Abstract
Water scarcity is an emerging multidimensional issue concerning not only the physical availability of resources but also is linked with poverty. The existing literature has established a relationship between income poverty and water poverty. In the Indian context, various studies have explored such issues using the Water Poverty Index (WPI), but only a few have analyzed downscale spatial units. This paper constructs district-level water poverty measures and maps its spatial heterogeneity for Maharashtra, India. Using an indicator-based approach, we aggregate various dimensions of water poverty into a single index. This composite index is formulated by normalizing the indicators and assigning weights using principal component analysis. After rescaling, the aggregate WPI score varies from 0 to 1, denoting lower to higher water poverty. The overall WPI estimate of Maharashtra is 0.47, implying high water stress. The study presents district-wise WPI information by combining the results with Geographic Information System (GIS). Our findings suggest that along with the physical abundance and accessibility to water, improvement in the determinants of capacity and environment is essential to tackle water poverty. Results highlight the intra-district variations among components of water poverty, indicating the importance of local-scale policy-making for better water resource management.
HIGHLIGHTS
Water stress is an increasing phenomenon in Indian states.
A district-scale Water Poverty Index is constructed for Maharashtra, India.
The regional disparities of water poverty in Maharashtra are analysed using spatial distribution and hotspot mapping.
Out of 33 selected districts for the study, around 22 are water-poor.
Results highlight the role of all five components of the Water Poverty Index.
INTRODUCTION
About 4 billion people are estimated to live with severe water scarcity conditions for at least one month in a year (Burek et al., 2016). Water scarcity, exacerbated by climate change, could cost some regions up to 6% of their gross domestic product (GDP), spur migration, and spark conflict (World Bank, 2016). In the context of arid and semi-arid areas of developing countries, climate change, water scarcity, and uneven distribution of wealth are interlinked, given their dependence on agriculture-based livelihood occupations (Salameh, 2000; Rijsberman, 2003). The nexus between water scarcity and poverty is complicated and less understood but essential for informed policy-making (Sullivan, 2001; Biltonen & Dalton, 2003). The role of water in fighting poverty has put water issues at the central agenda of various national and international platforms. Since the last decade, water scarcity conditions have worsened across India, including the wealthiest state, Maharashtra. According to Sen (1999), poverty results from capability deprivation. A healthy life with proper education and access to livelihood resources are the fundamental prerequisites for a wholesome life. Water is a production factor; hence, water deprivation indicates poverty (Sullivan et al., 2003; Kini, 2017). Cullis & Regan (2004) mentioned water poverty measures considering not only the shortage of water but also the insufficient adaptive social capacity to deal with such a situation. Improved water resource management and access to safe water and sanitation are essential for eradicating poverty and building peaceful and prosperous societies (WWAP, 2019). Escaping poverty requires access to water; thus, income and water poverty are essentially interconnected.
Notwithstanding, giving access to safe and clean water may not eradicate poverty and hunger permanently. However, it is necessary to sustain minimum living standards (Sullivan, 2002; Sullivan et al., 2003). Water scarcity emerges due to various physical, economic, financial, institutional, and political constraints. Molle & Mollinga (2003) have specified multiple dimensions of water scarcity: (i) the physical constraints are due to limited natural water sources; (ii) economic constraints result from a lack of human and financial resources; (iii) managerial induced scarcities arise with improper management of water systems; and (iv) institutional scarcity arises from the inability to address the demand-supply gap due to political subjection. Such scarcities may also vary in their temporal forms, which can be temporary or last longer, where climate change and/or its variability can be the critical factors. While water scarcity studies across different countries at macro scales have been highlighted, water scarcity is essentially a local or regional problem that shapes poverty in the region. Water availability is highly variable on a spatial and temporal scale (Sullivan & Meigh, 2007). The geographic attributes guide water poverty more than income poverty, given the crucial role of environmental factors. Furthermore, local water infrastructure development determines resource availability and people's ability to access it (Cullis & Regan, 2004). Hence, approaches to creating water-related livelihoods must be location-specific (Rijsberman, 2003).
Considering the varying scales, water poverty studies in India have emerged in some of the recent research contributions. Regarding assessing water poverty using index methods, Goel et al. (2020) measured water stress for 20 states of India using the WPI and found that most states are water-stressed, demanding attention on a priority basis. Prabha et al. (2020) computed the WPI at the city scale across 42 million-plus urban Indian agglomerations and identified that all the cities have performed poorly across the five components of WPI, requiring urgent government intervention. Chopra & Ramachandran (2021) used data envelopment analysis to study the water situation in 11 significant states of India, suggesting that geopolitical variation requires targeted water policies considering both physical availabilities and socioeconomic capacity. Yadav & Ibrar (2022) used the WPI to identify critical districts in Bihar facing water stress and flood damage. Pandey et al. (2022) argue that though the overall water situation has improved among Indian states from 2012 to 2018, the performance is poor across all five components of WPI, particularly resource availability and environment.
Earlier, on the other hand, Wilk & Jonsson (2013) approached participatory methods to assess water poverty at the community level in the state of Madhya Pradesh. They argued that highly aggregated data hide real disparities, thus making it challenging to target the vulnerable. While emphasizing the local influence on the methodology, this study highlights the importance of appropriate spatial representation and gender-sensitive assessment. Later, Juran et al. (2017) and Prince et al. (2021) used village-level survey methods to study water poverty in two tsunami-hit districts (Nagapattinam and Karaikal). The studies presented an advanced methodology to measure water access that supports statistical and spatial analyses at the community level, giving a fair picture of water poverty.
Nonetheless, index-based studies have their merits as they incorporate existing statistical data sets from census and local records to assess water stress in low-cost settings (Sullivan, 2002). Arguably, data collection at the community or regional level is not always cost-effective. Besides, water poverty assessment at a community scale through a participatory approach may encounter bias from the researcher's subjective view of the situation if time and finances do not allow for meeting stakeholders or conducting extensive surveys (Komnenic et al., 2009).
