Provided for non-commercial research and educational use only.
Not for reproduction or distribution or commercial use.
This article was originally published by IWA Publishing. IWA Publishing recognizes
the retention of the right by the author(s) to photocopy or make single electronic
copies of the paper for their own personal use, including for their own classroom use,
or the personal use of colleagues, provided the copies are not offered for sale and
are not distributed in a systematic way outside of their employing institution.
Please note that you are not permitted to post the IWA Publishing PDF version of
your paper on your own website or your institution’s website or repository.
Please direct any queries regarding use or permissions to email@example.com
Unsealed tubewells lead to increased fecal contamination
of drinking water
Peter S. K. Knappett, Larry D. McKay, Alice Layton, Daniel E. Williams,
Md. J. Alam, Brian J. Mailloux, Andrew S. Ferguson, Patricia J. Culligan,
Marc L. Serre, Michael Emch, Kazi M. Ahmed, Gary S. Sayler
and Alexander van Geen
Bangladesh is underlain by shallow aquifers in which millions of drinking water wells are emplaced
without annular seals. Fecal contamination has been widely detected in private tubewells. To
evaluate the impact of well construction on microbial water quality 35 private tubewells (11 with
intact cement platforms, 19 without) and 17 monitoring wells (11 with the annulus sealed with
cement, six unsealed) were monitored for culturable Escherichia coli over 18 months. Additionally,
two ‘snapshot’ sampling events were performed on a subset of wells during late-dry and early-wet
seasons, wherein the fecal indicator bacteria (FIB) E. coli, Bacteroidales and the pathogenicity genes
eltA (enterotoxigenic E. coli; ETEC), ipaH(Shigella) and 40/41 hexon (adenovirus) were detected using
quantitative polymerase chain reaction (qPCR). No difference in E. coli detection frequency was
found between tubewells with and without platforms. Unsealed private wells, however, contained
culturable E. coli more frequently and higher concentrations of FIB than sealed monitoring wells (p <
0.05), suggestive of rapid downward ﬂow along unsealed annuli. As a group the pathogens ETEC,
Shigella and adenovirus were detected more frequently (10/22) during the wet season than the dry
season (2/20). This suggests proper sealing of private tubewell annuli may lead to substantial
improvements in microbial drinking water quality.
Peter S. K. Knappett (corresponding author)
Larry D. McKay
Department of Earth & Planetary Sciences,
University of Tennessee,
Knoxville, TN 37996-1410, USA
Peter S. K. Knappett
Alexander van Geen
Lamont-Doherty Earth Observatory of Columbia
Palisades, NY 10964, USA
Daniel E. Williams
Gary S. Sayler
Center for Environmental Biotechnology,
University of Tennessee,
Knoxville, TN 37996-1605, USA
Md. J. Alam
Kazi M. Ahmed
Department of Geology,
University of Dhaka,
Dhaka 1000, Bangladesh
Brian J. Mailloux
Department of Environmental Science,
New York, NY 10027, USA
Andrew S. Ferguson
Patricia J. Culligan
Civil Engineering & Engineering Mechanics,
New York, NY 10027, USA
Marc L. Serre
Department of Environmental Sciences &
University of North Carolina – Chapel Hill,
Department of Geography,
University of North Carolina – Chapel Hill,
Carolina Population Center,
University of North Carolina – Chapel Hill,
adenovirus, Asia, Bacteroidales, E. coli, Shigella, tubewells
565 © IWA Publishing 2012 Journal of Water and Health
Fecal bacteria and viruses have been detected in ground-
water wells emplaced in aquifers of diverse geologic
material (Rudolph et al. ; Abbaszadegan et al. ;
Embrey & Runkle ; Borchardt et al. ; Johnson
et al. ; Kozuskanich et al. ). In developed countries,
such as the USA, wells are typically sealed with an expand-
ing clay which ﬁlls the annulus between the well casing and
the surrounding aquifer sediments from ground surface to
approximately 1 m above the screened interval to prevent
‘short circuiting’ by downward ﬂow of contaminated surface
water or shallow groundwater. In developing countries,
such as Bangladesh, the annulus of shallow drinking water
wells, referred to as tubewells, is typically ﬁlled with soil
or sediments obtained during drilling; a measure unlikely
to prevent rapid annular ﬂow. Recent programs have
encouraged the construction of cement platforms and drain-
age channels around well-heads to drain wash water before
it inﬁltrates down the tubewell annulus, but there is insufﬁ-
cient information to indicate whether this is successful
(Luby et al. ; Leber et al. ; van Geen et al. ).
Although tubewells are considered an ‘improved’ drink-
ing water source by the World Health Organization
(WHO ), these tubewells still do not have adequate pro-
tection from annular ﬂow and are considered unsealed
(personal communication, Peter Ravenscroft). The goal of
this study was to evaluate the impact of tubewell construc-
tion (speciﬁcally annular seals and cement platforms) on
the levels of human exposure to fecal contamination and
Over 10 million of these tubewells provide drinking
water for millions of inhabitants throughout rural Bangla-
desh. In this country, 11% of all deaths are estimated to be
caused by diarrheal disease (Streatfield et al. ) with
recent studies suggesting that sustained levels of diarrheal
disease are caused in part by drinking untreated ground-
water (Escamilla et al. , ; Wu
et al. ). The
widespread fecal indicator bacteria (FIB) and bacterial and
viral pathogens found in tubewells (Luby et al. ; van
Geen et al. ; Leber et al. ; Ferguson et al. ) are
known to be predictive of diarrheal disease in diverse popu-
lations around the world (Gundry et al. ).
Bangladesh is underlain by shallow aquifers consisting
of unconsolidated sand and silt laid down by streams and
rivers ﬂowing through the Ganges-Brahmaputra delta
(Goodbred & Kuehl ; Weinman et al. ). In addition
to annular ﬂow around the tubewell, it is possible that fecal
contamination enters aquifers and tubewells through inﬁl-
tration from latrines or seepage from the many ponds and
canals found in rural villages (Knappett et al. a).
