The social ecology of water in a Mumbai slum: failures in water quality, quantity, and reliability

KB is a non-notified slum of approximately 10,000 people located on a wharf on Mumbai’s eastern waterfront. Like other slums, living spaces consist of shanties built of corrugated metal, wood, and cement. Only 3% of homes have a latrine inside. As a result, the vast majority of children and 14% of adults engage in open defecation, while the remainder of adults use a handful of pay-for-use toilets in the community or travel long-distances to access pay-for-use toilets outside of the slum [11]. KB is located on central (federal) government property, which precludes it from receiving services provided by the city government, such as water and electricity [10]. Nearly all KB residents therefore purchase water via an informal distribution system run by private vendors.

The sources of KB’s water supply are two underground pipes that were installed decades ago by the fire department for emergency use. Water vendors, nearly all of whom are residents of KB, have created entry points in the fire brigade pipes from which water is extracted using motorized pumps. Water is then pumped through rubber hoses that extend hundreds of meters to reach community lanes. Given the absence of space on the main road of the community and the lack of underground piping infrastructure, the hoses travel through the surrounding ocean and trash dumps and are often visibly compromised by holes. Tracing each vendor’s water distribution area (i.e., the households receiving water from a particular vendor’s motor and hose system) formed the basis for sample selection in the SWA, as is described further below.

Within each household, water is stored in two types of containers. Large plastic drums with capacities of 100 to 300 liters are placed outside the home and hold “storage water,” which is used for bathing, toileting, and washing clothes (Figure 1). Water from these larger storage containers is used for drinking only during times of severe water scarcity. Smaller containers with a capacity of one to 50 liters are kept inside the home and are used to store “drinking water” (Figure 2). Notably, nearly all of the drinking water containers in KB are wide-mouthed and allow people to directly access water from the containers with their hands, a detail that has major implications for household-level water contamination.

KB’s informal water distribution system is vulnerable to widespread failure. Local officials (from the central government agency on whose land KB is situated) occasionally raid and confiscate motors tapping into the fire brigade pipes, cutting off water access to KB’s residents. Such episodes of “system failure” occur a few times a year. When this happens, most KB residents roll large storage drums at least one kilometer, and as far as two kilometers, to access taps in the next closest community, while others get water from private tankers.

The water vendors who run the informal water distribution system incur significant costs to maintain the system, all of which are passed on to community residents who purchase this water. For example, based on interviews with multiple water sellers, we estimated that they pay Indian rupees (INR) 15,000 to 20,000 for a new motorized pump (used for extracting water from underground pipes), which is US dollars (USD) 273 to 364. A used motorized pump costs INR 10,000 to 12,000 (USD 182 to 218); however, these require frequent repair every few months, which generally costs INR 500 (USD 9) per repair. When local officials confiscate motorized pumps every few months, water vendors pay bribes of INR 500 to 1000 (USD 9 to 18) to get motorized pumps back, though they are often unable to get the pumps back and are forced to buy new ones. The rubber hoses used to distribute water to community lanes cost INR 5000 (USD 91) per 100 meters, and a few hundred meters of hose are often required to provide water to community lanes from each motorized pump. Finally, each water vendor usually hires one or two other people to facilitate water distribution, and these individuals are remunerated by receiving free water.

The BNA is a 92-question household survey covering five domains: income and assets, access to government benefits, access to education, access to water and sanitation, and common causes of morbidity. Only the water-related data are reported here. The survey was implemented from August to December 2008 by a group of 15 uniformly trained interviewers. The questionnaire was administered to a head of household greater than 18 years of age after receiving written informed consent. Since men in KB work long hours outside of the home, 71% of respondents were women.

Researchers coded every living space in KB. Every occupied household was requested to answer the questionnaire at least once during the study period. Due to migrancy, many households were empty and locked at any given time. Out of 2,439 total living spaces, 900 were permanently locked with no identifiable occupants. Of the remaining 1539 occupied households, 959 responded to the survey, yielding an estimated 37.7% non-response rate. This relatively high non-response rate is attributable to the long work days of many adults, who return home at hours when the community is not safely accessible to researchers.

