thus, a positive directionality was allocated to components that likely increase social vulnerability (eg orphans, females, elderly), and a negative directionality was assigned to components that likely decrease vulnerability (eg income, employment, education, wealth). Table 4 presents SoVI scores for the community councils in the study area from most vulnerable to least vulnerable. The 2 rural and highland community councils in the western part of the study area exhibit the highest levels of social vulnerability, while Mashaleng, an urban community council also in the western part, exhibits the lowest SoVI score (Figure 3). Additionally, urban and lowland community councils such as those of Khoelenya and Liphakoe exhibit relatively low SoVI scores. Community councils showing moderate social vulnerability are found in the central part of the study area.
The SoVI scores were mapped in ArcMap using an SD classification to show spatial differences in levels of social vulnerability across the study area (Figure 3). Three potential clusters of social vulnerability emerged. The most vulnerable community councils were mostly clustered in Mohale's Hoek district, in the western and central parts. The least vulnerable community councils were also predominantly in Mohale's Hoek district and clustered in the western and central parts of the study area. Moderate levels were largely in the central part of the study area. In contrast, Quthing district exhibited clusters of moderate and low vulnerability levels.
Spatial patterns of high social vulnerability occurred in Matsatseng and Thabana Mokhele community councils. Other areas of high vulnerability were also rural, extending to the highlands in both districts (Likhutloaneng, Nkau, and Mphaki). Areas of low social vulnerability exist in Mashaleng. Areas with moderately low vulnerability levels

Factors influencing social vulnerability in Lesotho
A number of spatial patterns emerged for many components; these patterns influenced the overall pattern of social vulnerability in the study area and indicated where pockets of high social vulnerability are located. Therefore, it is essential to highlight the individual components that contribute to the overall SoVI score shown in Figure 3. By assessing the individual components, the main causes of social vulnerability in each community council can be clearly understood. These scores are discussed in the following 2 sections and mapped in Figures 4 and 5.
Social vulnerability index components 1-4 Access to resources: Factors affecting access to resources include employment and income and are constructed from economic activities engaged by households. The rural community councils of Likhohlong, Likhutloaneng, Mkhono, Qabane, Seforong, and Tsatsane have limited access to resources. Household interviews and field observations confirmed that there were limited formal employment opportunities in areas with limited access to resources, and hence high levels of poverty and social vulnerability. The urban community councils of Liphakoe and Motlejoeng exhibited low SoVI scores in this component. Vulnerable population groups: Vulnerable population groups include child-headed households, orphans, and people aged 65 years or older. These groups have high prevalence in the urban and peri-urban lowlands. Many orphans and child-headed households were observed on their own in the urban lowlands during field visits, mainly owing to the existence of social networks and cohesion still existing in rural areas of Lesotho, where most orphans are taken care of by relatives and neighbors. Low vulnerability scores for this component were common in the rural highlands.
Population density: The population density component includes total population, number of households, and percentage of female population. High vulnerability scores are evident in the western part of the study area, which consists mainly of the urban and peri-urban lowlands. The southeastern region, particularly Quthing district, had low vulnerability scores in this component. The spatial disparities presented in the population density map emanate from the rural-urban dichotomy, whereby low-lying urban areas are most favorable for settlements given the availability of arable land, economic opportunities, and social services.
Considerably lower population densities are found in remote mountain areas.

Discussion and conclusions
The adoption of the SoVI approach in Lesotho is of value for several reasons: N Its use of census data means that it can be updated as new data become available, allowing for time-series analyses of social vulnerability (Simpson and Katirai 2006).
N Finally, it offers a way to include local views in quantitative analyses (through household interviews, expert interviews, and focus group discussions) and thus to comply with the framework of the Intergovernmental Panel on Climate Change (IPCC 2012), which calls for the integration of local knowledge with scientific and technical knowledge to improve disaster risk reduction.
This research expands upon Cutter's (1996) Hazards of Place vulnerability model by incorporating qualitative and quantitative data to assess vulnerability. Its theoretical and practical contribution to the literature on natural hazards and vulnerability is that it advances both theory and knowledge about vulnerability within the context of a developing mountainous region in which the lack of accurate and reliable data is often a major limitation. Flexibility in the selection of data and input variables is required to fully capture the geography of Lesotho and the study region in particular. This study provides a spatially based method for identifying vulnerable populations and communities, which is particularly relevant to dynamic mountainous terrain settings, and compares their levels of vulnerability at district and community council levels. The results can help decision makers to produce more localized and hazard-specific mitigation strategies.
The vulnerability of rural highland communities-due in part to poverty, limited access to resources and services, family structure, and dependence on rain-fed agriculture-is an example of place-based vulnerability. Limited access to resources in these areas is often coupled with low education levels and lack of sanitation facilities and communication tools such as radios and mobile phones. Less socially vulnerable urban community councils have greater ability to cope with natural hazards. Access to economic resources and services, which is strongly influenced by components within given spatial realms in the mountainous regions of Lesotho, is an important factor determining the spatial dynamics of social vulnerability, as is also likely the case in other developing rural mountain regions.