In this study, we combine five components – resources, access, use, capacity, and environment – to compute a Water Poverty Index (WPI) in the districts of Maharashtra. The index would substantially reflect the link among the hydro-climatic, environmental, and socioeconomic aspects affecting accessibility to safe water and associated poverty. The WPI deals with this heterogeneity in the physical availability of water and socioeconomic differences, i.e., access to water within a community or in an area, given their income level and education. Although ecological sustainability and affordability require consideration in exercising water provisions through desalination and import (Cronin et al., 2014) to the water-poor regions, such variability is the essence of water poverty estimates to employ adequate policy formulations. The districts of Maharashtra can represent the local scale, helping the policymakers to target impoverished areas.
Moreover, as water is a state subject, measuring water poverty at the district level enables us to examine how multi-tier governments would address district-level water poverty. There is a dearth of studies at the district scale in India. This paper fills this research gap by constructing district-level water poverty measures (for 33 districts based on available data) and mapping the spatial heterogeneity for Maharashtra in India. The paper is structured as follows: the second section discusses the study area's brief profile. Data sources and methodology adopted to develop the water index are provided in Section 3. Section 4 comprises the results, followed by the discussion of our findings along with policy recommendations in Section 5. The last section concludes the study.
DESCRIPTION OF THE STUDY AREA
Maharashtra is the country's wealthiest state with the highest gross state domestic product (GSDP) and contributes around 15% of GDP. The state contributes more than 55% of the service sector, followed by industry, agriculture, and allied activities. The capital city, Mumbai, alone contributes around 6% to the nation's GDP. Nevertheless, the intrastate disparities are contrasting. In this relatively better-urbanized state, water supply has improved over the years, yet only 50% of households have access to tap water. Among these households, treated water is supplied to only 32% and less than 42% of the rural population gets access to drinking water within their premises. However, it can be argued that access does not, by any means, ensure a safe and continuous water supply (Khambete, 2020).
Furthermore, Maharashtra has vast economic, social, and climatic differences across the state. Despite substantial rainfall and major rivers flowing through the state, rising population, shrinking surface water bodies due to pollution, declining water quality, and poor water management have accentuated severe water scarcity. In light of the above context, constructing a multidimensional index seems crucial to assess district-level water poverty.
RESEARCH STRATEGY: MATERIAL AND METHODS
Water Poverty Index: A conceptual overview
Over the past three decades, research has seen many indices developed for quantifying water resource vulnerability – water scarcity and stress – based on human water requirements and available water resources (Brown & Matlock, 2011). The most prominent measure of water stress is the Falkenmark indicator (1989). It is the fraction of the total annual runoff available for human use. The indicator calculated water usage per person in various countries and proposed thresholds: no stress, stress, scarcity, and absolute scarcity. Above 1,700 m3, a nation has no water stress, but below 1,000 m3 per capita per year, there are chances of losing human capital and hampering economic development with the rise in population (Feitelson & Chenowith, 2002). However, this indicator cannot be used on a smaller scale, as the indicator uses country-level data, where local information like regional variability or cultural differences are lost while averaging. Another drawback of this water stress indicator is that it considers only the physical availability of water and does not deal with water quality or utilization of available resources. Gleick (1996) developed an index focusing on water use rather than physical availability called basic human water requirements. The index measures water required to meet basic human needs and maintain hygiene, minimum drinking water, and basic water requirements for sanitation, bathing, and food preparation. Gleick proposed a total demand of 50 L per person daily as the water requirement to meet basic human needs. However, on the downside, it ignores agricultural and industrial water uses, constituting a substantial portion of human water use. Meigh et al. (1999) presented an improvement over this index in the Global Water Availability Assessment model. This index included agriculture, industrial, and domestic water uses to compare total water demand, where physical water resources include groundwater and surface water.
However, these above-stated studies, being one-dimensional, have limitations in integrating different aspects of water, such as availability, use, supply, and quality, to generate a comprehensive index (Jemmali & Matoussi, 2013). OhIsson (2000) presented the Social Water Stress Index, where the UNDP Human Development Index (HDI) functions as a weighted measure of Falkenmark indicator to understand the ability to adapt to water stress. Subsequently, the term ‘water poverty’ is also used as an indicator. Salameh (2000) defined the measure as the ‘ratio of the amount of available renewable water to the amount required to cover food production and the household uses of one person in one year under the prevailing climatic conditions’. However, this index does not consider the socioeconomic reasons for water shortage.
Sullivan (2001, 2002) presented a holistic approach to creating an aggregate index, WPI, which is cost-effective to generate, can be developed locally, and further scaled up to the regional or national level. The WPI is a composite index incorporating significant water issues into a single number. The WPI, suggested by Sullivan (2002), is based on five components, as represented in Table 1.
Description of WPI components (adapted from Sullivan et al., 2003).
WPI components . | Description . |
---|---|
Resources | Considering its variability and quality, the total physical availability of water (surface and ground water) |
Access | Access to the safe water source, reflecting the total distance covered and time spent on water collection per household. It also represents access to water for irrigation and industrial purposes |
Capacity | Ability to manage water from improved sources. This represents lobbying capacity for water based on income level, education, health, and infrastructure |
Use | Water consumption by different sectors, i.e., agriculture, industry and domestic |
Environment | Impact of water management on different ecosystems to safeguard ecological integrity |
WPI components . | Description . |
---|---|
Resources | Considering its variability and quality, the total physical availability of water (surface and ground water) |
Access | Access to the safe water source, reflecting the total distance covered and time spent on water collection per household. It also represents access to water for irrigation and industrial purposes |
Capacity | Ability to manage water from improved sources. This represents lobbying capacity for water based on income level, education, health, and infrastructure |
Use | Water consumption by different sectors, i.e., agriculture, industry and domestic |
Environment | Impact of water management on different ecosystems to safeguard ecological integrity |
Globally, this composite index has been applied at different scales to assess the water poverty scenario. Using national-level data, Lawrence et al. (2002) ranked 147 countries for international comparison using the five components suggested under Sullivan's WPI. Ladi et al. (2021) assessed the impact of various WPI components on HDI in different parts of Iran. Alqatarneh & Al-Zboon (2022) estimated the WPI for Jordan and also used it as an indicator to evaluate current water strategies. Few studies have been conducted on the basin/district scale (Komnenic et al., 2009; Van Ty et al., 2010; Li et al., 2011; Manandhar et al., 2012; Koirala et al., 2020; Liu & Liu, 2021). Their findings comprehensively capture the linkage between water and economic poverty and suggest the need for location-specific policy interventions. Yuan et al. (2023) established a multidimensional water poverty framework for the Yangtze River economic belt in China, considering components of water technology, water assets and welfare, water resources, and water environment and management. Besides, at the community scale (Sullivan et al., 2003; Korc & Ford, 2013; Jaren & Mondal, 2021; Stewart et al., 2022), these studies show that the WPI is an effective indicator for tackling water scarcity. This comprehensive yet straightforward approach to measure water poverty highlights huge spatial variation among the five components of the WPI, which are constructed at different scales.