Recent ﬁeld experiments in Bangladesh indicate efﬁcient
spatial removal of FIB in typical medium-grained aquifers,
removal within 13 m from leaking latrine
ponds (Knappett et al. ). This ﬁnding suggests that
fecal contamination of unconﬁned aquifers from latrines,
ponds and other sources should have relatively limited
spatial extent. Contamination would be expected to be
even more limited in areas where the aquifer is overlain by
silt or clay layers. This is not consistent, however, with the
widespread occurrence of FIB in tubewells located in both
unconﬁned sandy aquifers and aquifers overlain by silt
(Leber et al. ), suggesting the presence of rapid ﬂow path-
ways, such as annular ﬂow around the tubewell casing.
Recent programs in Bangladesh and other developing
countries have encouraged the construction of 2–3m
cement platforms with drainage channels around tubewells
to remove standing water and reduce the likelihood of annu-
lar ﬂow of contaminated water (Luby et al. ; WHO/
UNICEF ). The utilization of cement platforms is sup-
ported by studies in rural Africa (Godfrey et al. ),
which found links between the presence of standing water
around the well-head and poor microbial water quality in
Studies spanning multiple villages and seasons in Ban-
gladesh, however, have shown that E. coli detection
frequency in private tubewells is typically insensitive to
both the presence and quality of a platform (Luby et al.
; van Geen et al. ; Leber et al. ) leading some
to suggest that tubewells are not subject to annular ﬂow
(Luby et al. ). Only one study (Escamilla et al. ),
considering a subset of approximately 90 wells from the
study by van Geen et al.(), found that the presence of
a well platform correlated to lower E. coli detection
566 P. S. K. Knappett et al.
Unsealed tubewells lead to increased fecal contamination of drinking water Journal of Water and Health
frequency in private wells. This effect was only signiﬁcant
during the early monsoon period (April–June) and at no
other time of year (Escamilla et al. ). Likely more critical
than the presence of an intact platform in preventing annu-
lar ﬂow is the presence of a seal between the borehole and
the annulus of a well from surface to screened depth.
E. coli prevalence peaks in tubewells during the wet
season (Leber et al. ), being closely associated with ante-
cedent rainfall events (van Geen et al. ). Increases in
allochthonous bacteria concentrations in aquifers following
rainfall events are attributed to vertical ﬂushing when bac-
teria are both introduced to the vadose zone and
mobilized from grain surfaces due to increasing water con-
tent and shear velocity (De Novio et al. ; Pronk et al.
). Pit latrines and ponds are the primary repository of
human feces in Bangladesh and the water table is very
shallow, lying 1–5 m below the surface, giving ample oppor-
tunity for contamination of the water table from these
numerous point sources in densely populated rural villages.
Further, ﬂooding is widespread during the late wet season,
potentially widening the spatial extent of fecal contami-
nation sources at the surface during the time of year when
the vadose zone is thinnest (van Geen et al. ; Knappett
et al. ).
The objective of this study is to quantify the impact of
private well construction on the frequency and concen-
trations of FIB detected in tubewells within a sandy
aquifer underlying a village in rural Bangladesh. A further
objective is to assess the utility of molecular FIB markers
in predicting the year-round risk of fecal contamination of
a well emplaced within a shallow, reducing aquifer. Quanti-
tative PCR (qPCR) was used to screen for the pathogenicity
genes elt A (enterotoxigenic E. coli; ETEC), ipaH(Shigella)
and 40/41 hexon (adenovirus) to assess the ability of FIBs to
indicate the presence/absence of pathogens.
Some of the monthly cultured E. coli measurements
reported in this study for private tubewells have been pub-
lished previously in the context of a larger study ﬁnding a
broad negative correlation between E. coli detection fre-
quency and arsenic (van Geen et al. ). The previously
published monthly E. coli detection frequencies for 35 pri-
vate wells at Site K is compared here to synoptic
measurements made on 17 additional monitoring wells.
Additionally, the previously unpublished results of two
snapshot monitoring events for FIB DNA and pathogens on
a subset of private and monitoring wells are presented here.
The village of Char Para is referred to herein as Site K
(Knappett et al. a, b). Char Para overlies a sand bar
deposit of the neighboring ‘Old Brahmaputra’ river, which
ﬂows throughout Araihazar upazilla and transported far
more water and sediment, in the past, than it does today
(Weinman et al. ). Bangladesh has a dry season
during which the region receives little rain from November
through May. The dry season is followed by the monsoon, a
period of 3–4 months during which Bangladesh receives the
vast majority of its total rainfall for the year. In this study,
rainfall was measured using a HOBO weather station
(ONSET, Bourne, MA) in the region of Matlab, 50 km
south of Site K. The unconﬁned water table at Site K ﬂuctu-
ates throughout the year from 4 m below the surface at the
end of the dry season to within 1 m of the surface during
the wet season (Knappett et al. ). Although there is an
inﬂuence of the local river which bounds three sides of the
village, lateral hydraulic gradients throughout Site K are
small and affected by irrigation pumping in the surrounding
rice ﬁelds. Approximately 1,500 people live in Site K.
Roughly 50 ponds and 180 latrines are scattered throughout
the site. Half of these latrines spill efﬂuent onto the open
ground (Knappett et al. a), consistent with the country-
wide improved sanitation coverage for rural Bangladesh in
2010 of only 43% (WHO/UNICEF ).
Private wells in Bangladesh, referred to as ‘tubewells’,
are inexpensive PVC pipes with 1–1.5 m screened intervals
equipped with iron hand pumps and are typically screened
at depths ranging from 8 to 30 m below the ground surface
(van Geen et al. ). As of 2009 our exhaustive survey of
Char Para (Site K) indicated that it contained 144 private
tubewells. Therefore 1 tubewell supplies drinking water for
10 people, as there are 1,500 inhabitants in the village
(Knappett et al. a ). Drillers will typically drill no further
than the depth necessary to ensure a year-round supply. In
the absence of poorly conductive surface deposits, tubewells
567 P. S. K. Knappett et al.
Unsealed tubewells lead to increased fecal contamination of drinking water Journal of Water and Health
will be quite shallow as they are at Site K (Leber et al. ).
