Professional Documents
Culture Documents
Science of The Total Environment
Science of The Total Environment
H I G H L I G H T S
a r t i c l e i n f o a b s t r a c t
Article history: The objective of this study was to quantify sample transportation times associated with mandated microbiolog-
Received 29 January 2014 ical monitoring of drinking-water in Colombia. World Health Organization Guidelines for Drinking-Water Quality
Received in revised form 26 March 2014 recommend that samples spend no more than 6 h between collection and analysis in a laboratory. Census data
Accepted 26 March 2014
were used to estimate the minimum number of operational and surveillance samples required from piped
Available online xxxx
water supplies under national regulations. Drive-times were then computed from each supply system to the
Editor: Damia Barcelo nearest accredited laboratory and translated into sample holding times based on likely daily monitoring patterns.
Of 62,502 surveillance samples required annually, 5694 (9.1%) were found to be more than 6 h from the nearest
Keywords: of 278 accredited laboratories. 612 samples (1.0%) were more than 24 hours' drive from the nearest accredited
Monitoring laboratory, the maximum sample holding time recommended by the World Health Organization. An estimated
Spatial analysis 30% of required rural samples would have to be stored for more than 6 h before reaching a laboratory. The
Drinking water analysis demonstrates the difficulty of undertaking microbiological monitoring in rural areas and small towns
Faecal indicator bacteria from a fixed laboratory network. Our GIS-based approach could be adapted to optimise monitoring strategies
Water supply
and support planning of testing and transportation infra-structure development. It could also be used to estimate
GIS
sample transport and holding times in other countries.
© 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.scitotenv.2014.03.127
0048-9697/© 2014 Elsevier B.V. All rights reserved.
644 J. Wright et al. / Science of the Total Environment 485–486 (2014) 643–652
for microbiological testing. The WHO (1997: p. 53) recommends “the de Salud Departamental (Secretary of Departmental Health), that report
time between sample collection and analysis should, in general, not results to the INS via the Sistema de Información de la Vigilancia de la
exceed 6 h, and 24 h is considered the absolute maximum.” It goes on Calidad del Agua para Consumo Humano (SIVICAP; Information System
to recommend that samples are transported in an insulated container for Drinking Water Quality Surveillance) (Instituto Nacional de Salud,
and under ice, so as to inhibit regrowth of indicator bacteria. Where 2012). The INS makes use of a network of both public and private labo-
ice is unavailable, the WHO recommends that sample transportation ratories across the country for surveillance monitoring, with at least one
time should not exceed 2 h. National advice on sample holding times INS laboratory per state. Surveillance monitoring results are then sent to
is similar. In Colombia, the Instituto Nacional de Salud (INS; National the INS Grupo de Salud Ambiental (Environmental Health Group) in the
Institute of Health) recommends that sample transport time should capital, Bogota. The frequency of sampling for operational monitoring is
not exceed 6 h (Instituto Nacional de Salud, 2011), whilst the US higher than that for surveillance monitoring and the total number of
Environmental Protection Agency recommends that the time between samples collected nationally has increased markedly since 2007 as a
sample collection and start of incubation does not exceed 30 h (US result of regulatory changes introduced in that year.
EPA, 2005). In response to the logistical difficulties of transporting
samples to distant laboratories, a number of microbiological testing 2.2. Data sources
technologies have emerged that are based around portable and on-site
‘field kits’. A recent review identified 44 such testing technologies To identify laboratories approved for microbiological testing of
(Bain et al., 2012a), though these field tests are not used for compliance samples for surveillance monitoring, we extracted the names and ad-
monitoring in most settings. dresses of laboratories listed under Articles 1 and 3 of Resolution 431
The scale of the logistical challenge in monitoring remote water of the Ministry of Health and Social Protection (Ministerio de Salud y
supplies has to date remained largely unquantified. This paper therefore Protección Social, 2012). Article 1 indicates laboratories that are ap-
aims to estimate drinking-water sample transportation and holding proved for microbiological and chemical testing, whilst Article 3 lists
times based on the Colombian network of fixed laboratories. In so laboratories that are approved only for microbiological testing.
doing, we aim to develop a GIS-based method that could be used to To map the spatial distribution of required water tests, we chose
estimate sample holding times in other countries. To quantify sample to use census data in preference to data on water service providers
transport logistics, we use three steps. First, we map laboratory facilities (obtainable from the SSPD web site) because census data provide
and the distribution of water supplies in Colombia. Second, we estimate information on the location of small community supply users, unlike
the number of microbiological samples required from these supplies the service provider data which only cover large piped supplies. The
under national regulations and standards. Third, we estimate the travel latest national census, conducted on 3rd October 2005, included a
times from water supplies to laboratory facilities so as to quantify the question on the type of household water source: “Where does the
scale of the logistical challenge associated with water quality monitor- household primarily collect water for drinking and preparing food?”