Scale selection is a significant issue for computing WPI, as upscaling or downscaling can lead to severe errors (Gibson et al., 2000). Indices formed at the national level fail to explain regional variations, whereas regional indices eliminate the distinctness between rural and urban populations (Sullivan, 2002). Therefore, scale-specific policy interventions are necessary to improve water poverty. Culis & O'Regan (2004) advocated developing WPI at a meso scale between a macro or national scale and a micro or community scale. Furthermore, the basin-scale WPI can give a moderate representation of spatial variability across the nation (Sullivan & Meigh, 2003). However, disconnection between hydrological and socioeconomic data may impede the outcome (Sullivan et al., 2006) since various water management decisions are taken at the municipality/district level, which is the unit where census and other socioeconomic data are generally collected. Hence, the WPI can be applied at the district scale most cost-effectively. Sullivan et al. (2003) noted that extreme local variations must be monitored locally, or the information on local diversity is lost while averaging.
Despite the usefulness of WPI and its far-reaching application, the index construction has some major conceptual weaknesses. Garriga & Foguet (2010) highlight its shortcomings in combining available data, poor statistical properties of the resultant index, and a high correlation between the final index and other indicators. The available data mainly govern the choice of indicators. Implausible assignment of weights and loss of information while aggregating various indicators emerged as significant drawbacks (Feitelson & Chenowith, 2002; Nardo et al., 2005). Furthermore, Garriga & Foguet (2010) presented an improved method to calculate the WPI, popularly known as a modified WPI, later adopted by various studies to estimate water poverty (Garriga & Foguet, 2011; Jemmali & Matoussi, 2013; Jemmali & Sullivan, 2014; Jemmali, 2017).
Research framework to construct WPI
Indicator selection
The choice of indicators in this study is based on a substantial literature review of relevant conceptual and empirical studies representing the five components of WPI: resources, access, capacity, use, and environment. Table 2 illustrates indicators used in the WPI literature over the years at different scales. It is evident from the table that indicators under various components do not necessarily belong to any particular component. Studies have adopted indicators and modified them as required. Subsequently, based on the available data, our study uses 16 indicators under the five components mentioned above to quantify the WPI. The selected indicators and their data sources are explained in Table 3.
Review of literature on WPI components and indicators (see Supplementary material for the indicator terminology).
Sr. no. . | Authors . | Unit of analysis . | Resource (R) . | Access (A) . | Capacity (C) . | Use (U) . | Environment (E) . |
---|---|---|---|---|---|---|---|
1. | Lawrence et al. (2002) | Cross-country scale across 147 nations | R1, R2 | A1, A2, A3 | C1, C2, C3, C4 | U1, U2 | E1, E2, E3, E4, E5 |
2. | Sullivan et al. (2003) | Community | R3, R4, R5 | A2, A4, A5, A6, A7, A8 | C2, C5, C6, C7, C8, C9 | U1, U3, U4, U5 | E6, E7, E8 |
3. | Komnenic et al. (2009) | River basin riparian nations | R1, R2 | A1, A2, A3 | C2, C3, C4 | U1, U2 | E1, E2, E3, E4, E5 |
4. | Van Ty et al. (2010) | River basin | R3, R4, R6 | A1, A2 | C6, C10, C11, C12 | U1, U2 | E2, E9 |
5. | Garriga & Foguet (2010) | District | R3, R4 | A4, A7, A9, A10, A11 | C13, C14, C15, C16, C17, C18 | U1, U4, U6, A5 | E1, E2, E5, E10 |
6. | Li et al. (2011) | River basin | R3, R4 | A2, A4, A8 | C2, C6, C9, C19 | U1, U3, U5 | E1, E6, E8 |
7. | Manandhar et al. (2012) | River basin | R3, R4 | A4, A9 | C6, C19, C20, C21 | U1, U3 | E2, E9 |
8. | Jemmali & Matoussi (2013) | National | R3, R4 | A1 | C6, C19, C22, C23 | U1, U3 | A2, E1, E3 |
9. | Korc & Ford (2013) | Community | E1 | A4, A9, A12 | C6, C19, A10 | U7 | E13 |
10. | Jemmali (2017) | Cross-country scale in Africa | R2, R3, R4 | A1, A2, A14, U3 | C1, C19, C25 | U1, U2 | E1, E2, E9 |
11. | Shalamzari & Zhang (2018) | District | R3, R4 | A1, A2 | C6, C9, C11, C24 | U1, U3, U4, U5 | E8, E11, E12 |
12. | Goel et al. (2020) | State | R3, R4 | A2, A3, A4, A7, A13 | C6, C19 | C8, U3 | E2, E9, R5 |
13. | Prabha et al. (2020) | City | R3, R4 | A4, A9 | C5, C6, C10 | U1, U6, A5 | R5 |
Sr. no. . | Authors . | Unit of analysis . | Resource (R) . | Access (A) . | Capacity (C) . | Use (U) . | Environment (E) . |
---|---|---|---|---|---|---|---|
1. | Lawrence et al. (2002) | Cross-country scale across 147 nations | R1, R2 | A1, A2, A3 | C1, C2, C3, C4 | U1, U2 | E1, E2, E3, E4, E5 |
2. | Sullivan et al. (2003) | Community | R3, R4, R5 | A2, A4, A5, A6, A7, A8 | C2, C5, C6, C7, C8, C9 | U1, U3, U4, U5 | E6, E7, E8 |
3. | Komnenic et al. (2009) | River basin riparian nations | R1, R2 | A1, A2, A3 | C2, C3, C4 | U1, U2 | E1, E2, E3, E4, E5 |
4. | Van Ty et al. (2010) | River basin | R3, R4, R6 | A1, A2 | C6, C10, C11, C12 | U1, U2 | E2, E9 |