The minimum and maximum depths of these private tube-
wells, reported by the owners, were 6.1 and 91.5 m,
respectively, with a median depth of 9.1 m. The positions
of all wells in this study were determined using high accu-
racy (sub-meter) GPS using a Trimble GeoXH receiver
and Terrasync 2.4 software. GPS data were post-processed
using Pathﬁnder Ofﬁce 3.0 (Trimble Navigation Ltd, Sunny-
Drillers in Bangladesh do not typically use any material
to seal the outside of the PVC pipe (personal communi-
cation, Peter Ravenscroft) from rapid annular ﬂow or
‘short-circuiting’ of surface water or near-surface water to
screened interval depth. In developed countries, bentonite,
an expanding clay, will typically be used to seal wells. To
control for the possibility of annular ﬂow, 11 monitoring
wells with cement seals (MS) were installed in January
2008 throughout Site K. Ten of these had 1.5 m screens at
the same depth as a nearby private well, and one was
installed at some distance from all the 35 monitored private
wells (Figure 1) and was screened at a typical depth. In
addition six monitoring wells were installed within two
multilevel nests, previous to 2008 without cement seals
(M) (Table 1). These unsealed monitoring wells served as
intermediates between private (P) and sealed monitoring
(MS) wells to test for an effect of regular pumping only (as
opposed to pumping absence and seal presence combined)
on the frequency of E. coli detections in the well. The
depths of the 35 class P wells varied between 5.8 and
30.5 m (5th and 95th percentile were 6.1 and 15.2 m,
respectively) and the median depth was 7.6 m. Depths of
the 11 MS wells varied between 7.2 and 15.4 m and the
median depth was 7.7 m. All wells were monitored for
E. coli monthly for the period from April 2008 through
Continuous pumping from both P and MS tubewells has
been observed to dramatically decrease measured concen-
trations of E. coli over a 24-hour period (Knappett et al.
b), therefore prior to sampling all wells were purged
for a consistent three well-bore volumes. Duplicate 100 mL
water samples were taken from every well. Private tubewells
Locations of 35 unsealed private tubewells (triangles), six unsealed (circles) and 11 sealed monitoring wells (squares) within Char Para (Site K). The six unsealed monitoring wells
are contained within two multilevel piezometer nests. Image produced in Google Earth
. The inset country map is from www.mapresources.com. The scale bar in the bottom
left corner represents 200 m.
568 P. S. K. Knappett et al.
Unsealed tubewells lead to increased fecal contamination of drinking water Journal of Water and Health
were sampled using the existing iron hand pump whereas
monitoring wells were pumped using submersible electric
pumps (Typhoon, Groundwater Essentials, LLC) at ﬂow
rates ranging from 2 to 8 L/min. In between monitoring
wells, all tubing and electric pumps were ﬂushed with a
bleach and TWEEN solution diluted in water derived from
the well just sampled, followed by rinsing once with well
water and once with sodium thiosulfate as detailed in Knap-
pett et al.(b).
Two ‘snapshot’ sampling events were performed, once
during the dry season (March 16–18) and once during the
wet season (July 3–7) in 2009 to analyze for a broad spectrum
of FIBs and pathogens. Cultured E. coli was measured con-
currently, only during the wet season sampling event.
E. coli was quantiﬁed using the Most Probable Numbers
(MPN) based Colilert™ test kit (IDEXX Laboratories,
Inc.). Duplicate 100 mL groundwater samples were col-
lected in sterile containers and measurements were carried
out in a laboratory within 8 hours of sample collection.
MPN of E. coli were determined by combining the numbers
of discrete positive wells in both trays (Hurley & Roscoe
; Knappett et al. b).
For enumeration of fecal bacteria genomes, 4–8L of
groundwater was ﬁltered onto 0.22 μm nitrocellulose ﬁlters.
The ﬁlters were removed from the plastic housing, placed in
sterile Petri dishes, frozen and transported on dry ice to the
University of Tennessee. DNA was extracted and puriﬁed
from the ﬁlters using the FastDNA
SPIN for Soil Kit (MP
Biomedicals, LLC, Solon, Ohio) following the manufac-
turer’s protocols. DNA was measured using a nanodrop
and the extracts were diluted to 5–10 ng/μL of total DNA
to avoid inhibition, and this was further veriﬁed by
measuring the amount of a known plasmid spike added to
each sample for each PCR run.
Quantitative PCR was performed to detect E. coli and
Bacteroidales using the same assays and laboratory methods
as described in our previous study (Knappett et al. b).
The gene targets for the E. coli (herein referred to as mE.
coli) and Bacteroidales assays were the 23S rRNA gene
and the 16S rRNA gene, respectively (Bernhard & Field
; Scott et al. ; Layton et al. ; Noble et al.
; Kildare et al. ) and the primer and probe
sequences are provided in Table 2. Pathogen genes (eltA,
ipaH and adenovirus 40/41) were assayed in duplicate or tri-
plicate using the primers, probes and master mix types listed
in Table 2 and following standard qPCR protocols described
in previous studies (Knappett et al. b). Standards for the
quantitative PCR reactions were made from a relevant gene
fragment cloned into PCR4-TOPO cloning vector (Layton
et al. ). The method detection limit (MDL) was deter-
mined from the standard curve to be 20 gene copies per
qPCR reaction. Data were calculated only for samples in
which at least two PCR reactions had >1 gene copy and
were quantiﬁed as copies/μL nucleic acid extract. Gene
copies were adjusted to copies/100 mL for tubewell water
based on the fraction of the ﬁlter extracted, multiplied by
the volume of DNA extract and divided by the ﬁltered
sample volume. Due to differences in the volume of water
ﬁltered for each sample, the MDL varied somewhat with
each sample (Knappett et al. b) with a mean detection
limit of four copies/100 mL for the groundwater samples
(Ferguson et al. ).
Experimental design and statistical analyses
The three well types (P, M and MS) were compared for E. coli
prevalence using binned wet/dry season box plots and a
monthly time series comparison between P and MS wells.