ing. Through this analysis we identify areas that are too remote for Data on household water source type were obtained for 1119 munici-
monitoring via the existing laboratory network. palities from an online portal (Departamento Administrativo Nacional
de Estadistica, 2013). Statistics for each municipality were further bro-
2. Methods ken down into urban and rural areas, with the urban population defined
as that of the cabecera or principal municipal town (average population:
2.1. Study setting 28,500) and the rural population comprising the remainder of the mu-
nicipality (average population: 9800). To further disaggregate these fig-
According to the latest WHO/UNICEF Joint Monitoring Programme ures, we also drew on gridded estimates of total ambient population
(JMP) estimates for 2011, 93% of Colombians make use of an improved from the LandScan database (Dobson et al., 2000). These estimates are
water source, with 88% having water piped into the household. However, derived by assigning population within each municipality to constitu-
in keeping with many other countries, access to piped water within the ent grid cells that have suitable characteristics for human habitation
home is lower in rural areas at 61%, compared with 97% in urban areas (e.g. proximity to roads and suitable topography).
(WHO and UNICEF, 2013). The analysis presented here focuses only on As a basis for estimating the travel times between water supplies
piped supplies. and laboratories, we downloaded map layers depicting the locations
Water quality monitoring in Colombia can be divided into operational of human settlements and the road and river networks from an online
monitoring, intended to inform operational planning and management portal (UN-OCHA, 2013).
of water quality, and surveillance monitoring, which provides oversight
of drinking-water safety. Overall regulation of drinking-water suppliers 2.3. Spatial distribution of laboratories and required samples
in Colombia is the responsibility of the Superintendencia de Servicios
Públicos Domiciliarios (SSPD; Superintendency of Public Utilities), which The addresses of Colombian laboratories were translated into a
prescribes the number of operational monitoring samples required latitude and longitude via the geocoding facility in the Google Maps
from water suppliers each year. Operational monitoring is the responsi- Application Programming Interface (API). The API provided one or
bility of water service providers, who must report data to government more candidate locations for each address, either at address block,
via the Sistema Único de Información (SUI; Single Information System). street, town, or department level. All output addresses were then manu-
Results are used for operational purposes, and are additionally sent to ally reviewed. Where multiple candidate locations were provided or the
the SSPD for compliance assessment, and to the Ministries of Social match address was incorrect, interactive Google Maps searches were
Protection and of the Environment upon request. Larger, formal pro- used to identify the laboratory's most likely location.
viders with structured distribution systems are found mainly in urban We used 2005 census figures for the number of urban and rural
areas, whilst small, often informal providers are dispersed in rural and households with piped water to estimate the number of surveillance
peri-urban areas and small towns (Instituto Nacional de Salud, 2012). and operational samples required. For urban piped supplies, we
The frequency of monitoring and related reporting often meets or assumed that households with piped water in each of 1093 cabeceras
exceeds recommended levels among the larger, urban providers (SSPD, were served by a single, integrated supply system. To represent the
2011), whereas monitoring by small, informal providers is more sporadic spatial distribution of rural piped supply systems, we assumed that
and sometimes non-existent. piped systems would be located in areas of higher population density.
Surveillance monitoring in Colombia is performed by the Secretaría LandScan gridded estimates of ambient population were used to identi-
de Salud Municipal (Secretary of Municipal Health) and/or Secretaría fy rural pixels with population counts of more than 200 people (a pixel
J. Wright et al. / Science of the Total Environment 485–486 (2014) 643–652 645
being approximately 1 km2). Contiguous groups of such high population times to the nearest laboratory, and catchment areas of each laboratory,
density pixels were then grouped together and their total populations based on this impedance surface.
and centroids were calculated. Following a dasymetric mapping We used the cost surface technique in preference to travel times
approach (Mennis, 2003), municipality-level counts of rural households automatically derived from the Google Maps API, since the terms of
with piped water were then divided up between these patches of higher use for the API only allow for web site development, not research
population density, pro rata to each patch's estimated population (Google, 2013). We also used the cost surface approach because we
according to LandScan. The threshold value of 200 people per pixel for found via manual searches that Google Maps could not resolve drive-
population density was chosen so as to generate a plausible number of times for approximately 10% of trips from settlements to laboratories,
discrete piped supply systems, whilst also empirically matching these typically being the longer trips. However, to evaluate our cost sur-
municipality-level counts of households with piped water. Thus, this face, we randomly selected 50 cabecera-to-laboratory origin-destination
procedure incorporated all the piped supply users reported in the pairs and manually obtained drive time estimates for these journeys
census and distributed them into areas of higher population density. from Google Maps. For this random sample of 50 journeys, we then
Almost all rural piped users were accounted for by the procedure, except calculated Pearson correlation coefficients between the Google Maps
for 26,650 (0.6% of rural piped users) who lived in municipalities contain- drive-times, those from our cost surface, and simple straight-line
ing no densely populated patches. This threshold generated 5729 discrete distance from cabecera to laboratory.