5. | Garriga & Foguet (2010) | District | R3, R4 | A4, A7, A9, A10, A11 | C13, C14, C15, C16, C17, C18 | U1, U4, U6, A5 | E1, E2, E5, E10 |
6. | Li et al. (2011) | River basin | R3, R4 | A2, A4, A8 | C2, C6, C9, C19 | U1, U3, U5 | E1, E6, E8 |
7. | Manandhar et al. (2012) | River basin | R3, R4 | A4, A9 | C6, C19, C20, C21 | U1, U3 | E2, E9 |
8. | Jemmali & Matoussi (2013) | National | R3, R4 | A1 | C6, C19, C22, C23 | U1, U3 | A2, E1, E3 |
9. | Korc & Ford (2013) | Community | E1 | A4, A9, A12 | C6, C19, A10 | U7 | E13 |
10. | Jemmali (2017) | Cross-country scale in Africa | R2, R3, R4 | A1, A2, A14, U3 | C1, C19, C25 | U1, U2 | E1, E2, E9 |
11. | Shalamzari & Zhang (2018) | District | R3, R4 | A1, A2 | C6, C9, C11, C24 | U1, U3, U4, U5 | E8, E11, E12 |
12. | Goel et al. (2020) | State | R3, R4 | A2, A3, A4, A7, A13 | C6, C19 | C8, U3 | E2, E9, R5 |
13. | Prabha et al. (2020) | City | R3, R4 | A4, A9 | C5, C6, C10 | U1, U6, A5 | R5 |
WPI components, indicators, data source, and period adopted for the study.
Components . | Sub-components . | Indicators . | Data sources . | Period . |
---|---|---|---|---|
Use | Water consumption by different sectors | Domestic (ha-m) & Industrial (ha-m); U1 Irrigation (ha-m); U2 | Department of Water Resources, RD and GR, Ministry of Jal Shakti, GOI | 2017 |
Access | Access to safe drinking water | HH with improved drinking water sources (%); A1 | National Family Health Survey-4 Department of Agriculture, Maharashtra and Input Survey, Agriculture Census, Ministry of Agriculture and Farmers Welfare, India National Family Health Survey-4 | 2015–2016 2016–2017 2015–2016 |
Access to sanitation Access to irrigation Time to obtain water (round trip) | HH with improved sanitation facilities (%); A2 The ratio of net irrigated area to net sown area; A3 Within premises, Less than 30 min and 30 min or longer; A4 | |||
Capacity | Economic | Per capita GDVA at constant prices – 2011–2012; C1 | Economic Survey of Maharashtra | 2019 |
HH with electricity (%); C2 | National Family Health Survey-4 | 2015–2016 | ||
Social | Infant mortality rate (%); C3 | District Socioeconomic Review, Directorate of Economics and Statistics, GoM | 2020 | |
Literacy rate (age 15–49); C4 | National Family Health Survey- 4 | 2015–2016 | ||
Resources | Water availability | The variance of groundwater level (sq. mm); R1 | Department of Water Resources, RD and GR, Ministry of Jal Shakti, GOI | 2018 |
The variance of surface water (BCM); R2 | 2000–2020 | |||
Seasonal Variation | Coefficient of variation of rainfall (30 years); R3 Evapotranspiration (mm); R4 | 1990–2019 2018 | ||
Environment | Vegetation | Forest cover (% of geographical area); E1 | India State of Forest Report, Forest Survey of India, Ministry of Environment, Forest, and Climate Change, GOI | 2019 |
Pollution | Fertilizers used per hectare of cultivable land; E2 | District Socioeconomic Review, Directorate of Economics & Statistics, GoM and Department of Agriculture, Maharashtra | 2020 |
Components . | Sub-components . | Indicators . | Data sources . | Period . |
---|---|---|---|---|
Use | Water consumption by different sectors | Domestic (ha-m) & Industrial (ha-m); U1 Irrigation (ha-m); U2 | Department of Water Resources, RD and GR, Ministry of Jal Shakti, GOI | 2017 |
Access | Access to safe drinking water | HH with improved drinking water sources (%); A1 | National Family Health Survey-4 Department of Agriculture, Maharashtra and Input Survey, Agriculture Census, Ministry of Agriculture and Farmers Welfare, India National Family Health Survey-4 | 2015–2016 2016–2017 2015–2016 |
Access to sanitation Access to irrigation Time to obtain water (round trip) | HH with improved sanitation facilities (%); A2 The ratio of net irrigated area to net sown area; A3 Within premises, Less than 30 min and 30 min or longer; A4 | |||
Capacity | Economic | Per capita GDVA at constant prices – 2011–2012; C1 | Economic Survey of Maharashtra | 2019 |
HH with electricity (%); C2 | National Family Health Survey-4 | 2015–2016 | ||
Social | Infant mortality rate (%); C3 | District Socioeconomic Review, Directorate of Economics and Statistics, GoM | 2020 | |
Literacy rate (age 15–49); C4 | National Family Health Survey- 4 | 2015–2016 | ||
Resources | Water availability | The variance of groundwater level (sq. mm); R1 | Department of Water Resources, RD and GR, Ministry of Jal Shakti, GOI | 2018 |
The variance of surface water (BCM); R2 | 2000–2020 | |||
Seasonal Variation | Coefficient of variation of rainfall (30 years); R3 Evapotranspiration (mm); R4 | 1990–2019 2018 | ||
Environment | Vegetation | Forest cover (% of geographical area); E1 | India State of Forest Report, Forest Survey of India, Ministry of Environment, Forest, and Climate Change, GOI | 2019 |
Pollution | Fertilizers used per hectare of cultivable land; E2 | District Socioeconomic Review, Directorate of Economics & Statistics, GoM and Department of Agriculture, Maharashtra | 2020 |
Note: ha-m: hectare metre, BCM: billion cubic metres, min: minutes, mm: millimetres, and sq.mm: square millimetre. RD: River Development; GR: Ganga Rejuvenation; GoI: Government of India; GoM: Government of Maharashtra.