A 6-month dry season was deﬁned here from November 15
through May 15 with the wet season being the other half of
the year. A total of 18 months were available (April 2008
through October 2009) for which all three classes of wells
were sampled each month (class P wells were also sampled
from January 2008 through April 2008). For statistical testing
on binned wet/dry season data a minimum of ﬁve sampling
events were required in each season for each well, causing
Classiﬁcation of tubewells at Site K
Well type (notation) Seal (y/n) Pumping frequency Count
Private (P) N Daily 35
Monitoring (M) N Monthly 6
Monitoring (MS) Y Monthly 11
569 P. S. K. Knappett et al.
the numbers of wells in each category to be reduced to 33, 6
and 11 for P, M and MS, respectively. Analysis of variance
(ANOVA) was performed three times with well class as the
‘treatment’ and E. coli frequency during year round, wet
and dry seasons as response variables, to determine differ-
ences between the classes of wells. Further, the non-
parametric Kruskal-Wallis test was performed on the ranks
to conﬁrm statistical differences between paired classes of
wells using the statistical software NCSS (version 07.1.14,
NCSS, LLC, Kaysville, Utah). E. coli prevalence in a given
well (across time) or during a given month (across space)
was accompanied by the approximation ±2[p (1 p)/n]
used to estimate 95% CIs for the proportion of wells p with
detectable E. coli where n is the total number of sampling
events or wells respectively (Gelman & Hill ).
RESULTS AND DISCUSSION
Monthly E. coli detection frequency in sealed and
E. coli prevalence in tubewells was observed to be substan-
tially higher in the monsoon than the dry season
(Figures 2(b) and 2(c)). An ANOVA on the E. coli preva-
lence data conﬁrmed that well class (P, M, MS)
signiﬁcantly impacted the frequency of E. coli detected in
a well, both year-round (Figure 2(a)) and during the wet
season (p < 0.05) (Figure 2(b)), but not during the dry
season (Figure 2(c)). Non-parametric Kruskal-Wallis tests
between pairs of well classes conﬁrmed that P wells were
more frequently contaminated than MS wells year-round
and during the wet season (p < 0.05) (Figures 2(a) and
2(b)). E. coli prevalence in M wells was intermediate and
not signiﬁcantly different from either P or MS wells. The
lesser E. coli prevalence in M wells than P wells suggests
that daily pumping from private tubewells plays a role in
fecal contamination. Other possible causes of more frequent
private tubewell contamination over monitoring wells (M
and MS) include bioﬁlm growth within the iron hand
pump (Ferguson et al. ) and the introduction of E. coli
into the well following pump priming of private wells (van
Geen et al. ).
Although annular sealing appeared to substantially
reduce E. coli detection frequency, the presence of an
intact cement platform had no impact on the microbial
drinking water quality from a private tubewell in either the
wet or dry seasons (Figures 2(e) and 2(f)). The WHO/
Quantitative PCR primer and probes used to target speciﬁc genes and organisms in tubewell water samples
Assay gene target (assay
type and annealing
temperature) Oligonucleotide sequences
E. coli 23S rRNA
GAG CCT GAA TCA GTG TGT GTG 3
Knappett et al.
ATT TTT GTG TAC GGG GCT GT 3
CGC CTT TCC AGA CGC TTC CAC
All Bacteroides 16S rRNA
Layton et al.
Shigella and entero-
(dysentery-type E. coli)
IpaH U1f- 5
et al.()IpaH L1r- 5
- CGGAATCCGGAGGTATTG C-3
Enterotoxigenic E. coli
eltA (heat labile toxin
LT) ( ﬂuorogenic
Persson et al.
Adenovirus 40/41 Hexon (ﬂuorogenic
- CAGCCTGGGGAACAAGTTCAG 3
Rajal et al.
All probes synthesized with FAM (ﬂuorescein) and black hole quencher 1 (BHQ1) from Biosearch Technologies.
570 P. S. K. Knappett et al.
UNICEF () well construction guidelines emphasize the
importance of an adequate drainage channel to allow the
spilled water to leave the well-head area; however, strictly
speaking this is not required for the classiﬁcation of a tube-
well as an ‘improved’ drinking water source (WHO/
UNICEF ). As of 2010, 80% of the rural population in
Bangladesh was recorded drinking from ‘improved’ water
sources (WHO/UNICEF ). In our exhaustive survey of
the 144 private tubewells at Site K in 2009, however, only
42% (61/144) of private tubewells had intact platforms,
and many of these did not have drainage channels. In a
related study combining 32 wells from Site K with 93
wells in Matlab upazilla, 50 km south of Site K, only 51%
(64/125) of all tubewells had intact platforms (van Geen
et al. ). In the present study, we did not differentiate
between intact platforms with and without good drainage
channels. The sample size in the present study for which
monthly E. coli measurements were available is small,
with 11 tubewells having good platforms and 19 that did
not have platforms or had broken platforms. Other studies,
however, have reported the insensitivity of platform quality
and presence for FIB detection frequency in private tube-
wells in Bangladesh (Luby et al. ; Leber et al. ;
van Geen et al. ). This ﬁnding has led some to conclude
that annular ﬂow is not important in Bangladesh where
wells are drilled and not dug (Luby et al. ). Based on
the information presented in this study, annular ﬂow
seems to degrade microbial water quality, but the presence
of an intact platform has little protective effect. This may
be due to the general absence of good drainage channels,
or the ﬂat terrain which leads to ponding around well-
heads, rendering intact platforms and long drainage chan-
nels irrelevant following large rainfall events.
It was hypothesized that E. coli detection frequency
would decrease with depth. Well depth was not found to
correlate to E. coli detection frequency for any class (P,
M, MS) of wells in this study for any month or binned
season. This is consistent with other studies measuring
Detection frequencies of E. coli in monthly monitored private (P), unsealed monitoring (M), and sealed monitoring (MS) wells from April 2008 through November 2009. For the
well type plots (a)–(c) the number of wells with at least 5 months of monthly data in each season were 33, six and 11 for P, M and MS respectively. There were a total of 12
possible wet season sampling events and 6 dry season months. In the platform presence plots (d)–(f), only private wells are presented here since no monitoring wells had
cement platforms. A reduced number of private wells (n ¼ 30) was available due to missing information.
571 P. S. K. Knappett et al.
FIB contamination risk factors in private tubewells in Ban-
gladesh (Luby et al. ; Leber et al. ; van Geen et al.
