Article pubs.acs.org/est
Assessing Drinking Water Quality and Water Safety Management in Sub-Saharan Africa Using Regulated Monitoring Data Emily Kumpel,*,† Rachel Peletz,† Mateyo Bonham,† and Ranjiv Khush‡ †
The Aquaya Institute, PO Box 21862, Nairobi, Kenya The Aquaya Institute, 12 E Sir Francis Drake Blvd, Suite E, Larkspur, California 94939 United States
‡
S Supporting Information *
ABSTRACT: Universal access to safe drinking water is prioritized in the post-2015 Sustainable Development Goals. Collecting reliable and actionable water quality information in low-resource settings, however, is challenging, and little is known about the correspondence between water quality data collected by local monitoring agencies and global frameworks for water safety. Using 42 926 microbial water quality test results from 32 surveillance agencies and water suppliers in seven subSaharan African countries, we determined the degree to which water sources were monitored, how water quality varied by source type, and institutional responses to results. Sixty-four percent of the water samples were collected from piped supplies, although the majority of Africans rely on nonpiped sources. Piped supplies had the lowest levels of fecal indicator bacteria (FIB) compared to any other source type: only 4% of samples of water piped to plots and 2% of samples from water piped to public taps/standpipes were positive for FIB (n = 14 948 and n = 12 278, respectively). Among other types of improved sources, samples from harvested rainwater and boreholes were less often positive for FIB (22%, n = 167 and 31%, n = 3329, respectively) than protected springs or protected dug wells (39%, n = 472 and 65%, n = 505). When data from different settings were aggregated, the FIB levels in different source types broadly reflected the sourcetype water safety framework used by the Joint Monitoring Programme. However, the insufficient testing of nonpiped sources relative to their use indicates important gaps in current assessments. Our results emphasize the importance of local data collection for water safety management and measurement of progress toward universal safe drinking water access.
■
INTRODUCTION Unsafe drinking water accounted for an estimated 500 000 deaths globally in 2012, almost half of which were in subSaharan Africa (SSA).1 The WHO/UNICEF Joint Monitoring Programme (JMP), however, estimates that the majority of people in SSA (68%) drink water from a source-type that is classified as “improved” by the JMP.2 To measure progress toward the drinking water targets established by the United Nations’ Millennium Development Goals (MDGs), the JMP currently uses source-type classifications as proxies for water quality: improved sources are considered more likely to provide a better quality of water than unimproved sources. Nevertheless, growing evidence indicates that improved sources do not necessarily provide water that is free of microbial and chemical contaminants.2−8 Therefore, many more Africans are likely drinking unsafe water than currently estimated based on these source-type proxies.4 This recognition has led the JMP to propose that drinking water quality be included as part of the indicator for the post-2015 Sustainable Development Goal (SDG) for safe drinking water.9,10 However, collecting reliable and actionable water quality information, particularly in lowresource settings, poses a significant challenge. Efforts are underway to include water quality testing in the nationally representative surveys that inform JMP estimates for water and sanitation access, such as the demographic and health © 2016 American Chemical Society
surveys (DHS) and the multiple indicator cluster surveys (MICS).11,12 These cross-sectional surveys, however, will not capture periodic fluctuations in water quality and have limited application for local water safety management. In recognition of these shortcomings, the JMP has also emphasized the need to strengthen water quality data collection by local institutions with regulatory responsibilities for monitoring drinking water services.9 Local monitoring includes operational monitoring by formal water suppliers to ensure process controls and verify supplied water quality, and surveillance monitoring of both formal and informal water sources by public health or water authorities to evaluate the quality and adequacy of supplies.13−15 National ministries and independent sector regulators are usually required to collect and analyze operational and surveillance data for both enforcement and resource allocation purposes. Our previous research shows that operational and surveillance monitoring are established practices in many SSA countries, although most of the institutions that we surveyed do not meet their regulatory requirements for monitoring Received: Revised: Accepted: Published: 10869
May 30, 2016 August 16, 2016 August 25, 2016 August 25, 2016 DOI: 10.1021/acs.est.6b02707 Environ. Sci. Technol. 2016, 50, 10869−10876
Article
Environmental Science & Technology
Table 1. Summary of Dataset by Country, Including the Number (n) of Supplier or Surveillance Agencies, The Methods Used for Testing in Each Country, The Years in Which Data Were Collected, The Number of Tests for Fecal Indicator Bacteria (FIB), and the Percent of FIB Tests from the Retrospective Dataset (As Compared to MfSW-Supported data) monitoring agencies (n)
a
water quality data set a
country
all
suppliers
surveillance
methods
Ethiopia Ghana Guinea Kenya Senegal Uganda Zambia Total
4 1 1 8 1 9 8 32
2 1 1 4 0 4 3 15
2 0 0 4 1 5 5 17
MF, MPN MPN MF MF, MPN, PA, PC MF MF, MPN PA, MF
number of tests for FIB years
FIB (n)
percent retrospective
2011−2015 2011−2012 2012−2015 2009−2015 2009−2015 2009−2015 2010−2015
8017 506 2954 10 623 4423 9427 6976 42 926
97% 100% 40% 44% 3% 36% 17% 44%
Methods: MF: Membrane Filtration; MPN: Most Probable Number; PA: Presence-Absence; PC: Plate count.