rural supply systems, alongside the estimated 1093 urban supply systems
described above. This figure exceeded the 4327 providers registered with 2.5. Estimation of sample holding times
government (SSPD, 2011), thereby plausibly reflecting incomplete regis-
tration among water service providers nationally. Additionally, a count of Finally, we translated the estimated journey times between sample
rural households living in such high density areas better predicted the locations and laboratories into sample holding times, as an approximate
number of rural households with piped water in each municipality estimate of the time each sample spent after collection before arriving at
(n = 1090; R2 = 0.65), compared to a regression model based on the laboratory, reflecting the way monitoring teams operate in practice.
all rural households per municipality (R2 = 0.53). Data on daily monitoring patterns are limited, but a study by Crocker
So as to generate the number of samples required for monitoring, we and Bartram (2014) found that the number of samples collected per
then applied the regulatory guidance in Resolution 2115 (Ministerio de day ranged from 5 in rural India (West Bengal) to 24 in urban Jordan.
Protección Social and Ministerio de Ambiente Vivienda y Desarrollo Based on these figures and the authors' personal experience, we made
Territorial, 2007) to the cabecera and rural populations using piped some simple assumptions in order to estimate sample holding times.
water (Tables 1 and 2). In translating the number of households using In rural areas, we assumed that monitoring teams would spend 2 h trav-
a particular water supply type into a head count, we assumed that elling to the field and that 45 min would elapse between the collection
household size was constant within the rural versus urban parts of of successive samples. This latter figure included time for travel
each municipality. between monitoring locations, as well as the task of collecting the
sample. In urban areas, we assumed that monitoring teams would
spend an hour travelling to the field and 30 min would elapse between
2.4. Calculation of sample travel times to laboratory the collection of successive samples. These assumptions suggested that
10 samples would be collected per day in urban areas and 4 in rural
To calculate travel times to the nearest laboratory, we generated a areas. On the basis of these assumptions, we added a delay to account
gridded cost or impedance surface (Martin et al., 2002). Each grid cell for the collection of multiple samples on the same day and travel back
in a cost surface contains an estimate of the time required to cross the from the field to the journey times to the nearest laboratory. We first
cell. These impedance values can then be aggregated to identify the doubled and then halved these estimates to assess the sensitivity of
least cost route from a given origin to a given destination. In compiling our findings to these assumptions.
our impedance surface, we used the national speed limit of 80 km/h on
principal roads, 60 km/h on secondary roads, and 30 km/h in urban 3. Results
areas and on unpaved roads, and, following Blyde (2012), 11 km/h on
tracks. For 7 departments in southeastern Colombia, where some 3.1. Spatial distribution of laboratories and required samples
remoter settlements in the Amazon rainforest are accessible by riverboat
only and there is no road network (Armenteras et al., 2006), we assumed In total, there were 257 laboratories accredited for both microbiolog-
a speed of 22 km/h along navigable rivers. We then calculated journey ical and chemical testing, and a further 21 laboratories accredited for
microbiological testing only. Of these, the locations of 195 were auto-
matically geocoded via the Google Maps API, with the remaining 83
Table 1 being located through interactive online searches of Google Maps. In
Required minimum frequency of testing for operational monitoring for total coliforms/
terms of precision, 64 laboratories were geocoded to a matching nearby
E. coli by piped water service providers in Colombia.
Ministerio de Protección Social and Ministerio de Ambiente Vivienda y Desarrollo town name, 29 to a matching street name within a town, and 185 to a
Territorial (2007). matching address block. The distribution of these laboratories is
shown in Fig. 1.
Population served by piped Frequency Minimum no of samples to be
supply system analysed per period
Fig. 2 shows the estimated number of microbiological surveillance
samples required annually from piped supply systems under national
≤2500 Monthly 1
regulations. In total 62,502 samples were required for the whole
2501–10,000 3
10,001–20,000 Fortnightly 4 country, with the majority (36,720) being required for rural areas. The
20,001–100,000 Weekly 8 estimated number of operational samples was larger (219,434) and
100,001–250,000 Daily 3 dominated by urban areas (140,098 samples).
251,000–500,000 5
500,001–1,000,000 6
1,000,001–1,250,000 7 3.2. Travel times to nearest laboratory for water samples
1,250,001–2,000,000 10
2,000,001–4,000,000 12 Fig. 3 shows the relationship between journey times derived from a
N4,000,000 12 plus 1 per each additional cost surface and those calculated via Google Maps for a random sample
200,000 population served
of 50 journeys from cabeceras to the nearest laboratory. Overall, the cost
646 J. Wright et al. / Science of the Total Environment 485–486 (2014) 643–652
Table 2 surface generated slightly longer journey times relative to Google Maps,
Stipulated minimum frequency of testing for health surveillance in Colombia. but there was a close correspondence between the two sets of journey
Ministerio de Protección Social and Ministerio de Ambiente Vivienda y Desarrollo
Territorial (2007).
time estimates (r = 0.79; p b 0.001). Relative to these cost surface-
derived estimates of journey times, there was a somewhat weaker rela-
Population served by Minimum no of samples to tionship between straight-line distances and journey times estimated
piped supply system be analysed per month
via Google Maps (r = 0.71; p b 0.001 — data not shown).