). One study showed consistent decreases in FIB con-
centration with depth in paired MS wells installed only
3 m apart vertically in highly contaminated aquifers in
the vicinity of latrine ponds (Knappett et al. ). It is
likely that sediment heterogeneity and well-speciﬁc pro-
cesses such as bioﬁlm growth (Ferguson et al. ) and
rapid ﬂow down unsealed annuli confound a simple
relationship between depth and E. coli detection frequency
in studies where comparatively shallow tubewells (<36 m)
are compared across village(s) (Leber et al. ; van
Geen et al. ).
Throughout the 18-month monitoring period MS wells
were typically less frequently contaminated on a month-to-
month basis (Figure 3(b)). During the early monsoon
season E. coli was just as prevalent in MS wells as P wells;
however, E. coli prevalence tended to decrease later in the
monsoon in MS wells but continued to increase in P wells
(Figure 3(b)). Two exceptions to this pattern were when
E. coli prevalence in MS wells exceeded P wells, following
major rainfall events in the 2008 wet season.
These ﬁndings can be explained the following way.
Sealed monitoring wells act to sample aquifer water that
becomes contaminated via vertical inﬁltration through the
vadose zone in a ‘ﬁrst ﬂush’ event when the water table is
still >4 m below the surface (Figure 3(a)). Since 90% of all
wells in this study have total depths that vary between 6
and 15 m, this ‘ﬁrst ﬂush’ water reaches the saturated water
table only 0.5–9.5 m above the screened interval (1.5 m).
As the water table rises the impetus for downward movement
of inﬁltrating rainwater through sediment decreases with the
thinning of the vadose zone. Similarly, E. coli prevalence
increases in P wells at the start of monsoonal rains, but in
contrast to the MS wells, as the water table rises unsealed
P wells become more frequently contaminated (Figure 3(b)).
This can be explained by vertical ﬂow along the annulus
of the private wells enhanced by regular pumping.
Dry and wet season snapshots of molecular FIBs and
During the wet season, snapshot sampling event concen-
trations of molecular E. coli and Bacteroidales in MS
Comparing culturable E. coli prevalence in wells with rainfall and water table levels. (a) Weekly precipitation (vertical grey bars) for Matlab located 50 km south of Site K (left-
axis). Manual groundwater levels are displayed at Site K (black line with grad symbols) from April 1 2008 through November 1 2009 (right-axis). Months assigned to the wet
season are indicated by boxes outlined by dashed lines. (b) Monthly proportion of private (P) (n ¼ 35) and sealed monitoring (MS) (n ¼ 11) wells testing positive for cultured E. coli
(left-axis). 75th percentile cultured E. coli concentrations (MPN/100 mL) for both P (dashed grey line) and MS (solid grey line) wells (right-axis).
572 P. S. K. Knappett et al.
wells were signiﬁcantly lower than in P wells as assessed by
the Kruskal-Wallis test (p < 0.05). Although FIB DNA con-
centrations were high during the dry season, no signiﬁcant
difference in concentration was observed between MS and
P wells (Figures 4(a) and 4(c)). This contrasts with the cul-
tured E. coli data, which shows all three classes of
tubewells have lower frequency of E. coli detected during
the dry season than the wet (Figures 2(b) and 2(c)). The
similar concentrations of molecular FIBs in MS and P
wells during the dry season suggests that short circuiting
is not active during the dry season, and rather these con-
centrations of FIB DNA represent background levels in
the aquifer sustained throughout much of the year.
This ﬁnding suggests that FIB DNA persistence in these
shallow anaerobic aquifers is longer than culturable E. coli.
This agrees with ﬁndings in Knappett et al.() where
mE. coli and Bacteroidales were detected well above detec-
tion limit in ﬁne sediments adjacent to latrine ponds in the
absence of culturable E. coli. Even in sediments containing
abundant culturable E. coli, Bacteroidales and E. coli
DNA was observed to be transported further laterally and
especially vertically (Knappett et al. ). Another study
found widespread high concentrations of FIB DNA in pri-
vate wells across several villages and seasons (Ferguson
et al. ). It is likely that the reduced metabolism and
size of starving fecal bacteria in oligotrophic aquifers
allows them to pass much further than readily culturable
E. coli. This was found in a study by Jansen et al.()
in sand columns with Pseudomonas ﬂuorescens where
the starved, coccoid bacteria were less efﬁciently removed
than metabolically active rod-shaped bacteria. This
phenomenon has important implications for understanding
the perceived threat of FIB and pathogen DNA in ground-
water. Traditionally culturing of E. coli has been relied
upon to indicate the relative health risk from drinking
water (WHO ). Fecal bacteria DNA, contained in
dead or starving cells, however, may be transported further
and persist longer in an aquifer than the more metaboli-
cally active culturable cells. Therefore, the relative health
risk represented by different concentrations of FIB and
Concentrations of mE. coli and Bacteroidales DNA in private (P) and sealed monitoring wells (MS) during dry season and wet season snap shot sampling events. Sample size for
each group is indicated in parentheses. Unsealed monitoring wells (M) were not included in this analysis due to low sample numbers (n < 5).
573 P. S. K. Knappett et al.
pathogen DNA occurring in groundwater (Ferguson et al.
) is currently unknown and needs to be assessed as it
has been for recreational water exposure (Wade et al.
One advantage of enumerating FIB DNA over culturing
E. coli is that it is present in high concentrations year round
at multiple ﬁeld sites (Figure 4)(Ferguson et al. ). When
concentrations of mE. coli and Bacteroidales from the dry
and wet season snapshot sampling events are plotted against
each other for 14 overlapping wells, Bacteroidales emerges
as more consistent year round, whereas mE. coli concen-
trations are uncorrelated in the same wells in different
seasons (Figure 5(a)). Furthermore, wet season Bacteroi-
dales were the only FIB (and the only season) found to be
somewhat predictive (R
¼ 0.33, p < 0.05) of annual E. coli
detection frequency (Figure 5(b)). Bacteroidales may be a
more seasonally unbiased estimate of year round suscepti-
bility of a groundwater well to fecal contamination than
either mE. coli or the sporadically detected cultured E.
coli. Sampling at one or two points in time, rather than fre-
quent monitoring over years, may save substantial labor.
