Complex interactions between climate change, sanitation, and groundwater quality: a case study from Ramotswa, Botswana

Abstract

Groundwater quantity and quality may be affected by climate change through intricate direct and indirect mechanisms. At the same time, population growth and rapid urbanization have made groundwater an increasingly important source of water for multiple uses around the world, including southern Africa. The present study investigates the coupled human and natural system (CHANS) linking climate, sanitation, and groundwater quality in Ramotswa, a rapidly growing peri-urban area in the semi-arid southeastern Botswana, which relies on the transboundary Ramotswa aquifer for water supply. Analysis of long-term rainfall records indicated that droughts like the one in 2013–2016 are increasing in likelihood in the area due to climate change. Key informant interviews showed that due to the drought, people increasingly used pit latrines rather than flush toilets. Nitrate, fecal coliforms, and caffeine analyses of Ramotswa groundwater revealed that human waste leaching from pit latrines is the likely source of nitrate pollution. The results in conjunction indicate critical indirect linkages between climate change, sanitation, groundwater quality, and water security in the area. Improved sanitation, groundwater protection and remediation, and local water treatment would enhance reliable access to water, de-couple the community from reliance on surface water and associated water shortage risks, and help prevent transboundary tension over the shared aquifer.

Full text

REPORT
Complex interactions between climate change, sanitation,
and groundwater quality: a case study from Ramotswa, Botswana
Bonnie M. McGill
1,2
&Yvan Altchenko
3,4,5
&Stephen K. Hamilton
1,6
&Piet K. Kenabatho
7
&Steven R. Sylvester
8
&
Karen G. Villholth
3
Received: 23 March 2018 /Accepted: 10 November 2018
#The Author(s) 2019
Abstract
Groundwater quantity and quality may be affected by climate change through intricate direct and indirect mechanisms. At the
same time, population growth and rapid urbanization have made groundwater an increasingly important source of water for
multiple uses around the world, including southern Africa. The present study investigates the coupled human and natural system
(CHANS) linking climate, sanitation, and groundwater quality in Ramotswa, a rapidly growing peri-urban area in the semi-arid
southeastern Botswana, which relies on the transboundary Ramotswa aquifer for water supply. Analysis of long-term rainfall
records indicated that droughts like the one in 2013–2016 are increasing in likelihood in the area due to climate change. Key
informant interviews showed that due to the drought, people increasingly used pit latrines rather than flush toilets. Nitrate, fecal
coliforms, and caffeine analyses of Ramotswa groundwater revealed that human waste leaching from pit latrines is the likely
source of nitrate pollution. The results in conjunction indicate critical indirect linkages between climate change, sanitation,
groundwater quality, and water security in the area. Improved sanitation, groundwater protection and remediation, and local
water treatment would enhance reliable access to water, de-couple the community from reliance on surface water and associated
water shortage risks, and help prevent transboundary tension over the shared aquifer.
Keywords Climate change .Nitrate .Socioecology .Botswana .Sub-Saharan Africa
Introduction
Groundwater supports about 75% of the Sub-Saharan African
(SSA) population as well as industry and some crop irrigation
(Calow and MacDonald 2009; Villholth 2013; Niang et al.
2014), and water insecurity across SSA results from a com-
plex interplay of natural and social systems (Howard and
Bartram 2010). Groundwater and surface-water quality and
quantity are affected by changes in climate, land use, demo-
graphics and economic activity (Jiménez Cisneros et al.
2014); therefore, ensuring water security in SSA under a
changing climate requires an understanding of these
interacting effects on groundwater (Cronin et al. 2007;
Taylor et al. 2013; Famiglietti 2014; Niang et al. 2014).
Rapid urbanization in SSA is another driver of change in
groundwater and frequently means that infrastructure, including
Electronic supplementary material The online version of this article
(https://doi.org/10.1007/s10040-018-1901-4) contains supplementary
material, which is available to authorized users.
*Bonnie M. McGill
bonniemcgill@gmail.com
1
Michigan State University Department of Integrative Biology &
Kellogg Biological Station, Hickory Corners, MI, USA
2
Present address: Kansas Biological Survey, University of Kansas,
Lawrence, KS, USA
3
International Water Management Institute Southern Africa,
Pretoria, South Africa
4
Present address: Ministère de l’Agriculture, de l’Agroalimentaire et
de la Forêt, Paris, France
5
Present address: Laboratoire METIS, UMR 7619 UPMC/CNRS,
Université Pierre et Marie Curie, Paris, France
6
Cary Institute of Ecosystem Studies, Millbrook, NY, USA
7
University of Botswana Department of Environmental Science,
Gaborone, Botswana
8
Washington State University Vancouver Department of Molecular
Biosciences, Washington, USA
Hydrogeology Journal
https://doi.org/10.1007/s10040-018-1901-4

sanitation, lags behind the needs of the growing population. Pit
latrines are the most common means of human waste disposal,
often in areas where communities also rely on groundwater for
drinking (Cronin et al. 2007;Lapworthetal.2017). Unlined pit
latrines leach to the groundwater, threatening groundwater qual-
ity and human health with pollutants like nitrate (NO
3
−
)and
pathogens (Cronin et al. 2007; Graham and Polizzotto 2013).
In studies of pit latrine impacts on groundwater quality, NO
3
−
is the most commonly detected pollutant (Graham and Polizzotto
2013). Though safe levels of NO
3
−
for humans in drinking water
are still debated (Schullehner et al. 2018), it is clear that this is a
critical and widespread pollutant.
Elevated concentrations of NO
3
−
have been found in
arid environments, including the Kalahari Desert, where
it is believed to be the product of the decomposition of
ancient vegetation deposits (Heaton et al. 1983;Stadler
et al. 2012; Stone and Edmunds 2014). Stadler et al.
(2012) differentiated between naturally occurring and an-
thropogenic NO
3
−
in groundwater in the Kalahari using
stable isotopes: NO
3
−
in the deeper groundwater was
older in age and tended to be naturally occurring, while
NO
3
−
in the shallower, younger groundwater tended to
come from human and cattle sources. In a review of stud-
ies of NO
3
−
pollution in groundwater across southern
Africa,themainsourceofNO
3
−
pollution, despite natu-
rally occurring NO
3
−
, was anthropogenic activities, in-
cluding pit latrines and livestock feedlots (Tredoux and
Tal ma 2006). Tillage and associated nitrification of organ-
ic matter can be a minor source of NO
3
−
leaching. The
agricultural soils in inland southern Africa tend to be low
in N, and where inorganic N fertilizer is used, it is typi-
cally applied at low rates (Tredoux and Talma 2006),
though fertilizer application rates may be increasing.
In order to identify health risks associated with ground-
water quality and identify remediation options, especially
for NO
3
−
, in these settings it is important to understand
the contamination source. However, discerning between
human and livestock waste contamination can be a chal-
lenge: humans and livestock have similar NO
3
−
stable
isotope signatures and fecal indicator bacteria (Ashbolt
et al. 2001; Fenech et al. 2012). In recent years, a suite
of chemicals called emerging organic contaminants
(EOCs), including pharmaceuticals and caffeine (1,3,7-
trimethylxanthine), have been used as tracers of human
contamination. These substances tend to persist in the
environment, and detection limits continue to decrease
(Fenech et al. 2012;Lapworthetal.2012). Caffeine’s
metabolite paraxanthine (1,7-dimethylxanthine) has also
been used, but not as widely. Caffeine and/or
paraxanthine have been measured in groundwater and sur-
face water in Europe and the US (Buerge et al. 2003;
Glassmeyer et al. 2005;Godfreyetal.2007; Hillebrand
et al. 2012; Reh et al. 2013;Stuartetal.2014; Hillebrand
et al. 2015) and more recently in surface waters, urban
groundwater, and roof-harvested rain water in South
Africa (Matongo et al. 2015;Sorensenetal.2015;Waso
et al. 2016; Wanda et al. 2017; Rodríguez-Gil et al. 2018).
Denitrification is a naturally occurring microbial process
that removes NO
3
−
from groundwater. In low oxygen condi-
tions, denitrifying microbes reduce NO
3
−
into nitrous oxide
(N
2
O) and di-nitrogen (N
2
) gases that either escape to the at-
mosphere or, in a closed system like an aquifer, remain in the
water in dissolved form (Seitzinger et al. 2006; Robertson and
Groffman 2015). Heterotrophic denitrification involves several
subreactions (one of which produces N
2
O), such that the com-
plete denitrification process can be written as (Schlesinger and
Bernhardt 2013):
5CH2Oþ4Hþþ4NO−
3→2N2þ5CO2þ7H2Oð1Þ
Given the aforementioned stoichiometry, denitrification re-
quires dissolved organic carbon (DOC), the electron donor,
and NO
3
−
, the electron acceptor, in a ~1:1 molar ratio. Where
naturally occurring concentrations of DOC in the groundwater
limit denitrification, bioremediation technologies inject DOC
into the aquifer to enhance denitrification (USEPA 2013).
Scientists have warned that direct anthropogenic impacts on
groundwater quality could be more consequential than any di-
rect impacts of climate change on groundwater quality (Taylor
et al. 2009; Calow et al. 2010;FergusonandGleeson2012),
although some of these anthropogenic effects may be the indi-
rect result of climate change, as for example where climate
change alters human behavior that in turn has an impact on
groundwater. However, few studies have looked at the indirect
effects of climate change on sanitation and nitrate pollution of
groundwater (Howard et al. 2016). Climate change is happen-
ing in southern Africa. Over the last 50–100 years, most of
southern Africa’s mean, maximum, and minimum temperatures
have risen, especially in the last two decades (Niang et al.
2014). In this century, mean temperatures in southern Africa
are projected to increase faster than the global average, espe-
cially in semi-arid areas like Botswana. Between 1950 and
2000, rainfall in Botswana and neighboring countries has been
declining (Niang et al. 2014). Climate models project decreas-
ing annual mean rainfall for Botswana and Namibia, with drier
and shorter wet seasons (Niang et al. 2014); thus, there is a need
for understanding how climate change affects human behaviors
that can affect groundwater quality.
The study system, the Ramotswa aquifer and community,
is a coupled human and natural system (CHANS; Liu et al.
2007), meaning there are strong interactions between the bio-
physical and social systems, where changes in one can drive
changes in the other. An interdisciplinary CHANS approach
was used to investigate these interactions, i.e., how climate
affects water supply infrastructure and sanitation access in
Ramotswa, and how those changes impact the quality of the
Hydrogeol J