Description of components, sub-components, and indicators
Resources: Under this category, we considered using two sub-components: the physical water availability and seasonal variability (as a measure of reliability). Groundwater (fluctuations from its mean levels) and surface water for 20 years is used as indicators to measure the physical availability of water. Considering only the physical availability of water as an indicator of water stress is inadequate (Shalamzari & Zhang, 2018). The variability of water over space and time has been used as a sub-component to measure the reliability of available water resources. Furthermore, evapotranspiration is another important variable that significantly affects the water availability in arid and semi-arid regions (Salameh, 2000). The temporal and seasonal variability has been measured in this study using the indicators, the coefficient of variation in rainfall as a proxy to estimate variability (Manandhar et al., 2012) for 1990–2019 and evapotranspiration.
Access: Under this component, we considered four sub-components: access to safe drinking water, access to sanitation, irrigation, and time to collect water. Access to safe drinking water and sanitation are measured by the percentage of households with access to safe drinking water and in-house sanitation. Time to collect water is used as one sub-component as inadequate access to water forces household members, especially women and children, to spend hours fetching water from far-off sources. Access to improved water sources would not be sufficient if the availability is not within a reasonable distance. Therefore, the time spent to obtain water (round trip) in minutes is used as a sub-component. This variable is categorized namely within premises (if water is delivered/available at the household), less than 30 min (if time spent on water collection is less than 30 min in a day), and 30 min or more (highlights that water source is far away from the household). Furthermore, we used access to irrigation, calculated as the ratio of net irrigated area to net sown area, as an indicator.
Use: This component has sectoral water consumption as a sub-component (Sullivan & Meigh, 2003). Hence, to measure this sub-component, we used water consumption for domestic, industrial, and irrigation purposes as the indicators.
Capacity: Capacity is defined as the ability of the person to demand better water services (Prabha et al., 2020). Broadly, two sub-components are used to quantify this component: (a) economic capacity explained by purchasing power and infrastructure and (b) social capacity to spread awareness about water issues (Manandhar et al., 2012), measured by literacy rate and infant mortality rate, respectively. The economic capacity sub-component is explained using per capita gross district value added (GDVA) at constant prices, 2011–2012, and the percentage of households with electricity as indicators. The social capacity is measured by the literacy rate and infant mortality rate, respectively. A higher literacy rate indicates a higher ability to understand the importance of safe water and sanitation facilities and the ability to negotiate for the same.
Environment: Water poverty increases with environmental deterioration. Sustainability can only be achieved if improvements in the water system do not lead to environmental damage (Sullivan et al., 2003). Some resources cannot be reproduced with physical and financial capital. At the same time, infinite resource recycling is impossible. We used two sub-components: vegetation and pollution, proxied by forest cover (% of total geographical area), and fertilizer consumption, respectively.
Water quality is another significant aspect of maintaining overall ecological health. Water availability and its standard quality can satisfy the water needs of the environment. Calculating the WPI must include water quality information, ensuring that the relevance of WPI is sustained over time (Sullivan, 2002). However, these data are unavailable for all the districts of the study area and hence are dropped as an indicator. Also, data for two urban districts, i.e., Mumbai city and Mumbai suburban, are missing for four indicators: access to irrigation, water use by domestic and industrial sectors, water use by irrigation, and environmental pollution. Therefore, these two districts are not incorporated into the study.
Normalization of data
As indicators are measured in different units, we normalized our data to make it comparable. Normalizing indicators also ensure the identification of outliers in the overall index. Moreover, a few indicators used in constructing the composite index are positively related to the final index, while others are negatively correlated. Thus, it is a crucial step, and any movement in the normalized indicator influences the final index (Mazziotta & Pareto, 2013). In this study, we have used Z-score standardization for individual indicators and a rescaling (min–max) approach to normalize the final composite WPI.
Issue of weights and aggregation
The indicators must be aggregated using some weighting scheme to compute the final WPI. With 16 indicators across five components, this study uses principal component analysis (PCA) to assign weights. The efficacy of any final composite indicator depends on the choice of correct weights, reflecting the relative importance of each component and sub-component. The most common practice is to assign equal weights to all components; hence each component gets the same relative importance in the final index. Despite its ease and simplicity, an equal weighting scheme fails to capture the difference between the significant and insignificant components. Also, there is a higher probability of suffering from double counting due to any correlation among variables. To overcome this shortcoming, other weighting techniques include participatory and statistical methods, depending on the structure and quality of data. Participatory methods seek the opinion of various stakeholders for assigning weights. With many indicators, stakeholders cannot achieve consensus; hence, the method fails. The associated subjectivity also affects the final index since the weights are arbitrary and based purely on the developer's viewpoint. Subjectivity in weight selection can be handled by more objective statistical methods, commonly known as data-driven techniques such as correlation analysis, regression analysis, PCA, or factor analysis (Greco et al., 2019).
PCA is a data reduction technique that creates factors (or principal components (PCs)) by linearly transforming an extensive data set into smaller components for meaningful interpretations. It is an orthogonal linear transformation to convert correlated variables into a set of uncorrelated variables reflecting prominent variations in the original data. The central objective is to explain the maximum variance in the original data set with few components. Abdi & Williams (2010) argue that the goals of PCA are to extract the most crucial information from the data table, compress the data set by keeping only the important information and simplify the description of the data set. PCA extracts the PCs and uses these as new variables. Extracted factors (or PCs) are a linear combination of variables. The first PC is a linear combination of variables explaining the maximum amount of variance, followed by the second PC explaining the next most significant amount of variance, not correlated with the first component, and so on. PCA uses factor loadings of the extracted PCs as weights for the indicators. Thus, the set of weights explains the maximum variation in the original variables. However, weighting only intervenes to correct the overlapping information of two or more correlated indicators. It is not a measure of the theoretical importance of the associated indicator. This method cannot obtain weights without a correlation between the indicators (Nardo et al., 2005). Therefore, statistical weights cannot be interpreted as always reflecting the decision-makers' priorities (Garriga & Foguet, 2010).