Although correlated with each other during the wet
season snapshot, no combination of FIBs was found to be
predictive of pathogen presence/absence (Figures 6(a) and
6(b)), with 10/22 samples found to be positive for at least
one pathogen. There were several false negatives with
respect to the indicators, with pathogens being found in
water samples containing low concentrations of E. coli,
mE. coli and Bacteroidales. Annual E. coli detection fre-
quency did not predict the wet season presence of
pathogens (Figure 5(b)), either. In contrast, although only
2/20 dry season snapshot wells were positive for pathogens,
both of these wells were highly contaminated with mE. coli
and Bacteroidales (cultured E. coli was not measured simul-
taneously during the dry season snapshot) (Figure 6(c)). It is
noteworthy that pathogens were detected in approximately
equal numbers of both unsealed (P) and sealed (MS) wells,
reinforcing the idea that although well construction effects
lead to increased contamination of the tubewells them-
selves, fecal bacteria and pathogen DNA are clearly
inﬁltrating the broader aquifer beyond the near-well
environment (Knappett et al. ).
Inter-seasonal comparisons of FIB and pathogens detected in unsealed private
wells (P), unsealed monitoring wells (M) and sealed monitoring wells (MS). (a)
Measured FIB marker gene concentrations in wells from during the wet and
dry season snap shot sampling events (n ¼ 14). Only the equation for the line
of best ﬁt for Bacteroidales is displayed since the ﬁt was very poor for mE. coli .
(b) Wet season concentrations of Bacteroidales 16S genes and pathogen
presence/absence plotted against cultured E. coli detection frequency for all
months (n ¼ 22). Samples where no pathogens were detected are indicated by
black symbols. Samples positive for Shigella and Adenovirus are indicated by
red and blue, respectively.
574 P. S. K. Knappett et al.
Multiple differences, including annular sealing and usage
frequency, between private wells and monitoring wells led
to more frequent detections of E. coli in unsealed tubewells.
Clearly some simple well construction improvements can be
made that will lead to decreases in FIB detections and cases
of diarrheal disease obtained by drinking untreated ground-
water (Gundry et al. ). In a related study, hand pumps
themselves were found to harbor E. coli long after being
exposed to high levels of E. coli (Ferguson et al. ). This
effect could have accounted for some of the more frequent
contamination seen here in private wells. Amongst unsealed
private wells, however, platform presence/absence had no
impact on microbial water quality. This ﬁnding could possi-
bly be due to the lack of adequate drainage channels
accompanying intact platforms; however, this lack of sensi-
tivity to platform presence and quality has been reported
previously in Bangladesh (Luby et al. ; Leber et al.
; van Geen et al. ). Together these ﬁndings suggest
that annular sealing or another private well construction
factor may be much more important in determining
Comparison of FIB concentrations and pathogen presence/absence during wet ( n ¼ 22) and dry season (n ¼ 20) snapshot sampling events. (a) Comparison of synoptic
measurements of cultured E. coli (MPN/100 mL) and mE. coli (copies/100 mL) from the wet season snapshot sampling event. The culturable E. coli method detection limit (MDL)
was 0.5 MPN/100 mL. (b) Comparison of Bacteroidales and mE. coli concentrations during wet season snapshot. (c) Comparison of Bacteroidales and mE. coli during dry season
snapshot. Culturable E. coli was not measured synoptically during the dry season snap shot sampling event. Samples where no pathogens were detected are indicated by black
symbols. Samples positive for Shigella, EltA (ETEC E. coli) and Adenovirus are indicated by red, green and blue, respectively.
575 P. S. K. Knappett et al.
microbial drinking water quality than platforms and drain-
age channels in Bangladesh.
The molecular FIB Bacteroidales sampled in 22 wells
during the wet season was found to be predictive of year-
round fecal contamination in both unsealed and sealed
wells, as assessed by annual E. coli detection frequency
and dry season Bacteroidales. Pathogens were more fre-
quently detected in the wet season, but their presence was
uncorrelated to any FIB or well type. The high prevalence
of E. coli, and high concentrations of FIB DNA, in sealed
monitoring wells, especially during the early monsoon
season, indicate that fecal contamination is indeed inﬁltrat-
ing and spreading over broad volumes of the aquifer.
Inﬁltration pathways likely include both leaky tubewell
annuli and inﬁltration from ponds (Knappett et al. ).
Other factors potentially contributing to E. coli prevalence
in tubewells in Bangladesh include the size distribution
(Leber et al. ; Knappett et al. ) and mineralogy
(Ryan et al. ; Flynn et al. ) of the overlying sedi-
ment, pore-water ionic strength and chemical composition
(Fontes et al. ; Loveland et al. ), and the local spatial
density of contamination sources (Knappett et al. a; van
Geen et al. ). Multi-season FIB monitoring could be con-
ducted on both unsealed and sealed private wells with
similar usage frequency to isolate for an effect of well sealing
only apart from regular pumping and well-head effects (Fer-
guson et al. ). Further, private wells could be tested for
leaks in the PVC pipe to determine how prevalent this is
and whether leaking well pipes can explain the fecal con-
tamination patterns in the unsealed private wells.
This study was supported by grant 5 R01 TW008066 from
the NIH/FIC Ecology of Infectious Disease program.
Additional funding was provided by the Center for
Environmental Biotechnology, the Institute for a Secure
and Sustainable Environment at the University of
Tennessee and the Marie Curie Training Network
GOODWATER program. An anonymous reviewer greatly
contributed to the ﬁnal product. Thank you to
V. Escamilla for carrying out a detailed survey of Char
Para in June 2009; to K. Radloff for providing long-term
groundwater levels; and to Md. R. Huq for his assistance
in the ﬁeld.
Abbaszadegan, M., LeChevallier, M. & Gerba, C. P.
Occurrence of viruses in U.S. groundwaters. J. Am. Water
Works Assoc. 95, 107–120.
Bernhard, A. E. & Field, K. G. Identiﬁcation of nonpoint
sources of fecal pollution in coastal waters by using host-
speciﬁc 16S ribosomal DNA genetic markers from fecal
anaerobes. Appl. Environ. Microbiol. 66, 1587–1594.