≤2500 0.5 Fig. 4 shows the estimated travel time to the nearest laboratory
2501–10,000 1
accredited for microbiological testing, based on the cost surface. Travel
10,001–20,000 2
20,001–100,000 5 times to the nearest laboratory in much of the more sparsely populated
100,001–250,000 10 areas, particularly in the south and east, exceed the 6 hours' sample
251,000–500,000 15 holding time recommended by WHO.
500,001–1,000,000 30 Fig. 5 shows the distribution of samples required for microbiological
1,000,001–2,000,000 60
surveillance and operational monitoring of piped supplies, versus journey
2,000,001–4,000,000 96
N4,000,000 96 plus 1 for each additional time to the nearest accredited laboratory. 4.5% (1158) of urban surveil-
50,000 population served lance samples and 10.7% (3924) of rural surveillance samples were
between 6 and 24 hours' travel from the nearest laboratory, with a
Fig. 1. Map of Colombian laboratories accredited for microbiological testing under Resolution 431 (inset: the archipelago of San Andres and Providencia).
J. Wright et al. / Science of the Total Environment 485–486 (2014) 643–652 647
Fig. 2. Map of number of microbiological surveillance samples required per year for piped water supplies (inset: the archipelago of San Andrés and Providencia).
further 0.8% (216) of urban samples and 1.1% (396) more tha 24 hours'
travel from the nearest laboratory. 3.2% (4472) of urban operational
samples were 6 to 24 hours' travel from the nearest laboratory, as were
10.1% (8028) of rural operational samples. 0.4% (560) of urban samples
and 1.0% (792) of rural samples were more than 24 hours' travel away.
An estimated rural population of 275,000 was served by piped systems
that were more than 6 hours' travel from the nearest laboratory.
Fig. 6 shows the estimated sample holding times for rural and urban
areas. For the base case of 4 samples per day in rural areas, 29.4% (25.8–
32.4%) of operational samples and 31.0% (27.2–34.1%) of surveillance
samples were estimated to arrive in the laboratory after the recom-
mended 6 h for holding samples. For the base case of 10 samples per
day in urban areas, these proportions were 11.9% (9.5–13.5%) for oper-
ational samples and 16.9% (13.5–20.3%) for surveillance samples. The
Fig. 3. Google Maps versus cost surface drive times for a random sample of 50 journeys results from the sensitivity analysis, derived by doubling and halving
from cabecera to the nearest laboratory. the number of samples collected per day, are shown in brackets.
648 J. Wright et al. / Science of the Total Environment 485–486 (2014) 643–652
Fig. 4. Travel times (in hours) to the nearest accredited microbiological laboratory.
4. Discussion total coliforms and Escherichia coli. Nationally, total coliforms, residual
chlorine and E. coli are the parameters that most frequently do not com-
4.1. Implications for drinking-water quality monitoring ply with guideline values. The IRCA Water Quality Index values for
piped water from 2010 were generally poorer in rural municipalities
There is a growing emphasis on ensuring equity in safe water access than in urban ones (SSPD, 2011), suggesting greater rural drinking-
(Luh et al., 2013; Yang et al., 2013), but despite this, rural areas continue water contamination for Colombia. One difficulty in addressing this
to lag behind urban areas (de Barros et al., 2009). Systematic review challenge is the problem of monitoring drinking-water in rural areas
evidence drawn from low and middle income countries (Bain et al., and small towns, given that laboratory facilities are typically in urban
accepted) suggests that the microbial quality of rural water supplies is centres (Fig. 1). Although the number of tests undertaken by larger
lower than in urban areas and there is evidence from Peru to suggest suppliers exceeded minimum requirements in many cases, the SSPD
that this applies even for the same type of water supply (Miranda (2011: p. 188) note: “In 2010, there was no information about the quality
et al., 2010). A study from seven developing countries found that mon- of water consumed by 30.3% of the population, because they are found in
itoring of small water supplies has much lower compliance for water zones that are difficult to access for surveillance by the sanitary authorities
safety parameters than large water supplies (Crocker and Bartram, (usually in rural areas)”. The results presented here quantify the implica-
2014). Specifically, it found that the compliance for large supplies is tions for sample holding times of undertaking water quality monitoring
100%, whilst smaller supplies have a compliance of approximately via fixed laboratory infrastructure. This difficulty in accessing fixed
60%. In Colombia, a Water Quality Risk Index (Indice del Riesgo de la laboratories potentially exacerbates a rural–urban disparity in access
Calidad del Agua — IRCA) is used to assess drinking-water quality, calcu- to safe drinking-water that has been documented for much of Latin
lated as a weighted linear combination of various parameters including America (de Barros et al., 2009). It may also impact on more remote
J. Wright et al. / Science of the Total Environment 485–486 (2014) 643–652 649
Fig. 5. Distribution of microbiological samples required from piped water for (a) surveillance and (b) operational monitoring, versus journey time to the nearest laboratory.
