1 Introduction

1.1 Arsenic contamination: a naturally occurring hazard

Groundwater arsenic (As) contamination is a global phenomenon affecting more than 70 countries on six continents (Ravenscroft et al. 2011). As a densely populated region, the Ganga–Meghna–Brahmaputra (GMB) plain intensively exploits groundwater resources, experiencing a myriad of environmental impacts associated with local, regional, and global issues. The vulnerability of GMB is enhanced due to threats from natural processes, and socioeconomic and demographic forces. The GMB plain in India and Bangladesh contains the highest As-contaminated contiguous area in South Asia (SA) affecting about 500 million people’s lives and encompasses an area of 569,749 km2 (Hossain et al. 2006). The groundwater As contamination in India was first noticed in 1976 in the Chandigarh district, about eight decades after the first case of As poisoning in Poland (1898) (Mandal and Suzuki 2002). Later in 1982, India received a wake-up call because of another case of As poisoning in North 24 Pargana district of the state of West Bengal (WB) (Chakraborti et al. 2003). The groundwater’s As contamination and its health impact on human beings has been extensively studied in WB (Chowdhury et al. 2000; Mazumder et al. 2010; Guha Mazumder and Dasgupta 2011; Rahman et al. 2013). A total of 111 community blocks (out of 341) have been reported as As-affected blocks in 12 As-affected districts (out of 19) (SOES 2006). About two decades after the WB case, very high concentrations of As were detected in the groundwater in Bihar (Chakraborti et al. 2003). Later, the groundwater As contamination was revealed in other geographic locations in India including Uttar Pradesh, Jharkhand, Assam, Tripura Arunachal Pradesh, Nagaland, Manipur, Punjab, Haryana, Himachal Pradesh, Chhattisgarh, Hyderabad, and Andhra Pradesh (Datta and Kaul 1976; Mukherjee et al. 2006; Nickson et al. 2007). As is a group “A” carcinogen, and the ill effects of exposure to it cause serious health problems among the affected groups, which can lead to their economic and social marginalization. Reducing the risk from hazards of natural origin (arsenic in this case) is a major challenge concerning global environment change (Birkmann et al. 2013). Groundwater As contamination is most appropriately considered a chronic hazard because of its long-term, cumulative harmful effects. A chronic hazard presents unique challenges in terms of the public perception, which may engender and influence subsequent policy responses.

1.2 Vulnerability: perspectives, indexes, and models

Vulnerability assessment of natural hazards and climate change has emerged in the past decades as an important research field bringing together scientists from different disciplines (Adger et al. 2004; IPCC 2007; Parry 2007; Birkmann et al. 2013). In the absence of a universal definition for vulnerability and common conceptualization, formulations of vulnerability have proliferated (Birkmann et al. 2013). The natural science research communities often focus on the quantification of different factors contributing to vulnerability, often to the exclusion of socioeconomic factors. These approaches highlight physical vulnerability and often attempt to quantify damage ranges. These are illustrated using vulnerability curves, which help determine acceptable levels of potential losses (Kienberger et al. 2009; Papathoma-Köhle et al. 2011). In contrast, social science approaches often encompass a broad focus and examine, in particular, the impact of exposure to environmental hazards on individual households or a community, as well as the contextual conditions that influence social vulnerability (Wisner 2004). According to Phillips and Fordham (2009), social vulnerability to natural hazards is driven by social inequality and is deeply embedded in social structures that are often resistant to change (Phillips and Fordham 2009). Many assessment approaches characterize vulnerability according to the degree of susceptibility or fragility of communities, systems, or elements at risk and their capacity to cope under adverse conditions. In recent years, different frameworks have been developed to better systematize different facets of vulnerability, most often in relation to climate change adaptation and disaster risk management (Birkmann et al. 2013). However, developing a universal metric or measurement tool for vulnerability assessments across all disciplines is challenging because of the definitional elusiveness and heterogeneous nature and scale of analysis (temporal and spatial). The vulnerability indexes or indicators are quantifiable measures envisioned to represent a characteristic or a parameter of a system of interest using a single value, and have been applied in different contexts using different variables (Cutter et al. 2008). Among all the variables used to derive vulnerability indicators, socioeconomic status (wealth or poverty), age, special needs population, gender, race, and ethnicity have been studied most extensively as social factors that increase or decrease the impact of specific natural hazard events on a local population (Tierney 2001; Tierney et al. 2001; Center 2002; Bates and Swan 2007). It has been reported that selecting a single variable (e.g., race, gender, or poverty) does not adequately capture the characteristics of the communities described. In most cases, it is a combination of several socioeconomic attributes and circumstances, which defines social burdens from natural hazards (Cutter et al. 2009).