groundwater (Fig. 1). CHANS methods offer a powerful tem-
plate for investigating feedbacks between biophysical and
social systems like those previously discussed. In fact,
Sivapalan et al. (2012) argued that it is impossible to manage
or predict water availability without accounting for the feed-
backs between water and human systems, because human
activity has such pervasive effects on the modern water cycle.
The CHANS literature demonstrates that in order to charac-
terize today’s most pressing socio-ecological issues, interdis-
ciplinary research approaches are required to understand their
complexities and develop realistic solutions (Millenium
Ecosystem Assessment 2005; Liu et al. 2007; Collins et al.
2011; Stevenson 2011).
In this paper climate data analyses, sociology, and biogeo-
chemistry are integrated to capture the socio-ecological inter-
actions potentially driving groundwater pollution in
Ramotswa. The overarching research question is: BDoes cli-
mate change impact groundwater quality in Ramotswa, and if
so, how?^Specifically, the following hypotheses are tested
(Fig. 1):
H
1
: Climate change-induced droughts are becoming
more frequent.
H
2
: Water shortages affect sanitation behavior, increasing
pit latrine use.
H
3
: Nitrate pollution primarily originates from human
waste contamination.
H
4
: Enhanced denitrification via in situ bioremediation of
the Ramotswa aquifer has the potential to reduce nitrate
contamination.
This work represents several innovations: the application
of a CHANS approach to a groundwater system potentially
impacted by climate change, the measurements of both nitrate
and caffeine in groundwater in Africa (cf. Sorensen et al.
2015; Rodríguez-Gil et al. 2018), and a preliminary assess-
ment of the potential for in situ bioremediation (ISB) of nitrate
pollution in the Ramotswa Aquifer.
Study area
Biophysical characteristics
The peri-urban town of Ramotswa (24.88°S × 25.87°E) is
semi-arid and subtropical, about 25 km
2
in size and 20 km
south of Botswana’scapital,Gaborone(Fig.2). The town’s
eastern boundary is the ephemeral Notwane (Ngotwane)
River, which flows northward to the Limpopo River, and
serves as the boundary with South Africa in this area.
The rainfall is highly seasonal: the wet season (summer)
lasts from October to March, and the dry season (winter) is
from April to September. Therefore, the water year is from 1
October through 30 September. The mean annual (water year)
rainfall in Ramotswa is 454 mm (1980–2016), 89% of which
falls in the wet season (data from Botswana Dept. of
Meteorological Services). Dry season average rainfall is 47
and 64 mm (1980–2016) in Ramotswa and Gaborone, respec-
tively. Average annual evaporation from two nearby reservoirs
in South Africa is 2,120 mm (Altchenko et al. 2016), nearly
five times the average annual rainfall.
The geology and hydrogeology of the Ramotswa area are
described in Altchenko et al. (2017). Briefly, the outcrop area
of the transboundary Ramotswa aquifer is 453 km
2
,spanning
both Botswana and South Africa—Fig. S1 of the electronic
supplementary material (ESM). The aquifer is considered un-
confined to semi-confined and consists of the Ramotswa
Dolomite formation. In an aerial electromagnetic (AEM) sur-
vey, the Ramotswa maximum aquifer thickness in the
Ramotswa town area was at least 340 m and on average
150 m thick (Altchenko et al. 2017). The AEM survey con-
firmed the complexity of the Ramotswa aquifer’s hydrogeol-
ogy due to karstification, faults, and deep vertical dikes
(Altchenko et al. 2017). The mean depth to the water table
as measured inthe present study was 13.7 m below the ground
surface (Table S1 of the ESM), while in the larger Ramotswa
aquifer area and over a longer time period it was 24 m
(Altchenko et al. 2017). The general groundwater flow
Fig. 1 Conceptual diagram of
Ramotswa groundwater coupled
human and natural system
(CHANS) illustrating how
climate change could affect
Ramotswa drinking water
quantity and groundwater quality.
H
1
–H
4
in red represent the four
hypotheses explained at the end
of the ‘Introduction’
Hydrogeol J

direction is to the northeast (Fig. S1 of the ESM). The
Ramotswa aquifer average specific capacity is 2.7 L s
−1
m
−1
,
average transmissivity is 1,170 m
2
day
−1
, and the storage co-
efficient is 5.7 × 10
−2
(Staudt 2003; Moehadu 2014).
Production borehole yields range from 15 to 150 m
3
h
−1
,in-
dicating relatively high yields compared to other aquifers in
the area. Diffuse recharge is low (<20 mm year
−1
)(Gieske
1992; Post et al. 2012; Altchenko et al. 2017), and active
recharge where there is little overburden or soil cover and
direct infiltration through dolomite outcrops occurs in the
Ramotswa area (Altchenko et al. 2017; Gieske 1992). The
latter is critical, as the town of Ramotswa sits on an outcrop
of the Ramotswa Dolomite (Fig. S1 of the ESM).
Water supply and treatment infrastructure
The Ramotswa wellfield is about 34 km
2
(Fig. 2) and consists
of 14 production boreholes (BHs) managed by the Water
Utilities Corporation (WUC), a parastatal organization, and
about 45 monitoring BHs, managed by the Botswana Dept.
of Water Affairs (DWA). The BHs are not located in a true
Bwellfield^as they are scattered throughout town and the
surrounding area. A series of events explains the current wa-
ter supply situation in Ramotswa. In the early 1980s produc-
tion BHs were drilled in Ramotswa to supply drinking water
to Gaborone and Ramotswa. When the Gaborone reservoir
was completed in 1984, the Ramotswa groundwater was no
longer piped to Gaborone and was only used to supply
Ramotswa. In 1996, the Ramotswa wellfield was abandoned
because NO
3
−
concentrations exceeded the national drinking
water standard, 50 mg NO
3
−
L
−1
or 11 mg
NO
3
−
−NL
−1
(Walmsley and Patel 2011), and because
Ramotswa lacked the capacity to treat the water (Moehadu
2014). Water was therefore piped in from Gaborone (Fig. 3).
A multi-year drought from 2013 to 2016 severely lowered
the Gaborone reservoir level to <5% capacity from Dec 2014
to Mar 2016 (Fig. S2 of the ESM). Below 5% capacity, the
reservoir fails to release water. In 2014, the Ramotswa
wellfield was reopened as an emergency source of water,
and the BHs continued to supply water to Ramotswa and
the South East District until Tropical Cyclone Dineo struck
Botswana in Feb 2017 and filled the Gaborone reservoir (Fig.
S2 of the ESM).
Because NO
3
−
concentrations in groundwater from the
Ramotswa well field are generally above the drinking water
standard (50 mg L
−1
NO
3
−
) and water treatment is unavail-
able, the drinking water supplied in Ramotswa is a blend of
groundwater from the Ramotswa wellfield and surface wa-
ter supplied by pipes from the Gaborone Dam, supplied by
the WUC’s Gaborone Water Works (GWW; Moehadu
2014;Fig.3). The WUC manages all Botswana water sup-
ply and sewage. The volumes of dilution water from GWW
and produced from the Ramotswa wellfield over time were
unavailable. Gaborone relies on supply from the Gaborone
reservoir, the North-South carrier, the Molatedi reservoir in
South Africa and other sources (Fig. 3). When Gaborone’s
water supply is low, the water supply to Ramotswa and
other towns Bdownstream^(south) of the GWW is period-
ically restricted (red Bx^in Fig. 3)—defined here as shut-
ting off the water supply for a period of hours or days.
Fig. 2 Maps of aSouthern Africa, bBotswana’s South East District,
where black triangles represent the three weather stations used in the
present study, and cRamotswa borehole locations. Boreholes are
labeled and circle size indicates number of times sampled in the present
study. Color indicates whether the borehole was sampled in 2001 and this
study (green) or only this study (pink). Arrow shows general groundwater
flow direction (see also Fig. S1 of the ESM). Yellow triangle is the piped
sewage network ponds. Shaded area is the urbanized area
Hydrogeol J

Human dimensions of water supply and sanitation
of Ramotswa town
Botswana’s population has rapidly shifted from 9% urban in
1970 to 64% urban in 2011 (Statistics Botswana 2015b).
Between 2001 and 2011, the populations of Gaborone and
Ramotswa increased by 25 and 50%, respectively (Statistics
Botswana 2015c;b), increasing pressure on water supply
and sanitation infrastructure (Table 1;Fig.1). Based on the
most recent census in 2011, the population of Ramotswa in
2017 was projected at 37,500 people with a population den-
sity of 1,500 persons km
−2
(Statistics Botswana 2015c).
The poverty rate has been declining over time in
Botswana, and in 2010 the South East District poverty rate
was 13.4% compared to 6.1% in Gaborone (Statistics
Botswana 2013).
The annual mortality rate in Botswana for infants is 16
deaths per 1,000 and for children under 5 years it is 27 deaths
per 1,000 (Statistics Botswana 2017). Of those deaths, 6.4%
of infant deaths and 8% of deaths of children under 5 years are
due to Bdiarrhea and gastroenteritis of presumed infectious
origin^(Statistics Botswana 2017), possibly related to con-
taminated water.
As of 2011, more than one third of Ramotswa households
have potable water piped indoors, while about half access
potable water using an outdoor pipe (Table 1; Statistics
Botswana 2015c). Thirty eight percent of the Ramotswa pop-
ulation had access to a flush toilet as of 2011 (Statistics
Botswana 2015c), the majority of which were assumed to be
connected to the piped sewage network, while some small
fraction used septic tanks (Table 1; Staudt 2003). Fifty six
percent of the Ramotswa population relies solely on tradition-
al pit latrines or ventilated improved pit latrines (VIP)
(Statistics Botswana 2015c), which translates into an estimat-
ed 3,900 pit latrines in the town in 2011. It is unknown how
many pit latrines are lined vs. unlined.
Fig. 3 Wa te r su pp ly s ch em e fo r
the South East District. The red
arrow illustrates the coupling of
water quality to water quantity
and the grey-dashed arrow indi-
cates how a Ramotswa water
treatment facility could de-couple
Ramotswa from the Gaborone
water supply, which is vulnerable
to droughts. Modified from
Altchenko et al. (2016). Not to
scale.
Hydrogeol J