The SWA evaluated microbiological and chemical quality of water, quantity of water used, and cost of water during four study periods: winter (February 2011), summer (early May 2011), monsoon (August 2011), and an episode of “system failure” in late May 2011 (see above), when KB residents had to seek out alternative water sources. This seasonal design originated from findings of prior studies [12, 13] and from observations by researchers of possible variations in household water availability and cross-contamination risk during Mumbai’s different seasons.

Researchers uniformly trained in sterile technique collected water samples in autoclaved bottles from many points along the informal distribution network. These points included: three motorized pumps that extract water from fire brigade pipes (i.e., the “point-of-source”), six hoses that transport water from the three pumps to lanes in KB (i.e., the “distribution network”), and three randomly selected households supplied with water by each of the six hoses (i.e., the “point-of-use”). An additional point-of-source sample was collected from a household with a tap that directly accesses a fire brigade pipe. Since some households buy water directly from this tap (i.e., not through hoses), three such households were also selected for water collection, bringing the total household sample size to 21 (Figure 3). The same 21 households were followed for all four periods of data collection.

For comparison to KB point-of-source samples, control point-of-source tap water samples were obtained from chawls (low-income housing) and from Dharavi, a notified slum with legal water access (Figure 3). During “system failure,” a point-of-source sample was collected from the Reay Road tap, the main tap outside of KB used by residents when the water distribution system fails (Figure 4). Given environmental exposure to seawater in KB, two ocean water samples were also collected for testing in each season.

All 229 water samples were transported to Equinox Labs, which is certified by the International Organization for Standardization (ISO 9001:2008) and the National Accreditation Board for Laboratories (NABL), within four hours of collection. Equinox Labs performs external quality testing by cross-checking multiple samples with five other NABL certified labs at least once a year, and internal quality checks for water testing are carried out on a daily basis. Samples were tested for total coliform counts, E. coli, and chemical parameters, as per protocols recommended by the American Public Health Association and the Indian Standards Chemical Division Council [14].

The chemical parameters tested included pH (normal 6.5-8.5), total dissolved solids (normal <500 mg/L), turbidity (normal <5 NTUs), total hardness (normal <300 mg/L), calcium (normal <75 mg/L), magnesium (normal <30 mg/L), and sulphates (normal <200 mg/L). Given the fact that water from the motorized pumps and distribution hoses in KB is at risk for contamination with ocean water, we also tested these samples in the winter season for salinity (normal <0.8 dS/m). Microbiological testing was performed as follows: 100 mL of each sample were filtered through a 0.22-micron, 47 mm diameter membrane using a vacuum pump. The filter was subsequently transferred to an m-Endo agar plate for culturing. Samples were incubated at 35°C for 24 hours, at which time coliform bacterial colonies were counted using an automated colony counter. E. coli was reported as being present or absent after interpretation of indole, methyl red, Voges-Proskauer, and citrate tests.

In addition to water sample testing, in all four study periods, researchers administered questionnaires to an adult older than age 18 in each of the 21 households. Since questionnaires were administered around the time of water collection, they were usually administered to the adults who most commonly interact with water vendors and who therefore were most likely to provide accurate information on water spending. The questionnaire assessed common health and social equity indicators, specifically monthly and weekly water charges and quantity of water used in the last week [2]. For each household water sample collected, it also assessed the following parameters possibly associated with contamination, which were selected based on a review of prior studies [15‐17]: gross appearance of the water, composition of the water container, and length of time since the container was last cleaned and filled with water.

Quantity of water used was assessed through a detailed inventory of every drinking and storage water container in each household. Prior to starting data collection in the field, researchers were uniformly trained to recognize the volume of water held by containers of various sizes and shapes commonly used to store water in the community. In each household surveyed, researchers carefully estimated the capacity in liters of each container and the number of times it had been filled in the last week. These numbers were tabulated to estimate the total amount water used in the last week by each household.

Multivariate logistic regressions were performed to understand associations between microbiological water quality (i.e., the presence or absence of coliforms and E. coli) and other drinking water and container characteristics (Table 2). Multivariate linear regressions were performed to understand associations between quantity of water consumed and season and cost variables (Table 2). Statistically significant variables (p < 0.05) were included in the multivariate models using the forward stepwise method, and the best reduced model was built based on a −2 log likelihood value. Statistical analyses were performed using SPSS (version 19; Chicago, IL, USA).