The choice of methodology to derive vulnerability indicators and appropriate weighting for each individual variable has been the most important constraint in vulnerability studies. Because of the absence of reliable theoretical and/or statistical evidence needed to assign weights, all indicators are usually assigned equal weight, thereby according them the same relative importance (Cutter et al. 2009). However, indicators can also be weighted according to the percent variance explained by each factor (Cox et al. 2006). Furthermore, vulnerability maps can be invaluable for adapting and planning mitigation frameworks in highly vulnerable areas. For the current study, we define vulnerability as the risk of experiencing economic loss, health problems, social isolation, social discrimination, loss of opportunities, or decline in socioeconomic status during the period of exposure to specific or multiple socioeconomic or environmental problems. No previous studies have been reported, which have applied the vulnerability framework to As contamination. This study is the first of its kind to propose a conceptual composite vulnerability framework and derive composite vulnerability indexes for As contamination.

2 Framework and methodology

2.1 Composite vulnerability framework

Deriving and mapping composite vulnerability (CV) to As contamination is a key step in quantifying total vulnerability due to existing biophysical and socioeconomic conditions in As-contaminated areas. The CV map has the potential to predict highly vulnerable areas where As mitigation is an urgent necessity; type(s) of As-mitigation technology(s) requires feasibility, sustainability for the specific vulnerable area(s); and the area(s) where the likelihood of success of an As-mitigation program is greatest. The components of CV framework are based on the concept that vulnerability is a function of exposure, sensitivity, and adaptive capacity (Fig. 1). The definitions of these three elements of vulnerability assessment have been adapted for our work. Exposure can be defined as the degree of environmental stress upon a particular unit of analysis; it may be represented as either long-term change in environmental conditions or changes in the magnitude and frequency of extreme events. Sensitivity is the degree to which a system will respond to a change in the environment, either positively or negatively. Adaptive capacity is the ability of a system to adjust to actual or expected environmental stresses, or to cope with the consequences. Adaptive capacity is also considered “a function of wealth, technology, education, information, skills, infrastructure, access to resources, and stability and management capabilities” (UNEP 2003). Although, theoretically, technology plays a vital role in enhancing societal adaptive capacity, we have omitted technology as a variable from our study because groundwater As contamination mainly affects rural areas in SA, where technological development has been minimal (Das 1999; Singh and Jha 2012).

Fig. 1
figure 1

Conceptual model of composite vulnerability framework

Full size image

2.1.1 Quantifying vulnerability indexes

The method to derive vulnerability indexes was modified and simplified for this study. The core of the methodology is the method used to derive vulnerability indexes in the guidelines of the United Nations Environment Program’s Assessing Human Vulnerability to Environmental Change (2003). Data were treated as per the UNEP guidelines (UNEP 2003). The variables used to derive vulnerability indexes are listed in Table 1.

Table 1 Indicator variables for vulnerability calculation
Full size table

The steps involved in quantifying vulnerability indexes are deriving vulnerability interval value (VIV) and vulnerability indexes (VI). The VIV is derived by using Eq. (1).

$${\text{Vulnerability}}\;{\text{Interval}}\;{\text{Value}}\,({\text{VIV}}) = \frac{{{\text{VV}}_{\hbox{max} } - {\text{VV}}_{\hbox{min} } }}{{{\text{VI}}_{\hbox{max} } }}$$
(1)

where VVmax is the maximum value of vulnerability variable; VVmin is the minimum value of vulnerability variable; and VImax is the maximum value of vulnerability index at “5 levels” of vulnerability scale. After deriving the VIV, the VIs can be calculated based on Eq. (2)

$${\text{VI}}_{\text{ndex}} = {\text{VV}}_{\hbox{min} } + {\text{VIV}}$$
(2)

where VVmin is the minimum value of vulnerability variables and VIV is the vulnerability interval value derived from the Eq. (1).

Furthermore, following these steps, five vulnerability indexes were derived and at every consecutive step, VVmin was replaced with VI value of the previous step. Finally, five levels of vulnerability indexes were derived and given appropriate weightage (weighted equally), and each As-affected district was assigned an appropriate VI values (Table 2).