Materials and methods
Changes in seasonal rainfall amount and variability
To assess whether the climate in the Ramotswa area has changed
over time, daily rainfall data were combined from three weather
stations in the broader area (Fig. 2): Ramotswa (24.88°S ×
25.87°E, record begins in 1980), Gaborone (20 km north of
Ramotswa; 24.70°S × 25.90°E, record begins in 1926), and
Lobatse (40 km south of Ramotswa; 25.25°S × 25.65°E, record
begins in 1922). These data were provided by the Botswana
Dept. of Meteorological Services. Combining the Ramotswa data
with the stations in Lobatse and Gaborone allows for analysis of a
much longer period (95 versus 37 years) and covers a larger area
that includes several reservoirs that supply water to Gaborone
(Fig. 3). Due to the strong seasonal pattern in rainfall, wet season
(October–March) and dry season (April–September) rainfall to-
tals (referred to hereafter as Bseasonal^) were analyzed separately.
Multiple linear regression analysis was used to test for trends
in five rainfall metrics over time. First, seasonal totals were
compared over time to see if annual (water year) volume of
rain (mm) is changing. Second, number of days with rain (here-
after Brainy days^) were counted per season and compared over
time. Third, rainfall intensity per season (mm day
−1
) was cal-
culated using Eq. (2)fromPryoretal.(2009):
Seasonyear intensity ¼total rain
ndays with rain ð2Þ
Fourth, climate change can have a greater effect on rainfall
extremes and variability than on averages, so variability was
compared over time using the coefficient of variation (CV =
standard deviation/mean) of monthly rainfall totals per season.
Fifth, extreme rainfall seasons were grouped using the upper
and lower 10th percentile of seasonal rainfall totals to see if
the magnitude of the wettest and driest seasons was changing
over time (Pryor et al. 2009;NCEI2017).
Documenting the impact of climate change
on sanitation practices
To document the human behavioral aspects of this study, a
small number of key informants were interviewed. The ques-
tions were developed to gather general information about san-
itation, including undocumented pit latrine details (history, lo-
cation, types), and how water restrictions affect sanitation prac-
tices at the community level. The questions (section S1 of the
ESM) were approved prior to the interviews by the Michigan
State University Institutional Review Board (IRB# ×16-325e).
The most relevant interview question for testing BH
2
: Water
shortages affect sanitation behavior, increasing pit latrine use^
was somewhat straightforward: BHow have water restrictions
affected sanitation access in Ramotswa? (access to flush toi-
lets, washing hands, etc.)?^(sectionS1oftheESM).
Anecdotal belief among researchers and community leaders
was that pit latrine use increased during water shortages.
Given the time limitations on the research, it was not possible
to fully explore the experiences of individual community
members using pit latrines. Rather, the focus of this inquiry
was to understand the extent of pit latrine use and how com-
munity sanitation is impacted by water shortages.
Three informants were chosen because of their access to
information, representation, and/or familiarity with a large group
of socio-economically diverse people in Ramotswa, and diverse
roles in community leadership and resource management in
order to expediently capture broad perspectives. Although peo-
ple in leadership positions can potentially hold an elite bias, it is
common practice to use community leaders as key informants
for this type of broad, community-level investigation.
The key informants were: (1) a leader in Ramotswa tribal
governance, which is independent from state government and
prioritizes the well-being of Ramotswa citizens; (2) a WUC
representative of the head office in Gaborone, who schedules
water shut offs; and (3) a manager at the WUC Ramotswa
Table 1 Selected demographic and sanitation indicators for Botswana and the study area populations for 2011
Parameter Botswana Gaborone Ramotswa Lobatse
Demography Population size 2,024,904 231,592 30,382 29,689
Annual pop. growth rate (%) 2001–2011 1.9 2.5 3.9 −0.2
Average household size NA 3.1 4.3 3.1
Unemployment rate 2011 20 9.2 26.5 12.3
Sanitation stats in %
of households
Access to potable water piped indoors NA 58 37 42
Access to potable water
piped outdoors
NA 32 55 41
Access to flush toilet NA 50 38 36
Access to pit latrine 47 3 56 10
Source Statistics
Botswana 2015b
Statistics
Botswana 2015b
Statistics
Botswana 2015a
Statistics
Botswana 2015b
Maximum values per row of the three cities/towns are shown in italics.NA is “not applicable”
Hydrogeol J

Water Works, which is the local office that manages and pri-
oritizes water supply and sewage in Ramotswa. Interviews
were conducted in English, audio recorded, transcribed, and
coded for thematic analysis.
Groundwater monitoring, water quality sampling,
and analyses
Available water quality data from various periods were com-
bined for existing production BHs from 1983 to 2016, a period
that has a large gap in measurements between 1999 and 2013,
roughly coinciding with the period when the pumps were
offline. The only historic water quality data available for mon-
itoring BHs comes from a baseline study conducted in 2001 by
Staudt (2003). In addition, data include those collected by
Modisha (2017) in August 2016 from a subset of Ramotswa
production and monitoring BHs. New data generated in this
study were collected from six monitoring and 14 production
BHs (Fig. 2) on four occasions in Ramotswa: mid October
2016 (onset of wet season), late November 2016, early
January 2017, and early February 2017. The wet season had
peaked during January 2017 (Fig. S3 of the ESM). Some data
on depth to the groundwater going back to the 1980s are avail-
able. However, inconsistency, data gaps, and the effects of pro-
duction BH pumping prohibit analysis of long-term trends in
the depth to water table.
Permission was granted from two farmers to sample their
private BHs (BH4157 and BHZ22547), which are equipped
with pumps; these data are grouped with the monitoring BHs.
At each of the four sampling events sampling from the same 20
BHs was attempted (Fig. 2). Sometimes this was not possible
due to inaccessibility or certain production BHs being offline
(production BHs are not all running at the same time). Each trip
averaged 14–15 BHs. The sampled BHs represent a majority of
the BHs available for sampling in Ramotswa; they are also a mix
of locations in relation to town (upstream, in town, downstream;
Fig. 1) to see if the pit latrines in town had a measureable effect
on water quality; and many were also chosen to repeat measure-
ments from 2001 as reported in Staudt (2003). Stratigraphy and
casing information from some BH logs are available in Staudt
(2003) and summarized in Table S1 of the ESM.
Samples from the monitoring BHs were collected using a
Grundfos submersible centrifugal pump with an unregulated
flow rate of 0.7–1.5 m
3
h
−1
, which was the same as used by
Modisha (2017) and Staudt (2003) (Grundfos models MP-1 and
SQ 1.2-3 N, Bjerringbro, Denmark). Samples from the produc-
tion BHs were obtained from a valve on the in situ pumps.
The following variables were measured continuously with a
calibrated Quanta Water Quality Sensor (Hach, Loveland,
Colorado, USA) while sampling: temperature, pH, dissolved ox-
ygen (DO), and oxidation reduction potential (ORP). The outlet
of the pump tubing and the sensor were placed at the bottom of a
15-L plastic bucket that was allowed to continuously overflow.
This ensured the sensor was measuring water fresh from the
pump (presumably equal to in situ water quality) and the water
at the top of the bucket acted as a barrier to atmospheric gas
exchange. BHs were pumped until physical and chemical vari-
ables stabilized (USGS 2006), typically after 30–60 min of
pumping, at which time the groundwater samples were collected.
An unfiltered 500-ml sample was collected into a sterilized
glass bottle for fecal coliform analysis, a standard indicator of
fecal contamination from warm-blooded animals, including
humans and livestock. Samples were stored on ice in a cooler
until analyzed within 24 h at the WUC Mmamashia Water
Treatment Works microbiology laboratory following the ISO
standardized membrane filtration method (ISO 9308-1).
A second, 350-ml sample was filtered through a new Supor
0.45-μm membrane filter (Pall Corporation, Ann Arbor, MI,
USA) and stored in plastic bottles. These samples were refrig-
erated and delivered to Waterlab, a private service laboratory in
Pretoria, South Africa for the following hydrochemical analy-
ses that were conducted within ten days: NO
3
−
(NO
2
−
+NO
3
−
)
was measured colorimetrically using hydrazine reduction and
mercuric thiocyanate methods, respectively, on an Aquakem
250 spectrophotometer (Thermo Scientific, Waltham, MA,
USA) (USEPA 1993); DOC was measured by UV oxidation
to CO
2
, which was measured using an infrared analyzer
(Sievers 900 Analyzer, GE Analytical Instruments, Boulder,
CO, USA). Ammonia (NH
3
) and ammonium (NH
4+
)werealso
measured but concentrations were <0.1 and < 0.6 mg L
−1
,re-
spectively, at all sites and dates. Several additional analytes
(section S2 of the ESM) were also measured on the same sam-
ples and results are reported in section S3 of the ESM.
Caffeine and paraxanthine were measured in groundwater
samples as indicators of human waste contamination.
Caffeinated sodas, instant coffee, and black teas are popular
beverages in Ramotswa (Fig. S4 of the ESM). The half-lives
of caffeine and paraxanthine were estimated in a US estuary to
be 3 to >100 days and 11 to >100 days, respectively (Benotti
and Brownawell 2009). In a German karst aquifer system, the
caffeine half-life was estimated at 89 days (Hillebrand et al.
2015). This high range of values indicates the need for site-
specific investigations under the prevailing local conditions.
Sample collection was conducted following Matongo et al.
(2015). During the November 2016 and January 2017 trips
only, 500-ml samples for caffeine and paraxanthine analysis
were filtered in the same way as the hydrochemistry samples.
Samples were stored in a refrigerator before extraction within
one week. Samples were collected and extracted by someone
who had not consumed caffeine in the previous 5 days, and
stored in coolers and refrigerators not used for food and
beverages.
Solid phase extraction for caffeine and paraxanthine
followed the method in Matongo et al. (2015). Caffeine stan-
dard (C
8
H
10
N
4
O
2
, CAS No. 58-08-2) and paraxanthine stan-
dard (C
7
H
8
N
4
O
2
, CAS 611-59-6) were purchased from
Hydrogeol J