frequency.16 Little is known, however, about whether the water quality data collected by local monitoring agencies agree with our current understanding of water quality from other sources of information such as previous research studies, systematic reviews, and nationally representative rapid assessments.8,17 In addition, although water quality monitoring is intended to lead to actions that improve water quality, official responses by formal water suppliers and surveillance agencies to water quality information have not been documented or evaluated. Efforts to improve water quality monitoring by local institutions should include strategies for optimizing the efficiency of data collection, which require an understanding of how water quality varies by source type in specific settings and how officials respond to evidence of poor water quality. Our research comprised three main objectives: (1) to determine the degree to which water sources and locations are monitored by water suppliers and public health agencies in SSA; (2) to use microbial water quality data collected by local monitoring agencies in SSA to understand how water quality varies by source type in different settings; and (3) to collect information on responses to water quality testing results. Our goal is to apply this information to identify priorities for regulated water quality monitoring programs.
2012 and July 2015 (detailed in the Methods section of Kumpel, et al.19). We observed each MfSW-supported agency’s sampling and testing processes during midterm assessments and scored their quality control procedures (SI Table S1). We tested retrospective data against Benford’s law to examine whether data were potentially fraudulent (SI Table S2).16 Data that did not meet these minimum quality control standards or Benford’s law test were excluded from the analysis set. We also excluded the results of tests on chlorinated samples if sodium thiosulfate had not been used during sampling.20 A description of the exclusion process is presented in SI Figure S1. Regional Classifications. To analyze water quality in regions within countries, we grouped monitoring agencies’ areas of jurisdiction according to the administrative designations specific to each country. These areas included provinces in Zambia (Eastern, Central, Lusaka, Western, and Southern) and Kenya (Nyanza, Western, Central, and Nairobi), and regions in Ethiopia (Oromia and Addis Ababa), Senegal (Diourbel, Dakar, and Kaolack), Uganda (Northern, Central, Eastern, and Western), and Guinea (Conakry, Kindia, Kankan, Labé, and Nzérékoré).21−26 In the Results section, we have made the regions anonymous by designating them as A−E within each country. Water Source-Type Classifications. We used the sample collection information recorded with water quality test results to classify each sample according to JMP source-type categories.27 Improved sources include high access (on-plot piped supplies) and basic access (shared piped water from standpipes and kiosks, protected springs, protected wells, boreholes, and harvested rainwater). Unimproved sources are unprotected springs, unprotected wells, and surface water. We relied on monitoring agency definitions of protected springs or dug wells. Water samples from household storage containers or water tanks were classified as stored water (this classification did not include samples from reservoirs supplying distribution systems), although no details were available regarding the types of containers or the source of the water. When agencies identified a proportion of their piped water samples as off-plot (e.g., standpipes or kiosks), we assumed that their other piped water samples were taken from on-plot locations. We assigned sample locations to the following categories based on information provided with the test results: business or NGO facilities, work or refugee camps, educational institutions, food service facilities (e.g., restaurant, manufacturing), government facilities, hospital or health centers, hotels, markets, or public transport stops, police, military, or prison facilities, religious
■