Table 2 Vulnerability indexes and associated vulnerability category
Full size table

The VI values for As-contaminated areas were derived based on the As-concentration profile of the area. The VI values from 1 to 5 were assigned based on As concentration in the drinking water between 0–10 µg/L, 11–50 µg/L, 51–100 µg/L, 101–200 µg/L, and As levels more than 200 µg/L, respectively. In the absence of fluoride (F), nitrate (N), and iron (Fe) concentration data, areas with these contaminants were given the VI value “1” for each contaminant and the area without these contaminations the VI value “0.” Areas affected with flood and droughts were treated similarly to F-contaminated areas. We did not characterize flood- or drought-affected areas according to the intensity of their occurrence. Flood and drought in the region are chronic natural hazards and affect a largely impoverished population, often resulting in mass displacement and loss of livelihoods. The damage is significant, and evaluating flood or drought intensity based on their occurrence and impacts in more than half of the districts in the state is beyond the scope of our work, and even more importantly, unlikely to yield additional insights. Therefore, we only considered the presence or absence of flood or drought in the As-affected areas as an additional environmental stressor. Hydrogeochemical conditions also significantly contribute to the sensitivity of As-prone areas. For example, the following characteristics of the area such as geological formation, lithology, physiography, high HCO3 − load in the groundwater have been correlated with the As concentration in affected areas (Lado et al. 2008; Winkel et al. 2008; Saha and Shukla 2013). Hydrogeochemical parameters such as hydraulic conductivity and chemical composition are aquifer-specific and could be used to predict As vulnerable areas (Lado et al. 2008, Winkel et al. 2008). For this study, we have used information about the geological formation (Quaternary to upper quaternary = 1 and Quaternary = 2), lithology of the areas (younger alluvial, older alluvial, and red sandy soil = 1; younger alluvial and older alluvial soil = 2; younger alluvial and calcareous alluvial soil = 3; and younger alluvial soil = 4), and the physiography (hill = 1; alluvial plain = 2) of the areas.

2.2 Principal component analysis (PCA)

Principal component analysis (PCA) has been widely used in vulnerability studies (Cutter and Finch 2008; Schmidtlein et al. 2008; Guillard-Gonçalves et al. 2014). PCA provides an opportunity to reduce the dimensionality of the multivariate data sets (Warner 2012). In large data sets with several variables, some of the variables may be positively or negatively correlated. For instance, some of the variables essentially contain the same information as the twenty-one vulnerability indicators that were used to derive composite vulnerability indexes in this study (Warner 2012). Therefore, PCA finds a new orthogonal coordinate system of uncorrelated predictors to represent the original vulnerability indexes data (Warner 2012). PCA was performed on the data used to derive vulnerability indexes in this study using the Statistical Package for the Social Sciences (SPSS) version 21 (SPSS 2012). Each component in PCA is a linear combination of the original variables used to derive vulnerability indexes (Warner 2012). The first principal component is in the direction of greatest variance in the original data set (Warner 2012). PCA produces eigenvalue and loading, which are, respectively, a sum of the squared loadings on a component and explain how strongly a variable is correlated with the component (Warner 2012). Loading of vulnerability indicators close to ±1 and greater than .5 was considered significant (Warner 2012). A correlation matrix between all the variables was derived to see the association between all the possible variables. PCA was performed with varimax rotation to obtain easily interpretable component loadings, and the components with eigenvalues greater than 1 were extracted (Warner 2012). A communalities test was also performed to see that how much of the variance in each of the original variables is explained by the extracted components (Warner 2012). Communalities for variables greater than 50 % were desired. A scree plot was also derived, which is a graphical presentation of the eigenvalues across the number of components (Warner 2012). Furthermore, the extracted components were named based on the higher loadings of the vulnerability indicators (Warner 2012). Since the number of indicators of exposure, sensitivity, and adaptive capacity were heterogeneous, we derived a ratio to name each component applying the following equation:

$${\text{Ratio}}\;{\text{to}}\;{\text{name}}\;{\text{the}}\;{\text{principal}}\;{\text{components (RNPC)}} = \frac{{{\text{Total}}\;{\text{number}}\;{\text{of}}\;{\text{indicators}}\;{\text{with}}\;{\text{high}}\;{\text{loadings}}\,(\!>0.40)}}{{{\text{Total}}\;{\text{number}}\;{\text{of}}\;{\text{indicators}}\;{\text{in}}\;{\text{the}}\;{\text{data}}\;{\text{set}}}}$$

Therefore, the components were named based on the highest RNPC values close to “1.”

2.3 The case study of Bihar state: a real-world case of a multi-stressor environment

2.3.1 Biophysical environment of the study area

Bihar, located in the GMB basin, is one of the worst As-affected states of India, sharing its geographic boundary with Bangladesh, West Bengal, and Nepal—three of the most As-affected vicinities of SA (Saha 2009). As contamination in the groundwater of Bihar is a phenomenon only recently discovered; it was first investigated in 2002 in the Semaria Ojha Patti village of the Shahpur block of the Bhojpur district (Chakraborti et al. 2003). The state still follows the old standard of 50 µg/L set for As in drinking water established by the World Health Organization (WHO 2004). Although the acceptable limit for As in drinking water is 10 µg/L, set by the Bureau of Indian Standards (BIS), however, in the absence of alternative sources, 50 µg/L of As is acceptable (BIS 2012). So far, out of a total of 82,000 groundwater samples tested for As contamination in 15 districts (of the total of 37 districts) of Bihar, 11 % exceed 50 µg/L, covering 57 community blocks in fifteen districts (Saha 2009). The concentrations of naturally occurring As in groundwater in Bihar (>1,000 µg/L) exceeded several times the As levels reported in groundwater in many countries such as Chile, Brazil, Mexico, Germany, Hungary, United Kingdom, and USA (Nordstrom 2002, Ghosh et al. 2005, 2007, 2008; Saha et al. 2009).