small towns, which are classed as urban in Colombia where they act as used to optimise the cost-effectiveness of different combinations of
cabecera. monitoring via fixed laboratories, mobile facilities, and field test kits,
There are several ways in which the analysis of sample transporta- as well as assess their equity implications. For example, collecting mul-
tion times presented here could be used to plan water quality monitor- tiple samples per trip lowers the transportation cost per sample, though
ing in rural areas. First, the travel time estimates could be used to target it lengthens sample holding times. Finally, by developing understanding
use of field-based test kits to distant water supplies, for example those of the logistical implications of any regulatory changes for sample trans-
beyond the 6-hour travel time contour in Fig. 5. Some microbiological port, the analysis presented here could be used to inform cost–benefit
testing procedures such as the DelAgua field incubator and test kit for analyses of any future regulatory changes to the surveillance and oper-
thermotolerant coliforms and the H2S test (Wright et al., 2012) are ational monitoring regimes.
intended for field use and do not require a laboratory. Where appropri-
ate and where cell phone coverage exists, test results from such field 4.2. Uncertainties affecting sample holding times
kits can be transmitted via cell phone as a means of reporting back to
the communities consuming water, a regulator or a service provider As a preliminary analysis, our findings are likely to be influenced by a
(Rivett et al., 2013). Testing potentially could also be undertaken via number of assumptions and limitations in the underlying data. The
mobile laboratories that visit more remote supply systems according census data are from 2005 and we have not attempted to project popu-
to a rota. Second, the spatial distribution of demand for water quality lation estimates forward to 2012. Our analysis will also be affected by
testing based on surveillance and operational monitoring regulations the uncertainty underlying our geocoding and travel time estimation
could be used to optimise the location of any new accredited laborato- techniques. In terms of geocoding, a study in urban Brazil (Davis and
ries, so as to maximise the laboratory network's coverage. By exploring de Alencar, 2011) found a similar pattern of Google Maps API geocoding
the population coverage resulting from the related investments in precision to that found here, with most addresses coded to address
transport, infra-structure and staff time, such an analysis could also be block level, some to town level, and the fewest to street level. They
650 J. Wright et al. / Science of the Total Environment 485–486 (2014) 643–652
Fig. 6. Distribution of number of microbiological samples required from piped water for (a) surveillance and (b) operational monitoring, versus estimated sample holding time.
found that address block level geocoding was usually accurate to within field staff collect on a single day, since this affects the delay in their
150 m, though accuracy was lower in low income neighbourhoods. transportation to the laboratory. These assumptions were based on
Elsewhere, a US-based study (Roongpiboonsopit and Karimi, 2010) data from other countries.
that evaluated geocoding via the Google maps API suggested that its A number of assumptions also underpin our findings. We assumed
performance was weaker in rural and industrial areas. On this basis, it that samples were processed in-country at the nearest laboratory and
is likely that the accuracy of the laboratory address geocoding was that there were no capacity constraints on sample processing by labora-
lower in rural areas and therefore rural water transportation times tories. We also ignored the regulatory requirement to undertake repeat
will be more uncertain. sampling in response to detection of substantial contamination, which
Our cost surface calculations ignore the effect of road quality, traffic would particularly affect rural areas, given the IRCA values described
congestion, and driving over the speed limit. The cost surface output above (SSPD, 2011). We assumed that samples were collected from
matched travel time estimates from Google Maps more closely than the centroid of each urban area and transported by road or boat. In
straight-line distances, and enabled us to estimate the travel time modelling the spatial distribution of operational and surveillance tests
contours in Fig. 1 and catchment areas for laboratories. However, the required, we used census data rather than data from water utilities
cost surface output remained an approximation of the travel times esti- and assumed that the rural population using piped supplies was
mated by Google Maps. The assumption about speeds for river travel concentrated in more densely populated areas. We assumed that each
also affects estimated journey times for samples from piped systems settlement had a single integrated water distribution system and did
in the more remote southeastern departments of Colombia. More gen- not attempt to project the 2005 census population forward to 2012,
erally, there remains very little independent literature internationally the year when our data on laboratories was collated. WHO's guidance
evaluating the accuracy of the travel times generated by Google Maps, is that the delay between sample collection and analysis should not
which we used alongside the cost surface methodology. Our findings exceed 6 h and we have only considered the time from collection to
are also sensitive to assumptions about the number of samples that analysis in a laboratory. We have not considered any subsequent delay
J. Wright et al. / Science of the Total Environment 485–486 (2014) 643–652 651
prior to sample analysis in the laboratory. Collectively, these assump- between 28% and 43% of monitoring costs depending on the setting,
tions are likely to lead to an under-estimate in the sample transportation based on data from seven countries.
times presented in Fig. 4. We did not account for 0.6% of geographically In using GIS for public health planning, there is a very large and well
dispersed piped water users and we focus here on piped water, ignoring established literature on the location of healthcare facilities and patient
the 12% of mainly rural Colombians who use other water sources. access to clinics and hospitals (Cromley and McLafferty, 2012). However,
Urban–rural disparities in sample holding times may therefore be even rather than focussing on population travel to healthcare facilities, here
greater than our results suggest. we have examined the logistics and distribution of samples to laboratory
The evidence underpinning current guidelines suggests a variable facilities and therefore focussed on the movement of goods. The ap-
and uncertain impact of sample holding times on subsequent indicator proach could be used to examine medical commodities where the cold
bacteria enumeration, depending on microorganisms, water treatment chain and associated transportation issues are important, such as vac-
prior to sampling, and physico-chemical sample characteristics cines and the transport of anti-venoms as suggested by Gutierrez et al.