Sigma-Aldrich (St. Louis, MO, USA). A 10-mg L
−1
stock
standard was made by dissolving 10 mg caffeine and 10 mg
paraxanthine in 1 L of 50:50 mix of methanol and deionized
water. Extracted samples were analyzed 16 weeks after extrac-
tion by GC-MS on an Agilent 6890 N gas chromatograph with
a 5973 inert mass selective detector (Agilent Technologies, Santa
Clara, CA, USA) in select ion monitoring mode utilizing a 15-m
Rtx-35 ms capillary column (Restek Corporation, Bellefonte,
PA, USA). The oven program was 70 °C for 1 min, a
20 °C min
−1
ramp to 190 °C, a 15 °C min
−1
ramp to 210 °C,
and a final ramp of 30 °C min
−1
to 300 °C. The quantifying ions
were m/z 194 and 180 for caffeine and paraxanthine, respective-
ly. Three micrometre injections gave a limit of quantitation for
caffeine of 0.5 μgL
−1
and for paraxanthine of 10 μgL
−1
.
Evidence of denitrification, i.e., the production of N
2
O, in
the aquifer was sought to determine its potential for ISB
(USEPA 2013). Dissolved gas samples were collected from
the groundwater for N
2
O analysis. Because gas samples were
transported to the USA for analysis, they were only collected
during the final sampling trip in February 2017 to minimize
storage time between collection and analysis. Previous unpub-
lishedworkhasshownthatN
2
O concentration in gas samples
collected this way are stable for 90 days when stored in the dark
at room temperature (K. Kahmark, Michigan State University
Kellogg Biological Station, personal communication, 2016).
The dissolved gas sample collection method described in
Hamilton and Ostrom (2007) was used here with a few mod-
ifications. A 30-ml water sample and a 30-ml ambient air
sample were drawn into one gas tight syringe, shaken for
5 min to achieve equilibration, and 10 ml of headspace gas
was injected to over-pressurize a 5.92-ml glass vial with a
rubber septum (Labco Ltd., High Wycombe, UK). To ensure
the sampled water was not exposed to the atmosphere, it was
drawn through a narrow piece of polypropylene tubing ex-
tending from the syringe tip and inserted directly inside the
end of the pump tubing or far below the water surface in the
bucket. Samples were collected in triplicate at each site along
with a fourth sample of ambient air. Gas samples were stored
in the dark at room temperature and were analyzed within
30 days on a gas chromatograph (Agilent 7890, Agilent
Technologies, Santa Clara, CA, USA). N
2
O was analyzed
with a
63
Ni electron capture detector at 350 °C coupled to a
Gerstel MPS2XL automated headspace sampler (Gerstel,
Mülheim an der Ruhr, Germany). The system had a two-
column back-flush setup using Restek PP-Q 1/8″OD, 2.0-
mm ID, 80/100 mesh, 3-m-packed columns (Restek,
Bellefonte, PA, USA). The oven was set to 90 °C.
Calculations of N
2
O concentrations in groundwater followed
those described in Hamilton and Ostrom (2007) and are the
result of several steps. Briefly, the concentration of N
2
Odis-
solved in the original liquid sample is back-calculated using
the calculated Bunsen solubility coefficient, Henry’sLaw,and
the ideal gas law (Weiss 1974; Weiss and Price 1980). The
ambient N
2
O concentration of the air drawn into the syringe
for headspace equilibration was subtracted from the
final headspace concentration. In addition, the amount of N
2
O
dissolved in the water when it infiltrated the soil was predicted
assuming it was in equilibrium with the atmosphere, and this was
also subtracted from the measured N
2
O concentration.Therefore,
reported N
2
O concentrations are the amount of N
2
O that the
water accumulated below ground in excess of what it contained
from air equilibration in the unsaturated zone.
Multiple linear regression analyses of rainfall and water
quality data were conducted in R 3.3.2 (R Core Team 2017),
and coefficients and models are reported here with adjusted R
2
values. Plots and maps were created using the R packages
ggplot2 (Wickham 2009) and ggmap (Kahle and Wickham
2013).
Results
Changes in seasonal rainfall and variability
between 1922 and 2017 in Ramotswa, Gaborone,
and Lobatse
Seasonal total rainfall has decreased significantly in both the
wet and dry seasons since the 1920s (Gaborone and Lobatse)
and 1980 (Ramotswa) (Fig. 4a, model R
2
=73%, p<0.001,
Table 2: model 1). The number of rainy days has decreased
significantly in both seasons as well, but more so in the wet
season (Fig. 4b, model R
2
=81%, p<0.001, Table 2: model
2). Daily rainfall intensity (Eq. 2) has increased significantly
in both the wet and dry seasons across all stations (Fig. 4c,
model R
2
=35%,p< 0.001, Table 2: model 3). Because total
rainfall has decreased, the increase in intensity is driven by the
decline in number of rainy days in both seasons. Rainfall
variability (Fig. 4d) measured as the seasonal CV among
monthly rainfall totals, has increased significantly in both sea-
sons (model R
2
=64%,p<0.001,Table 2:model4).TheCV
increased more quickly in the dry season.
Rainfall extremes have also risen. Organizing the data by
percentiles (Fig. 4e) illustrates how the wettest dry seasons
(upper 10th percentile) are getting significantly drier over time
(model R
2
= 90%, p<0.001, Table 2: model 5). Also, the
wettest wet-seasons have become significantly wetter over
time. The driest dry seasons (lower 10th percentile) and the
majority of wet and dry seasons (middle 80% of observations)
had no significant trends. The wet-season lower 10th percen-
tile showed a drying trend but is not statistically significant.
In summary, rainfall metrics show that the climate in the
Ramotswa area has become more extreme, with greater rain-
fall variability and intensity in both seasons. While both sea-
sons have grown drier over time, the results indicate an in-
creasing probability of both droughts and floods.
Hydrogeol J

Interviews on water restrictions and sanitation
behavior
The interviews confirmed that pit latrines are common-
place throughout Ramotswa and have been the most
common sanitation method in Ramotswa since at least
the 1950s. In addition, many of the households with a
flush toilet (~38% Ramotswa population, Table 1)also
have a new or remnant, functional pit latrine, which
they use when the water supply is restricted. Pit latrine
depth varies depending on soil depth, but typically they
are thought to be about 1–1.5 m deep. Pit latrines closer
to the Notwane River are known to flood as the water
table rises with heavy rains. The WUC reports that new
and existing Ramotswa households continue to install
flush toilets and connect into the sewage network to
the present day.
Many older homes have more than one pit latrine,
some of which are abandoned and/or full. When pit
latrines fill, a resident either pays the WUC US$50 to
pump the waste out of the pit latrine and truck it to the
sewage ponds, or the resident digs a new pit latrine and
leaves the old one unused. It was unclear what propor-
tion of households pay to pump out a pit latrine or how
frequently this occurs.
Sources explained that during the 2013–2016 drought pe-
riod, when the reservoir capacity was at <5% (corresponding
to failure capacity, Fig. S2 of the ESM), the water supply was
shut off for 1–3 days at a time per week without notice. It is
common practice during water restrictions for those people
who have flush toilets to use pit latrines. This was also con-
firmed in a 2016 survey of 100 households in Ramotswa,
which also found that people did not typically store water
for flushing their toilet during water restrictions, usually be-
cause they did not have a storage tank and the cost of a tank
was prohibitive (T. Matsoga, University of Botswana, person-
al communication, 2017). Indeed, the Ramotswa WUC repre-
sentative reported a drastic drop in flow to the sewage system
during the 2014–2015 period:
B[Water restrictions] do affect people because
without those flush toilets …our sewage system
is not working. Because of these water restrictions,
it’s only receiving very little water, which cannot
keep it running. It’s very visible at our [treatment]
ponds…There won’t be any flow from one pond to
another.^
Similarly, a source recalled that during the water restrictions,
they would Bgo for about three days without water. And
Fig. 4 Dry- and wet-season rainfall patterns over time at Gaborone,
Lobatse, and Ramotswa weather stations (color). Significance is indicated
with an asterisk. aTotal rainfall, bnumber of rainy days, crainfall
intensity, dseasonal coefficient of variation (CV) among monthly rainfall
totals, eextreme rainfall events including upper and lower 10th percen-
tiles of seasonal rainfall totals, shape). See Table 2for model results
Hydrogeol J

definitely then you would need to use the pit latrines.^The
interviews and household survey cited above suggest that wa-
ter restrictions increased pit latrine usage in Ramotswa. This
view was confirmed informally in conversations with people
in other water management positions in Botswana, residents in
Ramotswa, and people living in other communities in
Botswana affected by water restrictions.
Nitrate concentrations and in situ denitrification
High groundwater NO
3
−
concentrations remain a problem in
Ramotswa. The highest NO
3
−
concentration measured in the
present study, 29 mg NO
3
−
–NL
−1
, was recorded at two sites:
production BH4422 in town in October 2016 and private
BH4157 at the cattle kraal (pen) in November 2016 (Fig. S5
Table 2 Historic rainfall
regressions Model Estimate Standard error pAdjusted R
2
Model p
1. Seasonal rainfall total = year + season (reference
level is dry season)
73% <0.001
(Intercept) 76.572 7.511 <0.001
b
Year −0.552 0.193 <0.01
b
Wet season 359.565 10.469 <0.001
b
2. Number rainy days = (year × season) + station
(reference levels are dry season and Gaborone
station)
81% <0.001
(Intercept) 13.431 0.720 <0.001
b
Year −0.040 0.021 0.062
b
Wet season 30.451 0.779 <0.001
b
Lobatse station −6.580 0.848 <0.001
b
Ramotswa station −13.600 1.221 <0.001
b
Year × wet season −0.180 0.029 <0.001
b
3. Intensity
a
= year + season+ station (reference levels
are dry season and Gaborone station)
35% <0.001
(Intercept) 2.700 0.061 <0.001
b
Year 0.007 0.001 <0.001
b
Wet season 0.435 0.066 <0.001
b
Lobatse station 0.519 0.072 <0.001
b
Ramotswa station 1.023 0.103 <0.001
b
4. CV = year × season (reference level is dry season) 64% <0.001
(Intercept) 1.724 0.260 <0.001
b
Year 0.004 0.001 <0.001
b
Wet season −0.995 0.037 <0.001
b
Year × wet season −0.003 0.001 <0.05
b
5. Seasonal total mm =year × percentile group × season (reference
levels are middle 80% and dry season)
90% <0.001
(intercept) 65.162 5.214 <0.001
b
Year −0.145 0.194 0.456
Upper 10th percentile 142.046 15.237 <0.001
b
Lower 10th percentile −60.271 15.136 <0.001
b
Wet season 375.420 7.450 <0.001
b
Year × upper −1.160 0.607 0.057
b
Year × lower 0.080 0.510 0.875
Year × middle × wet −0.218 0.278 0.432
Wet season, upper 10th perc. 130.839 21.970 <0.001
b
Wet season, lower 10th perc. −157.296 22.076 <0.001
b
Year × upper × wet 2.056 0.892 <0.05
b
Year × lower × wet −0.804 0.764 0.294
For all models, year is centered on the mean year
a
(Intensity + 0.5)
1/2
b
Indicates statistical significance
Hydrogeol J