The spatial distribution of groundwater As in Bihar is irregular, and contamination occurs in patches. The hot spots (>50 µg/L) have been found to be confined within the Holocene newer alluvium of the thick multi-cyclic sand, clay, sandy clay, and silty clay sequence of a depth within 50 m below ground level, jeopardizing the hand pump-based rural drinking water supply in the state (Saha et al. 2009). The Pleistocene older alluvium was usually free from As contamination. Additionally, a high positive correlation between As and iron contamination has been found in the state, doubling the cost of filtration, and operation and maintenance of filtration equipment due to multiple metal contamination (Saha et al. 2010). Furthermore, the elevated As load is confined to the flood plain where rainfall facilitates percolation of organic carbon to the groundwater, which stimulates microbial respiration, triggering a reductive dissolution of As and iron in the solid phase (Mukherjee et al. 2012). These hydrogeochemical phenomena produce HCO3 − in shallow groundwater that helps mobilization of As in the groundwater (Saha et al. 2010). Geochemical analysis of As-contaminated groundwater in Bihar reveals that the contaminated groundwater was found to be near neutral to mildly acidic and dominated by alkaline earth (Ca2+ and Mg2+) and weak acid (HCO3 −) (Saha et al. 2008, 2010, 2011). The presence of high concentrations of HCO3 − in groundwater significantly contributes to the hydrogeochemical evolution of groundwater and trace metal mobilization in the area (Saha 2009). In a recent study, Saha and Shukla (2013) have explained that the As-contaminated groundwater in the state is primarily evidenced by three hydrogeochemical facies dominated by Mg and HCO3 (Fig. 2). The influence of redox conditions on As mobility in the groundwater has been widely reported in the literature (Ravenscroft et al. 2011). A mixed correlation between pH and As concentrations has been reported in the state (Singh et al. 2014). There are studies investigating the association between the depths of the tested hand pumps and the concentrations of As in groundwater in the state. However, none of them found any significant correlation between the two parameters (Saha and Shukla 2013; Singh et al. 2014).

Fig. 2
figure 2

Piper diagram showing different hydrogeochemical facies of As-contaminated groundwater in Bihar (Source: Saha and Shukla 2013)

Full size image

Although the state has rich groundwater, the fact is that shallow aquifers, on which the rural water supply is heavily dependent, are contaminated with As, and this situation makes it difficult to provide potable water in the affected areas (Saha and Shukla 2013). The state is endowed with 36 % replenishable groundwater resource for further extraction. The transmissivity of the aquifer (3,718 and 6,986 m2/day) indicates its higher potentiality than the aquifers in As-affected areas in WB (300–8,800 m2/day); hydraulic conductivity of the groundwater ranges from 64.88 to 82.00 m, indicating very good aquifer potential; and the deeper aquifer is protected by a middle clay, which may be developed for community drinking water supply by deep tube wells having a yield capacity of 150 m3/h (Saha et al. 2011). This hydrogeochemical information is valuable and could help creating mitigation strategies. However, further investigation in all the As-contaminated areas is warranted.

A toxic risk index ranging from .9 to 192.50 has also been derived in the state, exceeding the lower and upper end of the ranges of the typical toxic risk index 1.00, suggesting that the residents in the area might confront seriously adverse toxic health impacts (Singh and Ghosh 2012, Singh et al. 2014). The exposed communities were found to be consuming up to 1,469 µg/day of As against the maximum allowable limit of 200 µg/day through As-contaminated water and food materials (rice, wheat, maize, and lentils) in the state. Therefore, the cumulative effect is making children (57/1,000) susceptible to cancer with an average prevalence of skin pigmentation of 1.35 (Singh and Ghosh 2012, Singh et al. 2014). In a recent study, consumption of As in excess of 200 µg/kg through cooked rice in India has been linked with elevated genotoxic effect in human beings (Banerjee et al. 2013). A series of obstetric outcomes were also documented in women exposed to As-contaminated groundwater, and sixty persons with Arsenical skin lesions were reported in Bihar (Chakraborti et al. 2003). The As contamination zone is confined to the socioeconomically deprived communities of the state living along the river Ganges, making the exposed population (more than 8 million) highly vulnerable to its toxic effects (Singh et al. 2014).