Research protocols were approved by the Partners for Urban Knowledge, Action, and Research (PUKAR) Institutional Ethics Committee, which is registered under federal-wide assurance number FWA00016911.

Table 3 presents SWA data on household water costs during the four study periods. A majority of households paid a monthly base fee to water vendors of Indian rupees (INR) 150 to 400 per month, which is approximately US dollars (USD) 2.73 to 7.27, for water during all study periods. In summer and during episodes of system failure, most households pay additional weekly fees to acquire extra water. Based on monthly and weekly water costs, as well as water quantity data, we were able to estimate the cost spent per 1000 liters of water for each household. By comparing these costs to the standard government water charge in 2011 of INR 2.25 (USD 0.04) per 1000 liters, we estimate that KB residents spend 52 to 206 times more than residents of slums with legal water access, depending on the season (Table 3).

Based on income data for 521 households from another study (PUKAR's 2012 mental health study in KB), we estimate the mean monthly household income in KB to be INR 6411 (USD 116.56) [18]. Using this figure, we estimate that KB residents spend 5.9% to 15.9% of their monthly household income on buying water in different seasons (Table 3). We then estimated the mean total cost of water spent by a household over an entire year to be INR 6479 (USD 117.80), by assuming that winter season lasts from October to February, summer season from March to May, and monsoon season from June to September. Based on this figure, KB households spend approximately 8.4% of their yearly income on water.

A 2010 PUKAR census of KB found a population of approximately 10,000 people and an average household size of 4.8 people. Using these figures, we calculated the approximate yearly amount spent on water by the entire community to be INR 13,498,083 (USD 245,420.00). We compared this cost to an ideal scenario in which all 10,000 residents receive the WHO-recommended minimum of 50 liters per capita per day (l/c/d) of water for every day of the year at the standard government charge of INR 2.25 (USD 0.04) per 1000 liters. In such an ideal scenario, the entire community would spend INR 410,625 (USD 7,466.00) on water yearly.

This suggests that residents spend an excess amount of INR 13,087,458 (USD 237,954.00) yearly on water in the current informal system. We compared this excess amount to the cost of placing comprehensive water infrastructure in KB (in the form of a new pipeline with community taps), which is INR 2,500,000, or approximately USD 45,455.00 (personal communication from Ward Corporator Mangesh Bansod). Based on this figure, the excess amount spent on water under the informal system could pay for entirely new water infrastructure in KB more than five times every year.

The BNA provides additional data on water-related hardships. In the BNA, 952 households (99.3%) report having to regularly purchase water. The majority, 529 (55.2%), are only able to access water every three or more days (Table 4). Most households, 817 (85.2%), have water delivered via water vendors’ hoses, while 125 (13.1%) must fetch water from outside of their lanes. Due to queues at hoses or time involved in fetching water, 370 (38.5%) spend > ½ hour on obtaining water. Many households report a subjective perception that the local water situation affects their family members’ health, ability to go to work, and ability to go to school (Table 4).

Data from the SWA show that, in different seasons, 81-95% of households do not meet the WHO recommendation that all human beings use a minimum of 50 liters per capita per day (l/c/d) of water (Table 3 and Figure 5) [19]. A significant percentage use less than 20 l/c/d, a consumption level associated with a “high” or “very high” level of health concern per the WHO [19]. In the multivariate linear regression model (R² = 0.329), total money spent on purchasing water was positively associated with the quantity of water consumed (β-coefficient = 0.569, p < 0.001), while the cost of water in INR per 1000 liters was negatively associated with the quantity of water consumed (β-coefficient = −0.691, p < 0.001).

According to the BNA, most households, 568 (59.2%), do not use any method of water purification, while 25.8% use a cloth filter and 17.2% boil their drinking water prior to consumption (Table 4). Of note, our informal observations suggest that many adults also boil water prior to consumption in beverages such as tea or coffee; however, the majority of water is consumed “raw” without being boiled in beverages, especially for children, who are most likely to suffer from diarrheal illness. The SWA similarly found that 12 households (57.1%) did not use any method of water purification during any of the study periods. In the SWA, 35.7% of drinking water containers were made of plastic, 60.7% of metal, and 3.6% of clay. One hundred percent of storage water containers were made of plastic.