(Edberg et al., 2000). For example, McDaniels et al. (1985) found a (2009). In so doing, the approach could be used to support other public
significant but gradual decline of less than half a log in the coliform health initiatives in remote rural communities.
counts for piped water samples stored for up to 50 h at 5 °C. Pope
et al. (2003) noted a significant decline in E. coli counts from some sur- 4.4. Conclusion
face water samples stored at 4 °C for 24 h, whilst a more recent study of
surface waters (Aulenbach, 2010) identified a decline in total coliform The approach presented here provides a new means of using GIS to
counts in surface water stored at 4 °C for more than 27 h. However, support water quality monitoring, particularly in remote areas. Our
the evidence suggests that holding times have a greater impact on sub- findings demonstrate the difficulty of undertaking microbiological
sequent indicator bacteria enumeration when samples are stored at monitoring in rural areas and small towns from a fixed laboratory
higher temperatures and the impact is typically to reduce the bacterial network in Colombia. Our GIS-based approach could be adapted to
count. Thus, the impact of long (N 24 h) and even relatively short optimise monitoring strategies and support planning of testing and
(b6 h) sample holding times on the enumeration of indicator bacteria transportation infra-structure development. Such logistical challenges
remains uncertain, especially if sample temperatures are not carefully to monitoring in remote areas exist not only in Colombia but in other
controlled. middle and low income countries, so the approach could usefully be
applied elsewhere.
4.3. Future research
Conflicts of interests
The approach presented here could be developed further in a variety
of respects. Given that sample transportation poses challenges in many Profs Gundry and Bartram and Dr. Wright have benefited from the
other countries, and that the spatial distribution of laboratories and grant Aquatest funded by Bill & Melinda Gates Foundation, whose
water source types vary between countries (Rahman et al., 2011), the main purpose is to develop a low cost, field test for microbial contami-
study could be replicated usefully elsewhere. Here, we have used census nation of drinking water, suitable for use in developing countries. Bain
data to model demand for water quality monitoring. However, where and Crocker were employed on this grant. Prof SW Gundry was also
these exist, other data sources could be used instead, most notably the the named, principal investigator for the grant. Professor Gundry is
databases developed via water point mapping in sub-Saharan Africa the sole inventor on granted European patent number 1960104 entitled
by WaterAid (Jimenez and Perez-Foguet, 2008) or georeferenced data- ‘Apparatus for determining the presence of a contaminant in a sample of
bases of piped suppliers. Given the need for electricity to make ice for water or other fluid’. This patent is also granted in South Africa and
sample transport and to power some models of incubators and some member countries of the African Intellectual Property Organisation
field test kits, the approach could be further refined through the inclu- (OAPI) and is pending in 9 other countries or groups of countries. Prof.
sion of a map layer depicting electricity coverage and perhaps also cell Gundry and Bain are co-inventors on International Patent Applications
phone coverage to facilitate the communication of test results. To PCT/GB2010/050728 and PCT/GB2012/050452 both entitled ‘Apparatus
validate the broad pattern of travel times to laboratory identified for testing the quality of a fluid sample’. Other than these possible com-
here, our findings could be corroborated via examination of actual peting interests, there are no others known.
(as opposed to prescribed) spatial patterns of testing and interviews
with laboratory and environmental health field staff. Examining spatial
Acknowledgements
patterns of actual testing in relation to the distribution of water supplies
could also help identify gaps in the monitoring network. Much the same
This research has been partially funded by the Bill & Melinda Gates
difficulties of sample transportation affect the regulation and monitor-
Foundation, under a grant to the University of Bristol. The grant's
ing of environmental waters. The analysis could be replicated for envi-
purpose is to develop a low cost test of microbial water quality for use
ronmental and recreational/bathing waters, which in Colombia would
in developing countries. The Bill & Melinda Gates Foundation had no
entail drawing on the list of laboratories accredited by the Instituto de
role in the design and conduct of this study.
Hidrologica, Meteorologia y Estudios Ambientales (IDEAM; Institute of
Hydrology, Metrology and Environmental Studies) (Instituto de
References
Hidrologica Meteorologia y Estudios Ambientales, 2012).
The approach could also be developed further to investigate the Armenteras D, Rudas G, Rodriguez N, Sua S, Romero M. Patterns and causes of deforesta-
sample transportation component of different regulatory and laboratory tion in the Colombian Amazon. Ecol Indic 2006;6:353–68. http://dx.doi.org/10.1016/
j.ecolind.2005.03.014.
provision scenarios for water quality monitoring. For example, Blyde Aulenbach B. Bacteria holding times for fecal coliform by mFC agar method and total
(2012) translated Colombian freight travel times from a GIS-based coliform and Escherichia coli by Colilert-18 Quantitray method. Environ Monit Assess
analysis into economic costs by examining driver wage costs, fuel and 2010;161:147–59.