The next step is aggregating these weighted indicators. The most common aggregation method is the additive technique, i.e., weighted arithmetic means of indicators. However, linear aggregation suffers from a major drawback of compensability among indicators. Hence, poor performance by some indicators cannot be penalized fully (Munda & Nardo, 2005; Nardo et al., 2005). It assumes preferential independence among indicators, i.e., no synergies or conflicts exist among different components of the WPI. Consequently, theoretical inconsistency appears in how weights are used and their real theoretical meaning (Garriga & Foguet, 2010). The most appealing solution to this issue is adopting the geometric (multiplicative) aggregation technique, where poor performance by some indicators is penalized more heavily (Nardo et al., 2005). This aggregation method permits limited compensability, which restrains full compensation for indexes with lower scores (OECD, 2008).
Water Poverty Index construction
After selecting indicators, it is crucial to understand the linear relationship among indicators under each component. We have used SPSS software (version 20) to perform PCA. Before this, we examine the correlation across the indicators employing Pearson's correlation, and highly correlated variables are dropped to avoid redundancy in the final composite index. PCA requires some degree of correlation among variables, but when correlations are too low, the application of PCA becomes questionable (Hair et al., 2009). Only sufficiently related variables generate representative components. The remaining indicators are standardized using Z-score normalization to prevent the undue influence of any variable on PCs. In addition to the Pearson correlation coefficient, we use two statistical methods to determine the degree of interrelationship and the appropriateness of PCA. The Bartlett test of sphericity checks statistically significant correlations among variables.
Garriga & Foguet (2010) suggested applying a multivariate statistical technique such as PCA at the sub-components level to check whether the selected indicators for the WPI are statistically well balanced. To avoid double counting, a high correlation between these sub-components must be evaluated before calculating the final index value (Nardo et al., 2005). However, at the same time, the data set should be sufficient to give a comprehensive picture of the water situation (Jemmali & Sullivan, 2014). However, the prime motive behind this step is to reduce a large set of correlated indicators into a group of a few uncorrelated components without losing too much information. This study tried the application of PCA at the sub-index level for five components. Still, the KMO measure of sample adequacy falls outside the acceptable range of 0.5–0.7, except for access (0.59) and resources (0.57). Also, only one PC is extracted from each sub-index after performing PCA, which explains less than 80% of the total variation, whereas the variance explained criterion should keep enough factors to account for 80% of the total variation (Nardo et al., 2005). Therefore, after evaluating the correlation between sub-components and dropping those highly correlated indicators, we perform PCA only at the index level. The eigenvalue gives the total variance explained by each PC. Any PC with an eigenvalue less than one does not present enough total variance to be considered a unique PC (OECD, 2008). Therefore, we have considered only PCs with eigenvalues greater than one. The correlation of each indicator with all extracted could be explained through an unrotated component (factor) matrix. This matrix demonstrates the component loading, which is simply the correlation for each indicator on every component. An unrotated component solution serves only the purpose of data reduction and fails to adequately interpret the variables (Hair et al., 2009). Hence, the components are rotated to enhance interpretability (OECD, 2008). Though no specific rule was developed to guide the selection of any particular rotation technique, the most widely used technique is the orthogonal varimax rotation method used in this study to maximize the factor loading variance. Furthermore, extracted PCs are aggregated to generate the overall WPI. Garriga & Foguet (2010) discussed that all aggregation functions suffer to a certain extent. Among all the methods of aggregation, additive, multiplicative, or PCA, the result remains the same, and the methods are highly correlated (Prabha et al., 2020). Moreover, Goel et al. (2020) argue that the solution to the water stress problem lies in identifying the scores of the component. Loss of crucial information is bound to happen during aggregation. Subsequently, the application of the min–max method of normalization gives the WPI values between 0 and 1.
Spatial analysis of water poverty index
Using ArcMap 10.3, we illustrate the state's spatial distribution of water poverty. This study employs an equal-interval approach to set up five classes of attributes, initiating the demonstration of variation in all extracted PC scores and mapping the overall water poverty.
Tobler's first law of geography states that ‘everything is related to everything else, but near things are more connected than distant things’ (Tobler, 1970). We have analysed the WPI distribution characteristics in Maharashtra. There are two types of spatial effects: (a) spatial dependence/autocorrelation variables in one location depend on (or are correlated with) values of neighbouring areas and (b) spatial heterogeneity shows the variation in relationships across space. Exploratory spatial data analysis is used to visualize, summarize the spatial properties of a variable, and describe specific patterns in space (Oliveau & Guilmoto, 2005). Spatial autocorrelation is the most common technique to examine the nature of spatial variation, which shows the correlation of variables in close locational positions. Furthermore, we try to show hotspots or clusters using spatial autocorrelation. The first and most widely used measure of spatial autocorrelation is Moran's Index (or Moran's I). There is a two-step procedure to visualize spatial patterns using Moran's I: (i) global Moran's I, examining the extent to which similarity or dissimilarity pattern of an attribute exists in all neighbouring areas within the defined space or mapped region and (ii) local Moran's I to investigate the clusters or hotspots. Local indicator of spatial analysis (LISA) is a standard tool to check local autocorrelation once a global Moran's I pattern is observed (Anselin, 1995). The range of Moran's I lie between +1, denoting positive spatial autocorrelation, and −1, denoting negative spatial autocorrelation. Zero means no pattern. Positive spatial autocorrelation shows that geographically closer values of the variable tend to be similar on a map, i.e., high values tend to be located near high values and low values near low values (Griffith, 1987). Negative autocorrelation exhibits that geographically distant values of the variable have a strong dissimilarity pattern on a map, i.e., high values may be near high or low. LISA presents different types of spatial clusters, namely high-high (HH), also known as a hotspot, where high values surround a location with high values; low-low (LL), also known as cold spots, where low values surround a location with low values; low-high (LH) and high-low (HL) are potential outliers, where location with low values is surrounded by high values and location with high values are surrounded by low values, respectively. Moran's I scatterplot is an exploratory tool that plots cluster values in four quadrants and reflects the type and strength of spatial autocorrelations. The p-value of Moran's I is an approximate indicator of statistical significance. This work has used GeoDa version 1.20 to conduct spatial autocorrelation and exhibit spatial distribution. After calculating global and local Moran's I to check similarities and dissimilarities among spatial locations, we plot a map to display the locations of hotspots and potential outliers.