2.3.2 Socioeconomic–demographic health conditions of Bihar

Globally, India is home to one of the largest populations of impoverished people, who are highly vulnerable to health problems (Cord et al. 2009). Bihar is the second poorest state in the country, where more than 33 million people currently live below the poverty line, with a monthly per capita income of only Rs. 354.36 (about USD $7.8), and more than 2 million people have little or no food security, resulting in dietary deficiencies that increase susceptibility to Arsenicosis (BRLPS 2007, Census 2013). The state has a total population density of 1,102 person/km2 and a decadal population growth rate of 28.43 %. The total literacy rate is only 48 %, due to inequitable access to education and a high dropout rate (BRLPS 2007). The impacted communities are largely unaware of the contamination problem because of poor literacy, especially for women whose literacy rate is only 33.53 %. Consequently, women lack equal access to resources; are more likely to be impoverished and illiterate; and are susceptible to abandonment, divorce, ostracism, or domestic violence when afflicted with Arsenicosis (BRLPS 2007; Brinkel et al. 2009). Fear of contamination has affected social ties and resulted in fewer marriages, causing cultural stress between generations (Bihardays 2011). Arsenicosis-affected individuals and communities also face social marginalization, including alienation, stigmatization, and discrimination. A typical community in an As-affected rural area is presented in Fig. 3.

Fig. 3
figure 3

Typical community in an As-affected area in Bihar (Photograph by Sushant Singh 2013)

Full size image

Additionally, the high infant mortality rate, kala-azar prevalence, TB incidence, HIV prevalence, frequency of diarrhea, and other diseases significantly contribute to vulnerability of communities. Furthermore, multiple environmental contaminants such as F, N, and other contaminants, as well as other environmental stresses such as flood and drought incidence, in addition to exposure to stressors like climate change or regional biophysical or political issues further amplify vulnerability to As.

2.4 Data collection and processing

Arsenic, fluoride, nitrate, and iron contamination data in different districts of Bihar were extracted from the published literature and field survey (SOES 2006; Ghosh et al. 2007, 2008; Saha 2009; Saha et al. 2009; Singh 2011). The at-risk population was calculated based on the number of blocks contaminated with As. In this study, we defined the threshold for population at risk at 10 μg/L of groundwater As contamination (Nickson et al. 2007). Flood and drought incidence was pooled from federal reports. Socioeconomic, demographic, and health-related data were extracted from the Census-2001 and policy documents. Hydrogeological data were extracted from the published maps and state reports (BSPCB 2007; GSI 2012; BAMETI 2014). All the data were standardized and processed according to the UNEP guidelines (UNEP 2003).

A Shapefile for Bihar with district boundaries and rivers available at the http://www.diva-gis.org/ was downloaded. All the data were incorporated into the attribute table, and maps were created using ArcGIS 10.1 (ESRI 2012).

3 Results and discussion

3.1 Arsenic vulnerability maps (AVM)

The AVM explicates the areas contaminated with As ranging from 0 µg/L to >1,000 µg/L of As (Fig. 4) and the population at risk due to ready access to the drinking water sources with As levels more than 10 µg/L (Fig. 5). Five districts were found to be contaminated with >1,000 µg/L of As including Buxar, Bhojpur, Patna, Samastipur, and Bhagalpur. Nine districts were found to have As below the BIS standard of 50 μg/L in the state (Fig. 4). A total of about 9 million population was found to be at risk in the state. The population in about one-half of the As-affected districts was found to be highly vulnerable to As contamination, amounting to a total population of about 4.4 million out of the total at-risk population (Fig. 5). The least at-risk population in the As-affected districts was found to be 7 % in Darbhanga district, covering only one block; however, the highest As-affected population (63 %) was found in Khagaria district, covering a total of three blocks (Fig. 5).

Fig. 4
figure 4

Arsenic contaminated districts of Bihar. The map is based on more than 30,000 drinking water sources tested for As concentrations, pooled from several published sources. The mean value of As exceed the BIS standards set for drinking water; however, the map represents the highest concentration detected in the areas (SOES 2006; Ghosh et al. 2007; Saha 2009; Singh 2011)

Full size image
Fig. 5
figure 5

Arsenic vulnerable population of Bihar

Full size image

3.2 Mapping exposure

3.2.1 Mapping environmental vulnerability

Only Bhagalpur district was found to be a highly environmentally vulnerable area. Bhagalpur is a unique As-affected area because the groundwater in this area is contaminated with As, F, and N. Also, the district is affected by flood and drought incidences (Fig. 6). Five districts Bhojpur, Buxar, Patna, Samastipur, and Vaishali were found to be moderately vulnerable, followed by the less vulnerable two districts and four at risk. Only three districts including Darbhanga, Lakhisarai, and Purnia were the resilient districts because of the comparatively less biophysical stressors.