Bain RES, Bartram JK, Elliott M, Matthews RL, McMahan L, Tung R, et al. A summary
vehicle maintenance, suggesting that this is feasible. Previous research
catalogue of microbial drinking water tests for low and medium resource
suggests that the transportation costs of water quality monitoring settings. Int J Environ Res Public Health 2012a;9:1609–25. http://dx.doi.org/10.
are significant. Castillo (1994) compared the costs of field versus 3390/ijerph9051609.
laboratory-based testing in Chile, and estimated that transportation- Bain RES, Gundry SW, Wright JA, Yang H, Pedley S, Bartram JK. Accounting for water
quality in monitoring access to safe drinking-water as part of the Millennium
related costs contributed 35% of the total costs of laboratory testing. Development Goals: lessons from five countries. Bull World Health Organ 2012b;
Crocker and Bartram (2014) estimated that transportation contributed 90:228–35. http://dx.doi.org/10.2471/BLT.11.094284.
652 J. Wright et al. / Science of the Total Environment 485–486 (2014) 643–652
Bain RES, Cronk R, Wright JA, Yang H, Bartram JK. Fecal contamination of drinking water in Ministerio de Protección Social, Ministerio de Ambiente Vivienda y Desarrollo Territorial.
developing countries: a systematic review and meta-analysis; 2014d (accepted). Resolución 2115. Bogota, Colombia: Government of Colombia; 2007 [Accessed: Jan
Blyde J. Paving the road to export: assessing the trade impact of road quality. Int Econ J 2014. Available from: www.minambiente.gov.co/documentos/res_2115_220707.pdf].
2012;27:663–81. http://dx.doi.org/10.1080/10168737.2012.733721. Ministerio de Salud y Protección Social. Resolución 431. Bogota, Colombia: Government of
Castillo G. Water quality control Chile — phase II: final report. International Development Colombia; 2012 [Accessed: Jan 2014. Available from: www.dssa.gov.co/__media__/
Research Centre; 1994 [Accessed: Jan 2012. Available from:http://hdl.handle.net/ dssa/dssa.gov.co/documentos/resolucion-431-de-2012-laboratorios-de-agua-
10625/13856]. autorizados.pdf].
Chuang P, Trottier S, Murcott S. Comparison and verification of four field-based microbi- Miranda M, Aramburu A, Junco J, Campos M. Situación de la calidad de agua para consumo en
ological tests: H2S test, Easygel®, Colilert®, Petrifilm TM. J Water Sanitation Hyg Dev hogares de niños menores de cinco años en Perú, 2007-2010. Rev Peru Med Exp Salud
2011;1:68–85. http://dx.doi.org/10.2166/washdev.2011.026. Publica 2010;27:506–11. http://dx.doi.org/10.1590/S1726-46342010000400003.
Crocker J, Bartram JK. Comparison and cost analysis of drinking water quality monitoring Onda K, LoBuglio J, Bartram JK. Global access to safe water: accounting for water quality
requirements versus practice in seven developing countries. Int J Hyg Environ Health and the resulting impact on MDG progress. Int J Environ Res Public Health 2012;9:
2014. 880–94. http://dx.doi.org/10.3390/ijerph9030880.
Cromley EK, McLafferty S. GIS and public health. Guilford Press; 2012. Pope M, Bussen M, Feige M, Shadix L, Gonder S, Rodgers C, et al. Assessment of the effects
Davis Jr CA, de Alencar RO. Evaluation of the quality of an online geocoding resource in of holding time and temperature on Escherichia coli densities in surface water samples.
the context of a large Brazilian city. Trans GIS 2011;15:851–68. http://dx.doi.org/10. Appl Environ Microbiol 2003;69(10):6201–7.
1111/j.1467-9671.2011.01288.x. Rahman Z, Crocker J, Chang K, Khush R, Bartram JK. A comparative assessment of institu-
de Barros RP, Ferreira FHG, Vega JRM, Chanduvi JS. Measuring inequality of opportunities tional frameworks for managing drinking water quality. J Water Sanitation Hyg Dev
in Latin america and the Caribbean. The World Bank; 2009. http://dx.doi.org/10. 2011;1:242–58. http://dx.doi.org/10.2166/washdev.2011.002.
1596/978-0-8213-7745-1 [Accessed: Jan 2014]. Rivett U, Champanis M, Wilson-Jones T. Monitoring drinking water quality in South
Departamento Administrativo Nacional de Estadistica. Censo Basico 2005. 2013. Africa: designing information systems for local needs. Water SA 2013;39:409–14.
[Accessed: Jan 2014. Available from: www.dane.gov.co]. http://dx.doi.org/10.4314/wsa.v39i3.10.