of the ESM). BH4422 was unavailable for sampling after
October 2016. Monitoring BH4995 ranged from 23 to
27 mg NO
3
−
−NL
−1
for four measurements taken between
October 2016 and February 2017. These three BHs (4422,
4157, and 4995) also had three of the highest NO
3
−
–Ncon-
centrations in Staudt (2003)(Fig.5). Even though BH4422
and BH4995 showed a decline since 2001 (both around
40 mg NO
3
−
–NL
−1
then and near 30 mg NO
3
−
–NL
−1
in
2017), they are still well above the NO
3
−
water quality stan-
dard. Given variability in NO
3
−
concentrations and the lack
of long-term data (Fig. S5 of the ESM), it is unclear if the
measurements reported here indicate a long-term declining
trend. None of the BHs showed a statistically significant
trend in NO
3
−
concentrations over time between August
2016 and February 2017.
The molar ratio of NO
3
−
–N to DOC tended to be >1
and as high as 44 in the case of BH 4995, suggesting
DOC is potentially limiting denitrification at certain BHs
(Fig. S6 of the ESM), which indicates the feasibility of
ISB. The production BHs that exceeded the NO
3
−
–N
drinking water standard, BH4422 and BH4400, also
showed C-limitation.
N
2
O concentrations were orders of magnitude greater
than atmospheric equilibrium concentration (most sam-
ples had >1 μgN
2
O–NL
−1
compared to the atmospher-
ic concentration of 0.15 μgN
2
O–NL
−1
), indicating
high denitrification activity. N
2
O concentrations were
positively related to nitrate concentration (Fig. 6). A
linear regression of log(N
2
O concentration + 0.5) by
NO
3
−
concentration was highly significant (p< 0.0001)
andexplainsmuchofthevariabilityinN
2
O(R
2
=70%,
Table S2 of the ESM;Fig.6). BH4160 had 47.75 μg
N
2
O–NL
−1
(the maximum by far). Measurements with
the highest NO
3
−
and N
2
O were all C-limited (Fig. 6),
suggesting the potential for ISB.
Nitrate source tracking
Fecal coliforms were not detected in any of the production
BHs but were detected in several monitoring BHs (Fig. 7).
The maximum detection limit of 200 colony forming units
(CFU) per 100 ml was observed at least once in BH4371,
BH4995, and BH10129, indicating that actual concentrations
could be higher. The Botswana drinking water standard is zero
CFU per 100 ml (Walmsley and Patel 2011). The presence of
fecal coliforms suggests that other fecal pathogens are likely
to be present in the aquifer. High NO
3
−
–N concentrations did
not always coincide with the presence of fecal coliforms, but
the presence of fecal coliforms tended to coincide with high
NO
3
−
–N concentrations (Fig. S7 of the ESM). Regression
analysis of samples with fecal coliforms by NO
3
−
was signif-
icant (p<0.05, Table S2 of the ESM) but only weakly ex-
plains the variation (R
2
= 11%).
Caffeine and paraxanthine were present in several BHs
at both sampling times (November 2016 and January 2017;
Fig. 7b,c). The detection (and nondetection) of paraxanthine
at many of the same BHs as caffeine confirms the low
likelihood of false positives during sampling and analysis.
Both compounds were measured in more wells and at
slightly higher concentrations in November compared to
January. The few BHs where the compounds were detected
in January were BHs where they were also detected in
November (Fig. 7b,c). Where detected, caffeine concentra-
tions ranged from 14 to 56 ng L
−1
, and paraxanthine con-
centrations ranged from 180 to 770 ng L
−1
. Caffeine and
paraxanthine were found in several monitoring BHs as
well as production BHs, including production BH4340,
BH4358, and BH4373, which are upstream of town. To
test whether caffeine was tracing the NO
3
−
pollution, a
regression was used to predict caffeine with an interaction
between the NO
3
−
concentration and C-limitation status
Fig. 5 Changes in NO
3
−
–N concentrations in Ramotswa BHs over time and space. Circle size indicates BH NO
3
−
–N concentrations. aConcentrations in
2001 (Staudt 2003) compared to bmean concentrations measured in this study
Hydrogeol J

(whether the molar ratio of NO
3
–
−
N to DOC was >1 or <
1). The regression was significant (p< 0.01) and explains
67% of the variability (Table S2 and Fig. S8 both of the
ESM).
Discussion
Effect of climate change on water supply in Ramotswa
Declining rainfall and increasing variability
The analyses of the last 90 years of rainfall data strong-
ly suggest that the South East District’s climate is
changing (Fig. 4). Declines in seasonal total rainfall
and number of days with rain, together with increasing
variability and extremes, are likely to increase the fre-
quency and intensity of droughts and floods such as the
most recent 2013–2016 drought. This supports H
1
:
Climate change-induced droughts are becoming more
frequent (Fig. 1), which is in agreement with other stud-
ies of rainfall amounts in southern Africa (Niang et al.
2014; Hodnebrog et al. 2016). The climate models used
in the IPCC Fifth Assessment Report project Botswana
and Namibia will Bvery likely^(i.e., >90% probability)
continue to see declines in mean annual rainfall, a delay
in the onset of the wet season, and by 2100, drier and
shorter wet seasons and drier dry seasons (Niang et al.
2014). A delay in the onset of the wet season was not
detected in the rainfall data used here.
Changes in the variability and intensity of rainfall may
have more significant influences on drought and flood fre-
quency. The South East District’s rainfall data showed a de-
cline in the number of days with rain and increases in seasonal
rainfall intensity, in agreement with New et al. (2006).
Climate change impacts on Ramotswa drinking water supply
According to the aforementioned climate results and other
studies on Botswana’s rainfall trends (Kenabatho et al. 2012;
Byakatonda et al. 2018), the Gaborone reservoir is likely to
reach failure capacity more frequently in the future (Fig. 4).
As the temperature is likely to increase 3.4–4.2 °C by 2100
in southern Africa (Niang et al. 2014), evaporation from
reservoirs, already >2,000 mm year
−1
(Altchenko et al.
2016), will likely increase, reducing water security of reser-
voir storage. For people who rely on surface water for their
entire water supply or for diluting a contaminated ground-
water supply these changes in the climate threaten their wa-
ter security.
Effect of drought and water restrictions on sanitation
in Ramotswa
The key informant interviews strongly supported H
2
: Water
shortages affect sanitation behavior, increasing pit latrine use
(Fig. 1). Changes in the water system limited many people’s
sanitation options, and changed their sanitation behavior. With
an increasing likelihood of events like the recent 3-year
drought and no change in water supply and associated treat-
ment infrastructure, Ramotswa will experience more frequent
disruptions in water supply, and as a consequence there will be
more use of pit latrines (Fig. 1). Based on the most recent
population census, water restrictions potentially increase the
number of people using pit latrines by about 14,250 (Statistics
Botswana 2015c).
Groundwater quality and in situ denitrification
bioremediation potential
The Ramotswa groundwater continues to exhibit NO
3
−
con-
centrations in excess of the Botswana drinking water standard
of 50 mg NO
3
−
L
−1
. This includes two of the production BHs,
4422 and 4400, and several monitoring BHs (Fig. 5). The
concentrations on the whole do not appear to have increased
or decreased significantly since the Staudt (2003) observa-
tions, despite Ramotswa’spopulationincreasingby50%be-
tween 2001 and 2011 (Statistics Botswana 2015c). There may
be reasons why population growth and NO
3
−
concentration
Fig. 6 Dissolved N
2
O was significantly and positively related to nitrate
concentration. Error bars represent N
2
O standard error among three
replicate measurements on one sample date. Shaded area is the 95%
confidence interval for regression analysis (p< 0.0001 and adjusted
R
2
=0.70,TableS2oftheESM). BH 4422 and BH 4400 were inacces-
sible in February 2017 (only time N
2
O samples were collected). Many
BHs were C-limited over the four sampling trips (Fig. S6 of the ESM).
BH 4160 N
2
O concentration was 47.75 μgN
2
OL
−1
(not shown) and was
not C-limited
Hydrogeol J

are not directly related. Since approximately 2000, the newly
expanding peri-urban areas of Ramotswa are connected to the
sewage network and most households are believed to be
connected (source: interviews), though the recent drought
incentivized building pit latrines at new homes. Older
households located near the sewage network continue to
connect. Also, a greater proportion of new pit latrines,
compared to old pit latrines, may be lined, preventing
additional nitrate contamination. Given the data available,
it seems that the sewage network could be reducing the
volume of human waste entering pit latrines and poten-
tially affecting the aquifer. However, with increasing like-
lihoods of long-term droughts, water shortages, and water
restrictions, people with flush toilets may be forced to use
pit latrines, jeopardizing the preventive effect of the sew-
age network on groundwater pollution.
Fig. 7 Maps of fecal contamination indicators. aMean fecal coliform
count, an indicator of warm-blooded animal fecal contamination, from a
maximum of four measurements between the October 2016 and February
2017. No fecal coliforms were detected at any of the production BHs. b
Caffeine and cparaxanthine concentrations, which serve as indicators of
human waste contamination, measured in November 2016 (left panel)
and January 2017 (right panel)
Hydrogeol J