RESULTS
Using the framework suggested in the previous section, we estimate the WPI for 33 districts. The final WPI scores are generated by aggregating the values of extracted PCs. We explain the performance of different indicators across the districts under each PC, followed by the overall water poverty situation in the state. Besides, districts are divided into categories according to their WPI score. Furthermore, spatial clusters of the WPI are analysed through spatial autocorrelation techniques.
After performing PCA, our indicators converge into five PCs, namely PC1, PC2, PC3, PC4, and PC5. In Table 4, the second column demonstrates each PC with an eigenvalue greater than one, describing sufficient variation and distinctiveness. Overall, the five PCs explain 80% of the variation in the data.
Result of PCA; eigenvalue analysis of various components.
Components . | Initial eigenvalues . | ||
---|---|---|---|
Total . | % of variance . | Cumulative % . | |
PC1 | 4.528 | 28.298 | 28.298 |
PC2 | 3.480 | 21.753 | 50.051 |
PC3 | 2.081 | 13.009 | 63.059 |
PC4 | 1.313 | 8.206 | 71.266 |
PC5 | 1.289 | 8.058 | 79.324 |
Components . | Initial eigenvalues . | ||
---|---|---|---|
Total . | % of variance . | Cumulative % . | |
PC1 | 4.528 | 28.298 | 28.298 |
PC2 | 3.480 | 21.753 | 50.051 |
PC3 | 2.081 | 13.009 | 63.059 |
PC4 | 1.313 | 8.206 | 71.266 |
PC5 | 1.289 | 8.058 | 79.324 |
District-wise WPI scores in Maharashtra.
Overall WPI (rank) . | Very low (0–0.2) . | Low (0.2–0.4) . | Medium (0.4–0.6) . | High (0.6–0.8) . | Very high (0.8–1) . |
---|---|---|---|---|---|
Thane (30), Dhule (31), Nandurbar (32), Jalna (33) | Nanded (23), Ratnagiri (24), Raigad (25), Latur (26), Hingoli (27), Parbhani (28), Osmanabad (29) | Wardha (7), Amravati (8), Yavatmal (9), Satara (10), Kolhapur (11), Sindhudurg (12), Ahmednagar (13), Solapur (14), Jalgaon (15), Buldhana (16), Aurangabad (17), Washim (18), Akola (19), Sangli (20), Nashik (21), Bhandara (22) | Gondia (4), Pune (5), Beed (6) | Chandrapur (1), Nagpur (2), Gadchiroli (3) |
Overall WPI (rank) . | Very low (0–0.2) . | Low (0.2–0.4) . | Medium (0.4–0.6) . | High (0.6–0.8) . | Very high (0.8–1) . |
---|---|---|---|---|---|
Thane (30), Dhule (31), Nandurbar (32), Jalna (33) | Nanded (23), Ratnagiri (24), Raigad (25), Latur (26), Hingoli (27), Parbhani (28), Osmanabad (29) | Wardha (7), Amravati (8), Yavatmal (9), Satara (10), Kolhapur (11), Sindhudurg (12), Ahmednagar (13), Solapur (14), Jalgaon (15), Buldhana (16), Aurangabad (17), Washim (18), Akola (19), Sangli (20), Nashik (21), Bhandara (22) | Gondia (4), Pune (5), Beed (6) | Chandrapur (1), Nagpur (2), Gadchiroli (3) |
DISCUSSION
Our results highlight a great disparity relating to water poverty across the districts. The overall WPI score for Maharashtra is 0.47, signifying the state has performed poorly across all five components. Out of 33 districts selected for our analysis, 21 fall into medium to very high categories. This reveals the severity of water poverty situation in the state. Eleven districts have very low and low scores, indicating better performance in water poverty across all five components of WPI. There is an inter-district heterogeneity in terms of the five components of WPI across Maharashtra. We provide a detailed description of our findings considering the regional heterogeneity in this section.
- (a)
Konkan division
The districts in the Konkan division, namely Thane, Ratnagiri, and Raigad, have performed well in the overall WPI except for Sindhudurg, which has medium water poverty. This division belongs to the north and south Konkan coastal plain agroclimatic zone, which is the wealthiest part of the state; along with rich natural resources and abundant rainfall, the region has a share of around 40% in GSDP. These districts have performed well in overall WPI ranks. With high forest cover (as % of total geographical area), literacy rate, HH with improved water services, and low consumption of fertilizers per hectare of land with low variation in the groundwater level, surface water, and rainfall over the years, the region has maintained environment, resource, and capacity components. However, there is a need to improve sanitation facilities and better access to irrigation.
- (b)
Pune division
Pune division has medium to high overall WPI scores across the districts with coverage from the Western Ghats Mountain zone to Western Plain transition zone II, passing through sub-mountain transition zone I. In this region, rainfall varies from very high to low. Also, in the Krishna River, along with most of the west-flowing rivers passing through the area, which are primarily rain-fed, the surface water variation is very high, leading to inconsistent water availability for irrigation and other purposes. The region has a high per capita GDP, but the access to electricity and health facilities is below the state average.
- (c)
Nashik division
Nashik division primarily has tribal-dominated districts with low to medium WPI scores. Major rivers of the state, namely, Godavari and Tapi, flow through this part of the state; however, the physical availability of water is similar to the Pune division, with high variation in rainfall and surface water. The tribal districts with low literacy have abysmal access to sanitation and electricity. However, the overall region appears water secure, but there is scope for progress relating to specific indicators.