Table 5 shows the SWA’s microbiological testing results. The monsoon season had a different pattern of water contamination as compared to the pattern in the winter season, the summer season, or during the episode of “system failure.” We will first discuss the pattern of contamination in the winter and summer seasons.

During winter and summer, there was no coliform or E. coli contamination of any of the point-of-source samples from the two comparison groups, chawls (low-income housing) and Dharavi (a notified slum with legal water access). In winter and summer, none of the water samples collected from KB’s motorized pumps (which represent the point-of-source) or from KB’s hoses (which represent the distribution network) showed any evidence of contamination with coliforms or E. coli. Despite the absence of bacterial contamination of water at the point-of-source or in the distribution network of hoses in KB, there was significant contamination of drinking water and storage water at the household level (i.e., the point-of-use). For example, in the summer, 52.4% of drinking water samples were contaminated with coliform bacteria, and 42.9% were contaminated with E. coli. This suggests that all bacterial contamination of water during the winter and summer was happening at the household-level (i.e., the point-of-use) and not at point-of-source or in the distribution hoses.

During the episode of “system failure,” the pattern of water contamination was similar to the pattern in the winter and summer seasons, in that all contamination of water happened at the household level. The Reay Road tap, which represents the point-of-source for water during the episode of “system failure,” was not contaminated with coliform bacteria or E. coli. Again, despite receiving uncontaminated point-of-source water, 23.8% of drinking water samples were contaminated with both coliform bacteria and E. coli.

In contrast, during the monsoon season, we found multiple point-of-source samples to be contaminated. In the comparison groups, there was no contamination of any point-of-source samples from Dharavi, but one sample (25.0%) from the chawls was contaminated with both E. coli and coliforms. Two samples from KB’s motorized pumps (50.0%) were contaminated with coliforms but not E. coli, highlighting significant point-of-source contamination of water in KB in the monsoon season. During the monsoon, three hoses (50.0%) were contaminated with coliforms and one (16.7%) was contaminated with E. coli. Of note, all three of these contaminated hoses were connected to the motors in KB that showed evidence of contamination with coliform bacteria, suggesting that the water in these hoses became contaminated at the point-of-source and not secondarily as it ran through the distribution hoses. During the monsoon, at the household level, there was a very high rate of contamination of drinking water and storage water, with 76.2% of drinking water samples being contaminated with coliform bacteria. Of note, 75.0% of these contaminated drinking water samples were from households receiving water from uncontaminated distribution hoses. This suggests that, even in the monsoon, when there was significant point-of-source contamination of water, there was also superimposed household-level contaminated of stored drinking water at the point-of-use.

In the multivariate logistic regression analysis based on the variables noted in Table 2, season was the only significant predictor of coliform and E. coli contamination of drinking water (Table 6). Summer was significantly associated with both coliform and E. coli contamination, while the monsoon was significantly associated with coliform contamination only.

All of the water samples were also tested for common chemical parameters such as pH, total dissolved solids, turbidity, total hardness, calcium, magnesium, and sulphates. All samples were found to be within internationally accepted limits for all chemical parameters (normal ranges are noted above in the methods section). During winter season, we also specifically tested the point-of-source water samples (e.g., KB motors, Dharavi taps, and chawl taps) as well as the KB hose samples for salinity levels, given the risk for ocean water contamination of KB motor and hose samples. All of these samples had salinity levels of approximately 0.017 to 0.021 dS/m, which is well within the normal limit for human consumption of 0.8 dS/m.

Table 4 presents BNA data regarding residents’ opinions on the causes of, and responsibility for, water access problems. Most identify complications with land ownership and the slum’s unauthorized (i.e., non-notified) status as the main reasons for lack of formal water access. Three-fourths feel that the local government (either local politicians or the municipal system) should be responsible for improving water access. Most, 528 (55.1%), are willing to pay fees to the government for a reliable water supply. If such a supply were established, 874 (91.1%) expect that it would provide water every day, and 297 (31.0%) expect that water would be provided through home taps.