High N
2
O supersaturation in the groundwater suggests de-
nitrification is occurring in the aquifer (Fig. 6), while low
DOC concentrations relative to NO
3
−
concentrations indicate
that denitrification is C-limited (Fig. S6 of the ESM),
supporting H
4
: Enhanced denitrification via ISB of the
Ramotswa aquifer has the potential to reduce nitrate contam-
ination. Jacks et al. (1999) also demonstrated denitrification
occurrence in Ramotswa groundwater by measuring
15
N/
14
N
ratios in the groundwater and non N-fixing tree leaves.
Sources of nitrate contamination
N fertilizer, organic matter, livestock waste, and human waste
are potential sources of NO
3
−
contamination; however, N fer-
tilizer is ruled out as a significant contributor to NO
3
−
contam-
ination for several reasons. First, two of the production BHs
south of town (4358 and 4340) are located in cultivated farm
fields, yet their mean concentrations were consistently <1 mg
NO
3
−
–NL
−1
and N
2
O was also low. Second, samples with
fecal coliform contamination also tended to be consistently
high in NO
3
−
(Fig. S5 of the ESM), suggesting a livestock
and/or human source for both. Third, synthetic fertilizer only
became available to farmers in the area in the last few years as
a subsidy from the government, nevertheless NO
3
−
has been
high in the Ramotswa aquifer since the 1980s (Staudt 2003).
As for livestock waste, there are no feedlots in Ramotswa
where livestock waste is concentrated. Livestock waste on the
soil surface is commonplace throughout Ramotswa as animals
freely graze around town. Fences protect production BHs but
not monitoring BHs. Recharge mechanisms are complex.
Recharge can be <1 mm year
−1
(Post et al. 2012) when it
occurs by soil moisture infiltration but it can be rapid, up to
20 mm year
−1
, in localized areas of runoff percolation, ephem-
eral riverbed infiltration, and preferential flow through
karstified outcrops such as the outcrop the town sits on
(Gieske 1992). As a consequence, the likelihood of contami-
nation from dispersed livestock waste infiltrating from the
surface through soil moisture is low compared to the likeli-
hood of contamination from pit latrines. Livestock concentra-
tion in areas like Ramotswa with surface karstification might
present more significant livestock contamination risk. With
only one BH outside of town used for watering cattle
(BH4157), statistical conclusions about the direct impact of
livestock are not possible (at this BH, NO
3
−
mean concentra-
tion was 15 mg NO
3
−
–NL
−1
, fecal coliforms ranged from
none detected to 5 FCU, and caffeine and paraxanthine were
not detected). Thus, livestock as a source of some of the NO
3
−
cannot be completely ruled out.
Several lines of evidence suggest a predominantly human
source, supporting H
3
:Nitrate pollution is primarily originat-
ing from human waste contamination. Fecal coliform concen-
trations were highest in town (but also at BH4157) and con-
sistently low upstream. However, some samples had high
NO
3
−
and no fecal coliforms indicating that other sources of
NO
3
−
may exist or that fecal coliforms are not a conservative
or reliable tracer of NO
3
−
as in Peeler et al. (2006)orhuman
waste as in Glassmeyer et al. (2005). The strongest line of
evidence that NO
3
−
is from a predominantly human source
is that caffeine and paraxanthine were present in many BHs
in the town (Fig. 7) and caffeine concentrations correlated
significantly with NO
3
concentrations (Fig. S8 of the ESM).
However, this does not identify the exact source of human
waste. The thousands of pit latrines in Ramotswa are the most
likely sources given the locations and co-occurrence of the
aforementioned indicators. A study in Mochudi, a town
55 km north of Ramotswa and also located on the Notwane
River, found pit latrines were the source of NO
3
−
pollution of
the groundwater (Lagerstedt et al. 1994). It is possible that
some human waste comes from other sources like improperly
maintained or sited septic tanks, ruptures in the sewage net-
work, and/or leaking settling ponds. The sewage settling
ponds north (downstream) of town are unlikely sources of
contamination, but septic systems and sewage networks have
been shown to contaminate groundwater with N and other
pollutants (Katz et al. 2011).
The detection of caffeine in the southernmost BHs (4373,
4358, and 4340) is perplexing. These BHs had <0.5 mg
NO
3
−
–N(Fig.5) at all sample dates and <1.5 μgN
2
O–
NL
−1
, suggesting that there was either no NO
3
−
contamina-
tion or NO
3
−
had been fully denitrified to N
2
,orN
2
O gas had
escaped from the aquifer. In addition, the presence of caffeine
at these southern (upstream) locations could mean the caffeine
is not coming from the pit latrines. Below are a couple spec-
ulative explanations for this peculiarity.
First, these southernmost production BHs would have a
largeconeofdepressionwhentheyarepumping,sothat
their chemistry represents a wider area that potentially in-
cludes recharge zones affected by human waste. Along this
flowpath, caffeine could remain in the water, but NO
3
−
is
completely denitrified to N
2
so that neither NO
3
−
nor N
2
O
are present. The caffeine could persist, even along C-
limited flow paths (Fig. S8 of the ESM), because
denitrifying bacteria might not use caffeine as a source of
DOC. Indeed, caffeine is considered toxic to many bacteria
(Mazzafera 2004). Also, the conditions that enable micro-
bial caffeine degradation are different from the conditions
necessary for microbial NO
3
−
removal: denitrification re-
quires low oxygen conditions, whereas microbial degrada-
tion of caffeine seems to require aerobic conditions
(Mazzafera 2004). Therefore, samples collected down-
stream from a common source of NO
3
−
and caffeine could
exhibit caffeine without NO
3
−
or vice versa, depending on
the microbial community and oxygen availability.
Additionally, caffeine in these southern production
BHs could be from the Notwane River, which may car-
ry human and livestock waste contamination when it is
Hydrogeol J

flowing. The river likely loses water to groundwater
recharge and, anecdotally, BHs close to the river can
be submerged in flood water, potentially affecting the
chemistry of these southernmost BHs located in the
floodplain. Unfortunately, Notwane River chemistry data
at Ramotswa are not available to test this idea. But a
joint water quality report by the Departments of Water
Affairs (DWA) in Botswana and South Africa (2013)
found that NO
3
−
–Nwas8mgL
−1
after the Gaborone
waste water discharges into the Notwane River (down-
stream of Ramotswa and the Gaborone Reservoir),
which is low compared to >20 mg L
−1
in several
Ramotswa BHs reported here. Upstream of Ramotswa
is the town of Lobatse, which has a similar population
size as Ramotswa, confined animal feeding operations,
and a slaughterhouse. Its wastewater effluent feeds into
the Notwane Dam, 33 km upstream of Ramotswa. There
are no other large settlements on the Notwane River
between Lobatse and Ramotswa. Perhaps NO
3
−
but not
caffeine from Lobatse wastewater is removed by the
time the river recharges the aquifer in Ramotswa. If
the river is the source of caffeine for the southernmost
BHs, then it does not explain the presence of caffeine in
upland BHs like 287, 10129, or 4995—meaning, pit
latrines and not the river likely contribute caffeine to
the groundwater at least in these upland BHs.
Greater understanding is needed of the role of the Notwane
River in groundwater recharge and quality in Ramotswa to
assess its potential as a contamination source. However, the
co-occurrence of NO
3
−
, fecal coliforms, and caffeine in the
sampled BHs (except for the southernmost BHs) strongly sug-
gests human sources of contamination within Ramotswa,
which are most likely predominantly from the thousands of
pit latrines (Table 1).
Indirect impact of climate change on groundwater
quality
It was shown that human-induced changes in the climate,
sanitation behavior, and groundwater interact to put
Ramotswa’s water security at risk (Fig. 1). Water insecurity
in the district is likely to increase due to changes in environ-
mental and social conditions. Regardless of climate change,
social pressures on water supply will likely perpetuate water
scarcity in the South East District (Taylor et al. 2009;Calow
et al. 2010; Niang et al. 2014), and undoubtedly exacerbate,
and be exacerbated by, water quality threats. Ramotswa’sde-
pendence on Gaborone for diluting its groundwater means
Ramotswa’s access to safe drinking water (and water to flush
toilets) is paradoxically threatened by the combination of re-
duced surface-water quantity (via climate change) and poor
groundwater quality, i.e., a water quantity problem intensifies
awaterquality problem.
Recommendations
A three-pronged approach is recommended: reduce contami-
nation, in situ remediation of the groundwater, and local treat-
ment of Ramotswa groundwater. Individually, none of these
approaches provides an immediate or long-term solution, but
together they could turn Ramotswa groundwater management
into a success story.
Reduce contamination. If Ramotswa’s high annual population
growth rate from 2001 to 2011 of 3.9% continues (Statistics
Botswana 2015c), Ramotswa will see an increase in both wa-
ter demand and volume of human waste over time. Water and
sanitation infrastructure should be optimized to ensure safe
and reliable sanitation access, match current and future water
availability, and protect groundwater quality. Lining pit la-
trines should be a priority. When an unlined pit latrine is
pumped out, a liner could be installed. Incentives could be
designed to increase the number of pit latrines that get pumped
rather than abandoning full pit latrines and building new ones.
The climate analyses here suggest flush toilets will be less and
less usable with increasing frequency of droughts and water
restrictions. Captured and stored rain or grey water could be
used to flush toilets, but Ramotswa residents need assistance
in order to obtain the necessary equipment. Planning for im-
provements in infrastructure (e.g., expanding the sewage sys-
tem) requires information about the current system including
well production and water use. Useful spatial data that are
missing include: up to date maps of the piped sewage network,
homes connected, pit latrines and their status (including lined/
unlined, in-use/abandoned, etc.), and septic tanks.
Remediate. The presence of N
2
Odissolvedinthegroundwa-
ter confirmed that the microbes and conditions in the aquifer
are suitable for denitrification. The limited availability of
DOC relative to NO
3
−
in Ramotswa groundwater demonstrat-
ed that ISB with supplemental DOC has the potential to en-
hance removal of NO
3
−
via denitrification (Fig. 6). A vegeta-
ble oil amendment could act as a carbon substrate (USEPA
2013), and this has been proposed for ISB in Ramotswa. ISB
has the potential to be cost-effective and less energy intensive
than traditional water treatment such as reverse osmosis.
However, many factors must be taken into consideration
to test the feasibility of ISB in Ramotswa, including the
cost (relative to the funds needed to build and run a water
treatment facility), sustainability, dolomite suitability for
ISB (Tompkins et al. 2001), site specific hydraulic char-
acteristics, amendment longevity, performance monitoring
using tools such as stable isotopes, and effects on down-
stream geochemistry (USEPA 2013;Majoneetal.2015).
Both contamination prevention and remediation are im-
portant approaches to the mitigation of drinking water
contamination.
Hydrogeol J