Dobson J, Bright E, Coleman P, Durfee R, Worley B. Landscan: a global population database Roongpiboonsopit D, Karimi HA. Comparative evaluation and analysis of online geocoding
for estimating populations at risk. Photogramm Eng Remote Sens 2000;66:849–57. services. Int J Geogr Inf Sci 2010;24:1081–100. http://dx.doi.org/10.1080/
Edberg S, Rice E, Karlin R, Allen M. Escherichia coli: the best biological drinking water 13658810903289478.
indicator for public health protection. J Appl Microbiol 2000;88:106S–16S. SSPD. Diagnóstico de la calidad del agua suministrada por las empresas Prestadoras del
Google. The Google distance matrix API. 1 Nov 2013, 2013. Accessed: Jan 2014. Available servicio de acueducto en Colombia. 2009–2010; 2011 [Accessed: Jan 2014. Available
from: https://developers.google.com/maps/documentation/distancematrix, 2013. from: www.superservicios.gov.co/home/c/document_library/].
Gutierrez JM, Fan HW, Silvera CL, Angulo Y. Stability, distribution and use of antivenoms UN-OCHA. Humanitarian response common and fundamental operational datasets regis-
for snakebite envenomation in Latin America: report of a workshop. Toxicon 2009; try. 2013. [Accessed: Jan 2014. Available from: https://cod.humanitarianresponse.
53:625–30. http://dx.doi.org/10.1016/j.toxicon.2009.01.020. info].
Instituto de Hidrologica Meteorologia y Estudios Ambientales. Lista de Laboratorios US EPA. Manual for the certification of laboratories analyzing drinking water: criteria and
Ambientales Acreditados por el IDEAM. Bogota, Colombia: IDEAM; 2012 [Accessed: procedures — quality assuranceFifth ed. ; 2005 [EPA 815-R-05-004, Cincinnati, Ohio].
Jan 2014. Available from: www.andi.com.co/archivos]. WHO. Guidelines for drinking-water quality. Surveillance and control of community sup-
Instituto Nacional de Salud. Programa de Vigilancia por Laboratorio de la Calidad de Agua plies. 2. Geneva, Switzerland: World Health Organization; 1997 [Accessed: Jan 2014.
para Consumo Humano 978-958-13-0147-8; 2011 [Bogota, Colombia]. Available from: http://www.who.int/water_sanitation_health/dwq/gdwq2v1/en/
Instituto Nacional de Salud. Estado de la vigilancia de la calidad del agua para consumo index2.html].
humano en Colombia, periodo 2007–2011. 2012. [Accessed: Jan 2013. Available WHO. Guidelines for drinking-water quality. Geneva, Switzerland: World Health Organi-
from: http://www.ins.gov.co/sivicap/Documentacion/]. zation; 2011 [Accessed: Jan 2014. Available from: http://whqlibdoc.who.int/publica-
Jimenez A, Perez-Foguet A. Improving water access indicators in developing countries: a tions/2011/9789241548151_eng.pdf].
proposal using water point mapping methodology. Water Sci Technol Water Supply WHO, UNICEF. Drinking-water: equity, safety and sustainability, New York, USA; 2011
2008;8:279–87. http://dx.doi.org/10.2166/ws.2008.072. [Accessed: Jan 2014. Available from: http://www.wssinfo.org/fileadmin/user_up-
Luh J, Baum R, Bartram JK. Equity in water and sanitation: developing an index to mea- load/resources/report_wash_low.pdf].
sure progressive realization of the human right. Int J Hyg Environ Health 2013;216: WHO, UNICEF. Progress of sanitation and drinking-water: 2013 update. 2013. [Accessed:
662–71. http://dx.doi.org/10.1016/j.ijheh.2012.12.007. Jan 2014. Available from: http://www.wssinfo.org/fileadmin/user_upload/resources/
Martin D, Wrigley H, Barnett S, Roderick P. Increasing the sophistication of access mea- JMPreport2013.pdf].
surement in a rural healthcare study. Health Place 2002;8:3–13. http://dx.doi.org/ Wright JA, Yang H, Walker K, Pedley S, Elliott J, Gundry SW. The H2S test versus standard
10.1016/S1353-8292(01)00031-4. indicator bacteria tests for faecal contamination of water: systematic review and
McDaniels A, Bordner R, Gartside P, Haines J, Brenner K, Rankin C. Holding effects on co- meta-analysis. Trop Med Int Health 2012;17:94–105. http://dx.doi.org/10.1111/j.
liform enumeration in drinking-water samples. Appl Environ Microbiol 1985;50(4): 1365-3156.2011.02887.x.
755–62. Yang H, Bain RES, Bartram JK, Gundry SW, Pedley S, Wright JA. Water safety and inequality
Mennis J. Generating surface models of population using dasymetric mapping. Prof Geogr in access to drinking-water between rich and poor households. Environ Sci Technol
2003;55:31–42. http://dx.doi.org/10.1111/0033-0124.10042. 2013;47:1222–30. http://dx.doi.org/10.1021/es303345p.