Fig. 6
figure 6

Environmental vulnerability map of Bihar

Full size image

3.3 Mapping sensitivity

3.3.1 Mapping socioeconomic–demographic vulnerability

A total of five districts including Vaishali, Samastipur, Darbhanga, Purnia, and Katihar were found to be socioeconomically and demographically highly vulnerable (Fig. 5). Among these districts, only Samastipur falls in the 500–1,000 µg/L As contamination range (Fig. 4). Among the highly As-affected districts (>1,000 µg/L) and highly As vulnerable populations, only Patna district was found to be socioeconomically and demographically moderately vulnerable (Figs. 4, 7). The socioeconomic–demographic vulnerability is a function of the population below the poverty line, population growth, population density, SCs, and STs population (representing the lowest rungs of the caste hierarchy). The socioeconomic–demographic vulnerability map explicates that in spite of having a higher concentration of As and a highly vulnerable population size, the socioeconomic and demographic conditions of the exposed population reduced the vulnerability in the high As-contaminated areas and magnified the vulnerability in the areas with lower levels of As contamination. With more than 1,000 µg/L of As levels in the groundwater, Buxar, Bhojpur, and Bhagalpur districts were found to be socioeconomically and demographically at risk, which is the second level at the vulnerability index and only rank behind to reach the resilient stage, whereas the districts with As levels between 51 and 1,000 µg/L of As including Vaishali and Samastipur were found to be socioeconomically and demographically highly vulnerable. Moreover, Darbhanga, Purnia, and Katihar, which are comparatively less As-contaminated (<As 20 µg/L), were the other three highly vulnerable because of the poor socioeconomic and demographic conditions. Only two districts Lakhisarai (because of the lowest number of the most vulnerable population (STs) and the lowest population density) and Munger (because of the lowest population density in the areas) were found to be resilient districts in the state.

Fig. 7
figure 7

Socioeconomic–demographic vulnerability map of Bihar

Full size image

3.3.2 Mapping health vulnerability

The population of Bihar is affected by several diseases such as diarrhea, tuberculosis, filarial, polio, and kala-azar that are endemic in the state. This study demonstrates that Katihar and Patna districts were highly vulnerable due to very poor health conditions, followed by moderately vulnerable district Vaishali, less vulnerable districts Saran, Samastipur, Khagaria, and Purnia, and the at-risk districts Lakhisarai, Bhagalpur, Darbhanga, and Kishanganj (Fig. 8). In Katihar, the infant mortality rate (59 %) and the HIV prevalence (2.5 %) were very high, and in Patna, the tuberculosis incidence (8.5 %) was the highest among other districts (BRLPS 2007). In the moderately vulnerable district Vaishali, the kala-azar prevalence was the highest (11.4) followed by a high tuberculosis incidence of 4.3 (BRLPS 2007). People with poor health will be more susceptible to Arsenicosis; they lack the required physiological coping mechanism against any foreign agents in the body, therefore making them more vulnerable and sensitive to any additional health problems.

Fig. 8
figure 8

Health vulnerability map of Bihar

Full size image

3.3.3 Mapping geological vulnerability

Geologically, Darbhanga, Katihar, Khagaria, Kishanganj, and Purnia were found to be highly vulnerable as the lithology of the areas is shaped by only younger alluvium, which has been reported as the highly As-contaminated areas in the state. However, all these districts are on the northern side of the River Ganges and have not been investigated for groundwater As contamination (Figs. 4, 9). The moderately vulnerable districts were Buxar, Bhojpur, Saran, Vaishali, Samastipur, and Begusarai. All these districts except Begusarai have been found to be highly As-contaminated areas. The lithology of these districts is mainly shaped by the younger alluvial and calcareous alluvial soil.

Fig. 9
figure 9

Geological vulnerability map of Bihar

Full size image

3.4 Mapping adaptive capacity

Kishanganj, Purnia, and Katihar were found to be the highly vulnerable areas because of they have the lowest adaptive capacities (Fig. 10). The total literacy rates and the female literacy rates in these districts were far below the average literacy rates in the state (BRLPS 2007). Katihar and Purnia both had a total literacy rate of only 35.1 %, whereas Kishanganj had the lowest total literacy rate of only 31 % among all other districts (BRLPS 2007). Two highly As-affected districts Bhojpur and Patna, along with one of the lowest As-affected districts, Munger, were the resilient areas with higher literacy rates (Fig. 10). Other highly As-affected areas Buxar and Bhagalpur were at risk followed by Samastipur (Fig. 10).