Treat. A water treatment facility for Ramotswa’sgroundwater
using reverse osmosis has been planned in Boatle (grey
arrows in Fig. 3) but progress has stalled for over a year. If
Ramotswa were able to treat the groundwater and use it for
100% of its supply, then greater extractions will be required,
jeopardizing a sustainable pumping rate. Ramotswa’swater
security is also tied to demand and failures in infrastructure
in other parts of the South East District water delivery scheme
(Fig. 3). Therefore, it would be helpful tocompile past and on-
going water abstraction volumes and conjunctive use of sur-
face and groundwater across the water supply scheme in real
time in one database to manage both supply and use locally
and across the South East District.
To understand how NO
3
−
concentrations are responding to
the complex factors involved in this CHANS and the efficacy
of remediation efforts, systematic water quality and quantity
monitoring and associated funding should be a priority.
Further, the water quality of private BHs, which provide un-
treated water to many people during water restrictions, should
be monitored.
Conclusions
This work demonstrated the indirect effect of climate change on
groundwater quality in Ramotswa, Botswana. Historic rainfall
analyses suggest that events like the 2013–2016 drought are
likely to become more frequent in Ramotswa and the South
East District, increasing pressure on groundwater resources.
The drought led to water restrictions, inducing people with flush
toilets to use pit latrines. Groundwater NO
3
−
concentrations
exceeded the drinking water standard. The presence and extent
of NO
3
−
, fecal coliforms, caffeine, and paraxanthine in ground-
water wells suggest that N fertilizer and livestock waste were not
likely major sources of NO
3
−
pollution; the most reasonable
hypothesis is that human waste from the thousands of pit latrines
is a major source of the NO
3
−
. High dissolved N
2
O concentra-
tions in the groundwater suggested in situ denitrification is hap-
pening; however, low DOC concentrations relative to NO
3
−
concentrations suggest that denitrification is C-limited.
Therefore, ISB providing an additional C source in the aquifer
could enhance NO
3
−
removal by denitrification. The caffeine,
DOC, and N
2
O analyses provide new information that advances
the understanding of the drivers of Ramotswa groundwater
quality and potential solutions.
The CHANS approach allowed us to understand water qual-
ity in the larger context of interacting social and ecological
drivers. This perspective illuminated the interactions between
climate (drought), human behavior, and groundwater quality.
Without understanding the human behavior aspects of the ni-
trate contamination, a driver of the problem (pit latrine use) that
is likely to increase in frequency and intensity in coming
decades would remain obscured. This CHANS study provides
a template for socio-hydrological studies in other regions that
may lead to improved predictions of indirect impacts of climate
change and more sustainable water management.
In a global context, this story is not unique to Ramotswa.
Rapid urbanization in developing countries brings rapid in-
creases in the number of pit latrines, often in communities that
use shallow groundwater for drinking (Cronin et al. 2007).
Rather than thinking about and managing water supply and
sanitation separately, communities, resource managers, and
policy makers need to plan a water infrastructure system in-
corporating and addressing the interconnections between cli-
mate, water resources, sanitation, and water infrastructure.
Such an integrated system that protects groundwater quality
and quantity will ultimately reduce the cost of water treatment,
reduce water-borne disease transmission, and strengthen the
community’s water security and resilience in the face of ur-
banization and climate change.
Acknowledgements In-kind support was provided by the Botswana Dept.
of Water Affairs, Water Utilities Corporation, the Botswana Geosciences
Institute, the Dept. of Meteorological Services in Botswana; the South
African Council for Scientific and Industrial Research (CSIR) in Pretoria
and B. Pongoma; K. Kahmark; and the Kellogg Biological Station Long
Term Ecological Research program at Michigan State University. This
project would not have been possible without the invaluable support of
the following individuals: M. Molefe, P. Makoba, P. Moleje, M. Staudt,
K. Gunter, S. Ranamane, K. Gaoagelwe, B. Moodley, A. Naidoo, K. Puni
Gaboutloeloe, B. Mpeo, D. Weed, N. Martin, S. Baqa, and O. Modisha. A.
Reimer, D. Stuart, R.J. Stevenson, and G.P. Robertson also provided helpful
comments on an earlier draft.
Funding information This research was funded by the US Agency for
International Development (USAID) through a US Borlaug Fellowship in
Global Food Security (Award No. A1102.2) to McGill, who was kindly
hosted by the International Water Management Institute Southern Africa.
McGill was also supported by a US National Science Foundation
Graduate Research Fellowship (DGE-1424871). This work was under-
taken in the context of the project BThe Potential Role of the
Transboundary Ramotswa Aquifer^, funded by USAID under the terms
of Award No. AID-674-IO-17-00003 and the CGIAR Strategic Research
Program on Water, Land and Ecosystems (WLE), and implemented by
the International Water Management Institute (IWMI) and the
International Groundwater Resources Assessment Centre (IGRAC).
Open Access This article is distributed under the terms of the Creative
Commons Attribution 4.0 International License (http://
creativecommons.org/licenses/by/4.0/), which permits unrestricted use,
distribution, and reproduction in any medium, provided you give
appropriate credit to the original author(s) and the source, provide a link
to the Creative Commons license, and indicate if changes were made.
References
Altchenko Y, Lefore N, Villholth K, Ebrahim G, Genco A, Pierce K,
Woolf R, Moseltlhi B, Moyo T, Kenabatho P, Nijsten G-J (2016)
Resilience in the Limpopo Basin: the potential role of the
transboundary Ramotswa aquifer. Baseline report, International
Water Management Institute, Pretoria, South Africa. https://drive.
Hydrogeol J

google.com/file/d/0B-Ajpddeja2IX0JidWVlWE95aW8/view.
Accessed November 2018
Altchenko Y, Genco A, Pierce K, Woolf R, Nijsten GJ, Ansems N,
Magombeyi M, Ebrahim G, Lautze J, Villholth K, Lefore N,
Modisha RCO, Baqa S, McGill BM, Kenabatho P (2017)
Resilience in the Limpopo basin: the potential role of the
Transboundary Ramotswa aquifer. Hydrogeology report,
International Water Management Institute, Pretoria, South Africa.
http://ramotswa.iwmi.org/Data/Sites/38/media/pdf/
hydrogeological_report_2017.pdf. Accessed November 2018
Ashbolt NJ, Grabow WOK, Snozzie M (2001) Indicators of microbial
water quality. In: Fewtrell L, Bartram J (eds) Water quality: guide-
lines, standards and health. World Health Organization, Geneva
Benotti MJ, Brownawell BJ (2009) Microbial degradation of pharmaceu-
ticals in estuarine and coastal seawater. Environ Pollut 157:994–
1002. https://doi.org/10.1016/j.envpol.2008.10.009
Buerge IJ, Poiger T, Müller MD, Buser H-R (2003) Caffeine, an anthro-
pogenic marker for wastewater contamination of surface waters.
Environ Sci Technol 37:691–700. https://doi.org/10.1021/
es020125z
Byakatonda J, Parida BP, Kenabatho PK, Moalafhi DB (2018) Prediction
of onset and cessation of austral summer rainfall and dry spell fre-
quency analysis in semiarid Botswana. Theor Appl Climatol. https://
doi.org/10.1007/s00704-017-2358-4
Calow R, MacDonald A (2009) What will climate change mean for
groundwater supply in Africa? ODI background note, Overseas
Development Institute, British Geological Survey, London. https://
www.odi.org/sites/odi.org.uk/files/odi-assets/publications-opinion-
files/4120.pdf. Accessed November 2018
Calow RC, MacDonald AM, Nicol AL, Robins NS (2010) Ground water
security and drought in Africa: linking availability, access, and de-
mand. Groundwater 48:246–256. https://doi.org/10.1111/j.1745-
6584.2009.00558.x
Collins SL, Carpenter SR, Swinton SM, Orenstein DE, Childers DL,
Gragson TL, Grimm NB, Grove JM, Harlan SL, Kaye JP, Knapp
AK, Kofinas GP, Magnuson JJ, McDowell WH, Melack JM,
Ogden LA, Robertson GP, Smith MD, Whitmer AC (2011) An
integrated conceptual framework for long-term social–ecological
research. Front Ecol Environ 9:351–357. https://doi.org/10.1890/
100068
Cronin AA, Hoadley AW, Gibson J, Breslin N, Kouonto Komou F,
Haldin L, Pedley S (2007) Urbanisation effects on groundwater
chemical quality: findings focusing on the nitrate problem from 2
African cities reliant on on-site sanitation. J Water Health 5:441–
454. https://doi.org/10.2166/wh.2007.040
Famiglietti JS (2014) The global groundwater crisis. Nat Clim Chang 4:
945–948
Fenech C, Rock L, Nolan K, Tobin J, Morrissey A (2012) The potential
for a suite of isotope and chemical markers to differentiate sources of
nitrate contamination: a review. Water Res 46:2023–2041. https://
doi.org/10.1016/j.watres.2012.01.044
Ferguson G, Gleeson T (2012) Vulnerability of coastal aquifers to
groundwater use and climate change. Nat Clim Chang 2:342.
https://doi.org/10.1038/nclimate1413
Gieske A (1992) Dynamics of groundwater recharge: a case study in
semi-arid eastern Botswana. PhD Thesis, Free University,
Amsterdam
Glassmeyer ST, Furlong ET, Kolpin D, Cahill JD, Zaugg SD, Werner SL
(2005) Transport of chemical and microbial compounds from
known wastewater discharges: potential for use as indicators of hu-
man fecal contamination. Environ Sci Technol 39:5157–5169.
https://doi.org/10.1021/es048120k
Godfrey E, Woessner WW, Benotti MJ (2007) Pharmaceuticals in on-site
sewage effluent and ground water, western Montana. Ground Water
45:263–271. https://doi.org/10.1111/j.1745-6584.2006.00288.x
Graham JP, Polizzotto ML (2013) Pit latrines and their impacts on
groundwater quality: a systematic review. Environ Health Perspect
121:521–530. https://doi.org/10.1289/ehp.1206028
Hamilton SK, Ostrom NE (2007) Measurement of the stable isotope ratio
of dissolved N2 in 15N tracer experiments. Limnol Oceanogr
Methods 5:233–240
Heaton THE, Talma AS, Vogel JC (1983) Origin and history of nitrate in
confined groundwater in the western Kalahari. J Hydrol 62:243–262
Hillebrand O, Nodler K, Licha T, Sauter M, Geyer T (2012) Caffeine as
an indicator for the quantification of untreated wastewater in karst
systems. Water Res 46:395–402. https://doi.org/10.1016/j.watres.
2011.11.003
Hillebrand O, Nödler K, Sauter M, Licha T (2015) Multitracer experi-
ment to evaluate the attenuation of selected organic micropollutants
in a karst aquifer. Sci Total Environ 506–507:338–343. https://doi.
org/10.1016/j.scitotenv.2014.10.102
Hodnebrog Ø, Myhre G, Forster PM, Sillmann J, Samset BH (2016)
Local biomass burning is a dominant cause of the observed precip-
itation reduction in southern Africa. Nat Commun 7:11236. https://
doi.org/10.1038/ncomms11236
Howard G, Bartram J (2010) Vision 2030: the resilience of water supply
and sanitation in the face of climate change. World Health
Organization, Geneva, Switzerland. http://www.who.int/water_
sanitation_health/publications/9789241598422_cdrom/en/.
Accessed November 2018
Howard G, Calow R, Macdonald A, Bartram J (2016) Climate change
and water and sanitation: likely impacts and emerging trends for
action. In: Gadgil A, Gadgil TP (eds) Annual review of environment
and resources, vol 41. Annual Reviews, Palo Alto, CA, pp 253–276
Jacks G, Sefe F, Carling M, Hammar M, Letsamao P (1999) Tentative
nitrogen budget for pit latrines: eastern Botswana. Environ Geol 38:
199–203
Jiménez Cisneros BE, Oki T, Arnell NW, Benito G, Cogley JG, Döll P,
Jiang T, Mwakalila SS (2014) Freshwater resources. In: Field CB,
Barros VR, Dokken DJ, Mach KJ, Mastrandrea MD, Bilir TE,
Chatterjee M, Ebi KL, Estrada YO, Genova RC, Girma B, Kissel
ES, levy AN, MacCracken S, Mastrandrea PR, White LL (eds)
Climate change 2014: impacts, adaptation, and vulnerability part a:
global and sectoral aspects contribution of Working Group II to the
fifth assessment report of the Intergovernmental Panel of Climate
Change. IPCC, Geneva, pp 229–269
Kahle D, Wickham H (2013) Ggmap: spatial visualization with ggplot 2.
R J 5:144–161 URL http://journal.r-project.org/archive/2013-2011/
kahle-wickham.pdf. Accessed November 2018
Katz BG, Eberts SM, Kauffman LJ (2011) Using Cl/Br ratios and other
indicators to assess potential impacts on groundwater quality from
septic systems: a review and examples from principal aquifers in the
United States. J Hydrol 397:151–166. https://doi.org/10.1016/j.
jhydrol.2010.11.017
Kenabatho PK, Parida BP, Moalafhi DB (2012) The value of large-scale
climate variables in climate change assessment: the case of
Botswana’s rainfall. Phys Chem Earth A/B/C. 50–52:64–71.
https://doi.org/10.1016/j.pce.2012.08.006
Lagerstedt E, Jacks G, Sefe F (1994) Nitrate in groundwater and N cir-
culation in eastern Botswana. Environ Geol 23:60–64
Lapworth DJ, Baran N, Stuart ME, Ward RS (2012) Emerging organic
contaminants in groundwater: a review of sources, fate and occur-
rence. Environ Pollut 163:287–303. https://doi.org/10.1016/j.
envpol.2011.12.034
Lapworth DJ, Nkhuwa DCW, Okotto-Okotto J, Pedley S, Stuart ME,
Tijani MN, Wright J (2017) Urban groundwater quality in sub-
Saharan Africa: current status and implications for water security
and public health. Hydrogeol J 25:1093–1116. https://doi.org/10.
1007/s10040-016-1516-6
Liu JG, Dietz T, Carpenter SR, Folke C, Alberti M, Redman CL,
Schneider SH, Ostrom E, Pell AN, Lubchenco J, Taylor WW,
Hydrogeol J