- (d)
Aurangabad, Amravati, and Nagpur divisions
Districts of the Aurangabad division (Marathwada region) and the Amaravati and Nagpur divisions of the Vidarbha region primarily fall under the medium to high category of WPI scores. These are the drought-prone and impoverished regions of the state. Seasonal groundwater deviation is very high. The larger part of the population is engaged in agriculture, producing water-guzzling cash crops such as cotton and sugarcane. Additionally, the rainfall variance in these regions is excessive, and the dependency on rain-fed rivers to fulfil their water needs escalates the severity of the water situation. Yavatmal, Buldhana, and Akola are parts of the Vidarbha region located in the eastern part of the state, with India's most backward districts. Districts of the Marathwada and Vidarbha regions are known for the simultaneous existence of poverty and drought. These regions stare at water emergencies every summer and depend on water tankers to fulfil their water needs. The overall water poverty scenario is the worst in the eastern Vidarbha part of the state, particularly in districts namely Chandrapur, Nagpur, Gadchiroli, and Gondiya.
Pune and Nagpur are some of the state's prosperous and highly urbanized districts. Yet, a considerable population suffers from water poverty in these urban agglomerations. Nagpur presents unsatisfactory outcomes across all five components of WPI, capturing the region's distressing water problems. The district has a much higher infant mortality rate than the state average and lower access to improved water and sanitation. However, Jalna and Chandrapur appeared as outliers with extreme scores, which may be due to data constraints. Also, the decent WPI scores by districts in different parts of the state indicate that no one factor, such as resource availability, income level, or access to resources, can determine water prosperity.
Water resource availability should not be treated as the only factor in reducing water stress, but appropriate water management is necessary. Despite abundant water resources, a region can be less secure if it lacks water management (Yadav & Ibrar, 2022). The significant findings of our study are in contrast to the results by Goel et al. (2020), where Maharashtra appeared as a better performer, displaying low to medium WPI; the present work shows that the downscaled picture at the district level is significantly different. Maharashtra is facing severe water poverty. Pandey et al. (2022) presented comparable results using national sample survey data, suggesting that Maharashtra falls under the unsafe category regarding the WPI.
Decision makers should focus on the distinct intrastate impact due to an extensive economic, social, and climatic spatial variation across the state. There is an overdue task to investigate new ways of water availability, which must incorporate the massive spatial and temporal variations in the distribution of water resources in India, considering its access as well (Cronin et al., 2014). Policy suggestions from the current work recommend that along with the physical abundance and accessibility to water, improvement in the determinants of capacity and environment are essential to tackle water poverty. The results coincide with the arguments by Shalamzari & Zhang (2018) and Goel et al. (2020), stating that while attempting to improve water poverty, policymakers must pay attention to these components. Over the years, various policies and programmes have been drafted for integrated water resource management in Maharashtra. In conformity with the National Water Policy, Maharashtra has its own State Water Policy formulated in 2003, later revised in 2012 and 2019. The goal of this policy is to transform overall water governance by engaging different stakeholders in planning for improved drinking water supply and sanitation facilities, as well as strategic investments (Water Resources Department, 2019). Our findings also emphasize that the WPI is highly influenced by access and capacity indicators across the state. However, the sectoral water allocation in the policy neglects environmental water needs over other uses. The government of Maharashtra has also introduced a farm pond scheme and Marathwada water grid project in the drought-prone regions to support farmers and recharge groundwater. Our outcomes depict the water poverty scenario in these water-impoverished districts of the state and highlight the water requirement by the agricultural sector and the erratic water situation in these districts. Although the implementation of these schemes has remained questionable regarding their sustainability, little success has been yielded. Policy formulations need to promote multidimensional aspects to make any district water secure.
SUMMARY AND CONCLUSION
The present study uses the concept of the WPI to understand the water security situation in Maharashtra, India. The looming water stress in the region indicates substantial use of this resource, affecting its sustainability with rising possibilities of conflicts. A holistic measure of the water stress situation can be addressed through water poverty. The WPI is a systematic and integrated indicator to emphasize water stress. Liu & Liu (2021) argue that water poverty is associated with economic deprivation. Effective water management requires a linkage between water availability, accessibility, and the capacity to manage safe water from improved sources to meet water demand. Using PCA, this study chooses a set of 16 indicators under five components to evaluate the WPI. Based on the 33 districts' analysis, Maharashtra's estimated overall WPI implies high water stress. Our findings highlight poor performance in water poverty in the districts of Marathwada and Vidarbha regions of the state.
The WPI is mainly influenced by access and capacity components, as the extracted factors are primarily dominated by improved sanitation, followed by average literacy rate and real GDP per capita. Additionally, the indicators under use and resource components also affect the water poverty situation in the districts. Since around 50% of the population is dependent on agriculture, improvement in water availability for agriculture throughout the year and its fair distribution might help reduce water poverty. Simultaneously, developing access facilities and capacity building by improving literacy, health provisions, and income can significantly improve the situation. Apart from this, the WPI is also found to be sensitive to environmental components. Therefore, the ecological needs of water must be addressed while targeting water poverty. Additionally, the geographic information system tool is used to understand the spatial distribution of water poverty across districts. Water resource management using the spatial distribution of indicators measuring water poverty is crucial to reflect the accurate picture. Spatial mapping identifies the water poverty hotspots and may assist policymakers in making informed decisions while dealing with the factors responsible for poor performance and highlighting water-poor locations.
Nonetheless, one of the significant constraints of this study is the inability to create this composite index for any particular year. The unavailability of data for all five components every year produces such an outcome. Data available in the public domain for a few components are slightly old. However, using the available data would help assess the disparity since forecasting or extrapolating variables based on previous data may produce an ambiguous result. Some crucial variables are dropped due to the inability to find data for all the units of analysis. Besides, an impediment to the final WPI index is that it is effective depending on its ability to minimize the loss of information retained in each indicator. This study uses secondary data to assess water poverty across the districts of the state facing crucial water management issues. The evidence supports integrated water resource management. Thus, the index may not be exhaustive due to the abovementioned constraints, but the results may help serve the decision-making purpose. Incorporating soil erosion and desertification data may diversify the environmental component by presenting detailed information on ecological water needs. The subsequent research work can capture a thorough picture of further downscaled study units, including more indicators under each component of the WPI.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.