Fig. 10
figure 10

Adaptive capacity map of Bihar

Full size image

3.5 Mapping composite vulnerability

A composite vulnerability map (CVM) is the map based on the average mean value of all the vulnerability indexes covered in this study. It is important to mention here that the VI values for exposure and sensitivity trended in a positive direction (the greater the VI, greater the vulnerability). However, for adaptive capacity, the VI values trended in a negative direction (the greater the VI, less adaptive capacity). The CVM elucidates that Katihar was the only district found to be a highly vulnerable area, followed by the moderately vulnerable areas of Vaishali, Samastipur, Khagaria, and Purnia; less vulnerable areas were Buxar, Patna, Begusarai, Bhagalpur, Saran, Darbhanga, and Kishanganj districts, respectively (Fig. 11). The Bhojpur and Lakhisarai districts were found to be the districts at risk. Only one district, Munger, was found to be resilient with comparatively higher adaptive capacities (Fig. 11). Surprisingly, between the two at-risk districts, Bhojpur was one of the highly As-contaminated districts, with As levels in the groundwater of more than 1,000 µg/L (Fig. 4).

Fig. 11
figure 11

Composite vulnerability map showing total vulnerability in Bihar

Full size image

Additionally, the study suggests that besides higher concentrations of a carcinogenic material (arsenic in this case), other biophysical, socioeconomic, demographic, educational, health, and environmental factors have a significant effect on the total vulnerability in the area. Therefore, a CVM could be a very important decision-making tool to evaluate the actual vulnerable areas as a function of composite vulnerability indicators. Among the highly As-affected districts, which include Buxar, Bhojpur, Patna, and Bhagalpur (As > 1,000 µg/L), Bhojpur dropped two levels on the vulnerability index scale and improved to be classified as an “at-risk” district (Fig. 11). The other three districts dropped one level to being less vulnerable areas (Fig. 11). These findings suggest that vulnerability of the area or the communities in the state do not depend entirely on the environmental stressors like groundwater contaminations or other environmental hazards. Other factors, including the socioeconomic and demographic status of the communities and geological properties of the areas, play a vital role in shaping the total vulnerability of the population and the region. These finding are in line with previous studies, which argue that there is a need for multiple indicators to adequately assess the vulnerability of areas impacted by natural hazards (Cutter et al. 2009).

3.6 Principal component analysis

According to the correlation matrix, a significant positive correlation between rural population, flood incidence, and geological formation was found (Appendix 1). This positive correlation suggests that flood incidences are more prevalent in rural areas, and the geological formation is of quaternary age. These factors will contribute to the vulnerability in the rural areas in the state. Furthermore, a very high positive correlation between BPL, kala-azar prevalence, TB incidences, flood incidences, lithology of the areas, and physiography of the areas was also derived (Appendix 1). A high positive correlation was found between the SC population and population density. Total literacy and female literacy were found to be significantly negatively correlated with the ST population. Population growth only had a strong negative correlation with total literacy but was found to be strongly positively correlated with the lithology and the physiography of the areas. This further explains that population growth is greater in the areas with younger alluvial soil and in the alluvial plain in the state. Population density showed a strong positive correlation with kala-azar prevalence and TB incidence (Appendix 1).

Communalities of one-third of the variables were close to 1 and were reasonably high for the rest of the variables, with only one variable drought incidence (.566) close to the lower end of the acceptable value (Appendix 2). The extracted principal components have been represented in Table 3.

Table 3 Eigenvalues of the principal components
Full size table

The eigenvalues in the extracted five components that were found to be greater than one ranged between 1.2 and 6.8 (Table 3). The first principal components explain about 33 % of the variance; the second components accounted for 21 % of the variance; the third with 14.3 % of the variance; the fourth components for 7.5 %; and the fifth components for 5.9 % of the variance (Table 3). Together, the first five components accounted for a cumulative 81.6 % of the variance in the vulnerability data set.

The first PC contrasts the total literacy, female literacy, F-affected districts, and drought incidence with other variables in the component. As discussed earlier, literacy rates (both total literacy rate and female literacy rate) reduce total vulnerability, as they constitute a very important component of the adaptive capacity of the communities. The extreme negative loadings of total literacy rate (−.942) and female literacy rate (−.930) reflect that the composite vulnerability is highly negatively affected by adaptive capacity (Table 4). The sensitivity indicators, lithology (.760) and infant mortality rate (.740) with comparatively higher loadings, indicate that the composite vulnerability is affected by the sensitivity of the areas. Other variables with loadings close to zero indicate an average contribution to composite vulnerability (Saha and Shukla 2013). The second PC was primarily marked by high positive loadings of eight sensitivity indicators with only one variable geological formation with negative loading (Table 4). The third component was marked by only the sensitivity indicators. The fourth component and the fifth component were marked, respectively, with only three elements and two elements with loadings more than .40.