MATERIALS AND METHODS Water Quality Data. Water quality data were collected over two stages of The Aquaya Institute’s (Aquaya) Monitoring for Safe Water (MfSW) research program, which studies water quality monitoring by formal water suppliers and public health agencies in SSA.16,18 In the first stage, retrospective water quality results from samples tested between January 2009 and December 2013 were collected, when available, from the 72 monitoring agencies that applied for participation in MfSW (application process and data described in Peletz, et al.16). In the second stage, the results of water quality tests conducted between July 2013 and July 2015 were emailed to Aquaya monthly by the 26 monitoring agencies participating in MfSW (selected institutions described in Peletz, et al.18). The MfSW program included financial incentives to promote increased water quality monitoring by collaborating water suppliers and public health agencies.18 The program did not, however, dictate sampling priorities; each agency determined sampling locations according to applicable standards or guidelines. We collected additional data regarding the monitoring agencies’ testing programs and procedures, including sampling, testing, and quality control methods, through surveys, interviews, and documented communications between November 10870
DOI: 10.1021/acs.est.6b02707 Environ. Sci. Technol. 2016, 50, 10869−10876
Article
Environmental Science & Technology institutions, community sources, or households. Samples lacking clear locational information were classified as unknown. We excluded water quality data from our analysis if the samples were taken from water treatment plants, reservoirs, or unknown sources (SI Figure S1). Ten percent of the 10 275 samples collected from groundwater sources (dug wells, boreholes) and 21% of the 2271 samples collected from springs could not be classified as protected or unprotected and were also excluded from this analysis. Similarly, the 10% of 30,413 piped water samples that could not be identified as onor off-plot were excluded (this included all data from Benin). Samples taken from vended (by cart) water, and tanker truck water were excluded due to sample sizes below 50 (24 and 20 samples, respectively) (SI Figure S1). Microbiological Analysis Classifications. The data that we received included assays for heterotrophic bacteria, hydrogen sulfide (H2S) producing bacteria, total coliforms, fecal coliforms, Enterococcus species, and Escherichia coli (E. coli) (Table 1). We selected assay results for the following fecal indicator bacteria (FIB) for inclusion in this analysis: E. coli, fecal coliforms, Enterococci, and H2S producing bacteria.28 We also included total coliform results if the monitoring agency performed a follow-up test for another FIB when total coliforms were detected. Testing methods included presence/ absence (H2S or E. coli detection in 100 or 10 mL sample volumes), membrane filtration-derived colony counts (E. coli, fecal coliforms, or total coliforms in 100 mL sample volumes), multiple test tube-derived most probable numbers (fecal and total coliforms in 100 mL sample volumes), and direct plate counts (E. coli in 1 mL sample volumes). Microbial Contamination Categories. Given the variety of testing methods and corresponding detection limits used, we analyzed the microbial water quality data using categories. We classified quantitative FIB data measured in 100 mL samples into four log-based categories: 100 CFU or MPN FIB/100 mL (presence/absence data were excluded from analyses based on this classification). We categorized positive presence/absence assays into contamination levels according to the tested sample volume: positive 100 mL tests were classified as ≥1 FIB/100 mL, positive 10 mL tests were classified as ≥10 FIB/100 mL, and positive 1 mL tests were classified as ≥100 FIB/100 mL. The numbers of samples classified into each category based on methods and corresponding sample volumes are described in SI Table S3. We used the same categories for quantitative test results. We used the R statistics software package (R Core Team, 2015) to analyze and visualize data.
Figure 1. Surveillance areas and urban water suppliers covered by the regulated monitoring agencies represented in the data set. The country shapefiles were obtained from DIVA-GIS (diva-gis.org).
Sixty-four percent of the water samples in the data set were collected from piped supplies (both on- and off-plot) and the remaining samples were collected from nonpiped improved sources (17%) unimproved sources (10%), surface water ( 25% of samples from both improved and unimproved sources showed ≥10 FIB/100 mL, but in Uganda region (D), fewer samples (