Borchardt, M. A., Bradbury, K. R., Alexander Jr., E. C. , Kolberg, R.
J., Alexander, S. C., Archer, J. R., Braatz, L. A., Forest, B. M.,
Green, J. A. & Spencer, S. K. Norovirus outbreak caused
by a new septic system in a dolomite aquifer. Ground Water
De Novio, N. M., Saiers, J. E. & Ryan, J. N. Colloid
movement in unsaturated porous media: recent advances and
future directions. Vadose Zone J. 3, 338–351.
Embrey, S. S. & Runkle, D. L. Microbial Quality of the
Nation’s Ground-Water Resources, 1993–2004. National
Water-Quality Assessment Program Principal Aquifers. US
Geol. Survey Scientiﬁc Investigations Report 2006-5290. US
Geol. Survey, Reston, Virginia.
Escamilla, V., Knappett, P., Yunus, M., Streatﬁeld, P. K. & Emch,
M. Inﬂuence of latrine proximity and type on tubewell
water quality and diarrheal disease in Bangladesh. Ann.
Assoc. Am. Geogr. In Press.
Escamilla, V., Wagner, B., Yunus, M., Streatﬁeld, P. K., van Geen,
A. & Emch, M. Effect of deep tubewell use on childhood
diarrhoea in Bangladesh. Bull. WHO 89, 521–527.
Ferguson, A. S., Layton, A. C., Mailloux, B. J., Culligan, P. J.,
Williams, D. E., Smartt, A. E., Sayler, G. S., Feighery, J.,
McKay, L., Knappett, P. S. K., Alexandrova, E., Arbit, T.,
Emch, M., Escamilla, V., Ahmed, K. M., Alam, Md. J.,
Streatﬁeld, P. K., Yunus, Md. & van Geen, A.
Comparison of fecal indicators with pathogenic rotavirus in
groundwater. Sci. Total Environ. 431, 314–322.
Ferguson, A. S., Mailloux, B. J., Ahmed, K. M., van Geen, A.,
McKay, L. D. & Culligan, P. J. Hand-pumps as reservoirs
for microbial contamination of well water. J. Water Health 9,
Flynn, R., Hunkeler, D., Guerin, C., Burn, C., Rossi, P. & Aragno, M.
Geochemical inﬂuences on H40/1 Bacteriophage
inactivation in glacioﬂuvial sands. Environ. Geol. 45, 504–517.
Fontes, D. E., Mills, A. L., Hornberger, G. M. & Herman, J. S.
Physical and chemical factors inﬂuencing transport of
microorganisms through porous media. Appl. Environ.
Microbiol. 57, 2473–2481.
Gelman, A. & Hill, J. Data Analysis Using Regression and
Multilevel/Hierarchical Models. Cambridge University Press,
576 P. S. K. Knappett et al.
Godfrey, S., Timo, F. & Smith, M. Microbiological risk
assessment and management of shallow groundwater sources
in Lichinga, Mozambique. Water Environ. J. 20, 194–202.
Goodbred, S. L. & Kuehl, S. A. Floodplain processes in the
Bengal Basin and the storage of Ganges-Brahmaputra river
sediment: an accretion study using 137Cs and 210Pb
geochronology. Sediment. Geol. 121, 239–258.
Gundry, S., Wright, J. & Conroy, R. A systematic review of
the health outcomes related to household water quality in
developing countries. J. Water Health 2,1–13.
Hurley, M. A. & Roscoe, M. E. Automated statistical-analysis
of microbial enumeration by dilution series. J. Appl.
Bacteriol. 55, 159–164.
Jansen, S., Vereecken, H. & Klumpp, E. On the role of
metabolic activity on the transport and deposition of
Pseudomonas ﬂuorescens in saturated porous media. Water
Res. 44, 1288–1296.
Johnson, T. B., McKay, L. D., Layton, A. C., Jones, S. W., Johnson,
G. C., Cashdollar, J. L., Dahling, D. R., Villegas, L. F., Fout,
G. S., Williams, D. E. & Sayler, G. Viruses and bacteria
in karst and fractured rock aquifers in East Tennessee, USA.
Ground Water 49,98–110.
Kildare, B. J., Leutenegger, C. M., McSwain, B. S., Bambic, D. G.,
Rajal, V. B. & Wuertz, S. 16S rRNA-based assays for
quantitative detection of universal, human-, cow- and dog-
speciﬁc fecal Bacteroidales: a Bayesian approach. Water Res.
Knappett, P. S. K., Escamilla, V., Layton, A., McKay, L. D., Emch,
M., Williams, D. E., Huq, Md. R., Alam, Md. J., Farhana, L.,
Mailloux, B. J., Ferguson, A., Sayler, G. S., Ahmed, K. M. &
van Geen, A. a The impact of sanitation on fecal bacteria
and pathogens in village ponds in Bangladesh. Sci. Total
Environ. 409, 3174–3182.
Knappett, P. S. K., Layton, A., McKay, L. D., Williams, D.,
Mailloux, B. J., Huq, Md. R., Alam, Md. J., Ahmed, K. M.,
Akita, Y., Serre, M. L., Sayler, G. S. & van Geen, A. b
Efﬁcacy of hollow-ﬁber ultraﬁltration for microbial sampling
. Ground Water 49,53–65.
Knappett, P. S. K., McKay, L. D., Layton, A., Williams, D. E.,
Alam, Md. J., Huq, Md. R., Mey, J., Feighery, J. E., Culligan, P.
J., Mailloux, B. J., Zhuang, J., Escamilla, V., Emch, M.,
Perfect, E., Sayler, G. S., Ahmed, K. M. & van Geen, A.
Implications of fecal bacteria input from latrine-polluted
ponds for wells in sandy aquifers. Environ. Sci. Technol. 46,
Kozuskanich, J., Novakowski, K. S. & Anderson, B. C. Fecal
indicator bacteria variability in samples pumped from
monitoring wells. Ground Water 49,43–52.
Layton, A., McKay, L. D., Williams, D., Garrett, V., Gentry, R. &
Sayler, G. Development of Bacteroides 16S rRNA gene
TaqMan-based real-time PCR assays for estimation of total,
human, and bovine fecal pollution in water. Appl. Environ.