Table 4 Loading matrix of the principal components
Full size table

The bend that appears on the scree plot also suggests that the variance will essentially be explained by the five PCs (Appendix 3). A loading of more than .40 of each variable in each component was interpreted and was considered in naming the components (Warner 2012). The first component had high loadings of a total of 16 variables; both the adaptive capacity indicators, total literacy rate and female literacy rate, had the highest loadings (Table 4). Furthermore, five exposure indicators out of a total of six and nine sensitivity indicators out of a total of thirteen indicators were the other elements in the first component, with more than .40 loadings. Based on the RNPC, the first component could be labelled “Adaptive Capacity”; the second and the third component as “Sensitivity”; and the fourth and the fifth component as “Exposure” (Table 4). The fifth component had only two variables with loadings more than .40, each from exposure and sensitivity. Considering the comparatively higher loading of the iron-affected district (.610) over the HIV prevalence (.514), we named the fifth component “Exposure.”

4 Conclusion

There exists a unique combination of biophysical and socioeconomic conditions in Bihar, making it highly vulnerable to the effects of groundwater As contamination. The CVM shows that Katihar is the only “highly vulnerable” district, followed by the “moderately vulnerable” areas of Vaishali, Samastipur, Khagaria, and Purnia, and the “less vulnerable” areas of Buxar, Patna, Begusarai, Bhagalpur, Saran, Darbhanga, and Kishanganj, respectively. Bhojpur and Lakhisarai districts were found to be “at risk.” Only Munger was found to be a “resilient” district with a comparatively higher coping capacity. Vaishali, Samastipur, and Katihar districts should be given priority in As-mitigation plans; low-cost As-mitigation technology(s) would be the best fit for Vaishali and Purnia followed by Darbhanga, Katihar, Samastipur, and Begusarai, respectively. The Bhojpur district would be an ideal As-affected district to begin with an As-mitigation plan.

The CVM of Bihar has three major potential uses: first, to identify the “highly vulnerable” areas where As mitigation is the most urgent requirement; second, to identify types of feasible and sustainable As-mitigation options for a specific vulnerable area; and third, to identify the area(s) where the likelihood of success of an As-mitigation plan is greatest (Fig. 11). Our study demonstrates that among the 15 As-affected districts, Katihar can be considered “highly vulnerable”; therefore, this district should be given priority in an As-mitigation plan where As mitigation is an urgent necessity. Although Katihar has very low levels of As concentration, socioeconomically and demographically, health wise, and geologically, it is one of the “highly vulnerable” areas in the state with the least adaptive capacity. As-mitigation policy should accord the next highest priority to Buxar, Patna, Begusarai, Bhagalpur, and Saran followed by other As-contaminated districts including Vaishali, Samastipur, and Khagaria. Although Bhojpur was among the highly As-affected districts, because of its high level of adaptive capacity, it was found to be “at risk.” Therefore, the likelihood of success of an As-mitigation program in this district is very high. Vaishali, Samastipur, Khagaria, and Purnia are the “moderately vulnerable” areas that are socioeconomically and demographically heterogeneous in nature and fall under the socioeconomically and demographically highly vulnerable areas in the state. Therefore, the expensive As-mitigation options would not work in any of these districts. Considering the adaptive capacity of these three “highly vulnerable” districts, Vaishali has a better adaptive capacity than the other districts. Therefore, among the “moderately vulnerable” districts, Vaishali should be given priority, followed by Samastipur with low-cost As-mitigation policies.

The PCA suggests that the five principal components essentially explain the variance in composite vulnerability. The total literacy rate, the female literacy rate, rural population, population growth, population below the poverty line, scheduled caste population, infant mortality rate, flood incidence, drought incidence, As, F, N, lithology, and the geological formation were found to be the most important variables to explain the variance of composite vulnerability. The first component was the adaptive capacity, which further suggests that the adaptive capacity of the communities makes relatively greater contributions to reduce the total vulnerability. Moreover, other indicators of adaptive capacity could be identified and studied to derive total vulnerability in areas under multiple environmental and social stressors. For instance, additional information about people’s risk perception of the As problem, the presence and functionality of institutions, interpersonal trust, and trust in institutions working in those areas would be very helpful in understanding the adaptive capacity of the exposed communities. In our ongoing study, we gathered household-level information about these indicators along with other socioeconomic and demographic variables in three As-affected villages in Bihar, India, which would be incorporated in the proposed vulnerability framework, and a CV could be derived at the household level. This will make the CV decision-making tool more precise. The empirical data from these three As-affected villages would help calibrate the CVM and will testify to the predictive capacity of the CVM.