Microbiol. 72, 4214–4224.
Leber, J., Rahman, M. M., Ahmed, K. M., Mailloux, B. J. & van
Geen, A. Contrasting inﬂuence of geology on E. coli
and arsenic in aquifers of Bangladesh. Ground Water 49,
Loveland, J. P., Ryan, J. N., Amy, G. L. & Harvey, R. W. The
reversibility of virus attachment to mineral surfaces. Colloid
Surf. A 107, 205–221.
Luby, S. P., Gupta, S. K., Sheikh, M. A., Johnston, R. B., Ram, P. K.
& Islam, M. S. Tubewell water quality and predictors of
contamination in three ﬂood-prone areas in Bangladesh. J.
Appl. Microbiol. 105, 1002–1008.
Noble, R. T., Grifﬁth, J. F., Blackwood, A. D., Fuhrman, J. A.,
Gregory, J. B., Hernandez, X., Liang, X., Bera, A. A. &
Schiff, K. Multitiered approach using quantitative
PCR to track sources of fecal pollution affecting Santa
Monica bay, California. Appl. Environ. Microbiol. 72,
Persson, S., Olsen, K. E. P., Scheutz, F., Krogfelt, K. A. &
Gerner-Smidt, P. A method for fast and simple
detection of major diarrhoeagenic Escherichia coli
routine diagnostic laboratory. Clin. Microbiol. Infect. 13,
Pronk, M., Goldscheider, N. & Zopﬁ,J. Particle-size
distribution as indicator for fecal bacteria contamination of
drinking water from karst springs. Environ. Sci. Technol. 41,
Rajal, V. B., McSwain, B. S., Thompson, D. E., Leutenegger, C. M.,
Kildare, B. J. & Wuertz, S. Validation of hollow ﬁber
ultraﬁltration and real-time PCR using bacteriophage PP7 as
surrogate for the quantiﬁcation of viruses from water
samples. Water Res. 41, 1411–1422.
Rudolph, D. L., Barry, D. A. J. & Goss, M. J. Contamination
in Ontario farmstead domestic wells and its association with
agriculture: 2. Results from multilevel monitoring well
installations. J. Contam. Hydrol. 32 , 295–311.
Ryan, J. N., Elimelech, M., Ard, R. A., Harvey, R. W. & Johnson, P.
R. Bacteriophage PRD1 and silica colloid transport and
recovery in and iron oxide-coated sand aquifer. Environ. Sci.
Scott, T. M., Rose, J. B., Jenkins, T. M., Farrah, S. R. & Lukasik, J.
Microbial source tracking: current methodology and
future directions. Appl. Environ. Microbiol. 68, 5796–5803.
Streatﬁeld, K., Persson, L. A., Chowdhury, H. R. & Saha, K. K.
Disease Patterns in Bangladesh: Present and Future
Needs. International Centre for Diarrhoeal Disease Research,
Bangladesh, Dhaka, Bangladesh.
Wade, T. J., Calderon, R. L., Brenner, K. P., Sams, E., Beach, M.,
Haugland, R., Wymer, L. & Dufour, A. P. High
sensitivity of children to swimming-associated
gastrointestinal illness. Epidemiology 19, 375–383.
Weinman, B., Goodbred Jr., S. L. , Zheng, Y., Aziz, Z., Steckler,
M., van Geen, A., Singhvi, A. K. & Nagar, Y. C.
Contributions of ﬂoodplain stratigraphy and evolution to the
spatial patterns of groundwater arsenic in Araihazar,
Bangladesh. Bull. Geo. Soc. Am.
World Health Organization Global Water Supply and
Sanitation Assessment 2000 Report. World Health
577 P. S. K. Knappett et al.
Organization, Geneva, Switzerland. Available from: www.
World Health Organization Guidelines for Drinking-water
Quality 3rd Ed. Incorporating the First and Second Addenda
Volume 1, Recommendations. World Health Organization,
Geneva, Switzerland. Available from: www.who.int/
World Health Organization & United Nations Children’s Fund
Joint Monitoring Programme World Health
Organization & United Nations Children’s Fund, Geneva,
Switzerland. Available from: www.wssinfo.org/data-
Wu, J., van Geen, A., Ahmed, K. M., Akita, Y., Alam, Md. J.,
Culligan, P. J., Escamilla, V., Feighery, J., Ferguson, A. S.,
Knappett, P., Mailloux, B. J., McKay, L. D., Serre, M. L.,
Streatﬁeld, P. K., Yunus, M. & Emch, M. Increase in
diarrheal disease associated with arsenic mitigation in
Bangladesh. PLoS One 6, e29593.
van Geen, A., Ahmed, K. M., Akita, Y., Alam, M. J., Culligan, P. J.,
Emch, M., Escamilla, V., Feighery, J., Ferguson, A. S.,
Knappett, P., Layton, A. C., Mailloux, B. J., McKay, L. D.,
Mey, J. L., Serre, M. L., Streatﬁeld, P. K., Wu, J. & Yunus, Md.
Fecal contamination of shallow tubewells in Bangladesh
inversely related to arsenic. Environ. Sci. Technol. 45,
van Geen, A., Zheng, Y., Versteeg, R., Stute, M., Horneman, A.,
Dhar, R., Steckler, M., Gelman, A., Small, C., Ahsan, H.,
Graziano, J. H., Hussain, I. & Ahmed, K. M. Spatial
variability of arsenic in 6000 tubewells in a 25 km
Bangladesh. Water Resour. Res. 39, 1140.
von Seidlein, L., Kim, D. R., Ali, M., Lee, H., Wang, X., Thiem, V.
D., Canh, D. G., Chaicumpa, W., Agtini, M. D., Hossain, A.,
Bhutta, Z. A., Mason, C., Sethabutr, O., Talukder, K., Nair, G.
B., Deen, J. L., Kotloff, K. & Clemens, J. A multicentre
study of Shigella diarrhoea in six Asian countries: disease
burden, clinical manifestations, and microbiology. PLoS Med
First received 5 June 2012; accepted in revised form 9 September 2012. Available online 25 October 2012
578 P. S. K. Knappett et al.