Ouyang ZY, Deadman P, Kratz T, Provencher W (2007) Coupled
human and natural systems. Ambio 36:639–649. https://doi.org/
10.1579/0044-7447(2007)36[639:chans]2.0.co;2
Majone M, Verdini R, Aulenta F, Rossetti S, Tandoi V, Kalogerakis N,
Agathos S, Puig S, Zanaroli G, Fava F (2015) In situ groundwater
and sediment bioremediation: barriers and perspectives at European
contaminated sites. New Biotechnol 32:133–146. https://doi.org/10.
1016/j.nbt.2014.02.011
Matongo S, Birungi G, Moodley B, Ndungu P (2015) Occurrence of
selected pharmaceuticals in water and sediment of Umgeni River,
KwaZulu-Natal, South Africa. Environ Sci Pollut Res 22:10298–
10308. https://doi.org/10.1007/s11356-015-4217-0
Mazzafera P (2004) Catabolisms of caffeine in plants and microorgan-
isms. Front Biosci 9:1348–1359
Millenium Ecosystem Assessment (2005) Ecosystems and human well-
being: synthesis. Island: Washington, DC
Modisha RCO (2017) Investigation of the Ramotswa Trandboundary
Aquifer Area, groundwater flow and pollution. MSc Thesis,
University of Witwatersrand, Johannesburg, South Africa
Moehadu M (2014) Ramotswa wellfield monitoring report 2013-14. In:
Unit WUC-WRMS-G (ed) Water utilities corporation. Gaborone,
Botswana
North American Climate Extremes Monitoring (2017) Indices. https://
www.ncdc.noaa.gov/extremes/nacem/indices. Accessed 9
June 2017 2017
New M, Hewitson B, Stephenson DB, Tsiga A, Kruger A, Manhique A,
Gomez B, Coelho CAS, Masisi DN, Kululanga E, Mbambalala E,
Adesina F, Saleh H, Kanyanga J, Adosi J, Bulane L, Fortunata L,
Mdoka ML, Lajoie R (2006) Evidence of trends in daily climate
extremes over southern and West Africa. J Geophys Res-Atmos
111. https://doi.org/10.1029/2005jd006289
Niang I, Ruppel OC, Abdrabo MA, Essel A, Lennard C, Padgham JPU
(2014) Africa. In: Barros VR, field CB, Dokken DJ, Mastrandrea
MD, Mach KJ, Bilir TE, Chatterjee M, Ebi KL, Estrada YO, Genova
RC, Girma B, Kissel ES, levyAN, MacCracken S, Mastrandrea PR,
white LL (eds) Climate change 2014: impacts, adaptation, and vul-
nerability, part B:regional aspects contribution of Working Group II
to the fifth assessment report of the Intergovernmental Panel on
Climate Change. IPCC, Geneva, pp 1199–1265
Peeler KA, Opsahl SP, Chanton JP (2006) Tracking anthropogenic inputs
using caffeine, indicator bacteria, and nutrients in rural freshwater
and urban marine systems. Environ Sci Technol 40:7616–7622.
https://doi.org/10.1021/es061213c
Post DA, Vaze J, Teng J, Crosbie R, Marvanek S, Wang B, Mpelasoka F,
Renzullo L (2012) Impacts of climate change on water availability
in Botswana. CSIRO: Water for a Health Country Flagship. https://
publications.csiro.au/rpr/download?pid=csiro:EP121349&dsid=
DS4. Accessed November 2018
Pryor SC, Kunkel KE, Schoof JT (2009) Did precipitation regimes
change during the Twentieth Century? In: Pryor SC (ed)
Understanding climate change: climate variability, predictability
and change in the midwestern United States. Indiana University
Press, Bloomington, IN
R Core Team (2017) R: a language and environment for statistical com-
puting. R Core Team, Vienna, Austria. http://www.R-project.org.
Accessed November 2018
Reh R, Licha T, Geyer T, Nodler K, Sauter M (2013) Occurrence and
spatial distribution of organic micro-pollutants in a complex
hydrogeological karst system during low flow and high flow pe-
riods, results of a two-year study. Sci Total Environ 443:438–445.
https://doi.org/10.1016/j.scitotenv.2012.11.005
Robertson GP, Groffman PM (2015) Nitrogen transformations. In: Paul
EA (ed) Soil microbiology, ecology and biochemistry, 4th edn.
Elsevier, Amsterdam, pp 421–426
Rodríguez-Gil JL, Cáceres N, Dafouz R, Valcárcel Y (2018) Caffeine and
paraxanthine in aquatic systems: global exposure distributions and
probabilistic risk assessment. Sci Total Environ 612:1058–1071.
https://doi.org/10.1016/j.scitotenv.2017.08.066
Schlesinger WH, Bernhardt ES (2013) Biogeochemistry: an analysis of
global change, 3rd edn. Academic, Waltham, MA
Schullehner J, Hansen B, Thygesen M, Pedersen Carsten B, Sigsgaard T
(2018) Nitrate in drinking water and colorectal cancer risk: a nation-
wide population-based cohort study. Int J Cancer 143:73–79. https://
doi.org/10.1002/ijc.31306
Seitzinger S, Harrison JA, Böhlke JK, Bouwman AF, Lowrance R,
Peterson B, Tobias C, Drecht GV (2006) Denitrification across land-
scapes and waterscapes: a synthesis. Ecol Appl 16:2064–2090.
https://doi.org/10.1890/1051-0761(2006)016[2064:dalawa]2.0.co;2
Sivapalan M, Savenije HHG, Blöschl G (2012) Socio-hydrology: a new
science of people and water. Hydrol Process 26:1270–1276. https://
doi.org/10.1002/hyp.8426
Sorensen JPR, Lapworth DJ, Nkhuwa DCW, Stuart ME, Gooddy DC,
Bell RA, Chirwa M, Kabika J, Liemisa M, Chibesa M, Pedley S
(2015) Emerging contaminants in urban groundwater sources in
Africa. Water Res 72:51–63. https://doi.org/10.1016/j.watres.2014.
08.002
Stadler S, Talma AS, Tredoux G, Wrabel J (2012) Identification of
sources and infiltration regimes of nitrate in the semi-arid
Kalahari: regional differences and implications for groundwater
management. Water SA 38:213–224
Statistics Botswana (2013) Botswana Core Welfare Indicators Survey
2009/10. Statistics Botswana, Gaborone, Botswana. http://www.
statsbots.org.bw/sites/default/files/BCWIS%202009%2010%
20MAIN%20REPORT.pdf. Accessed November 2018
Statistics Botswana (2015a) Population census atlas 2011: Botswana.
Statistics Botswana, Gaborone, Botswana. http://www.statsbots.
org.bw/sites/default/files/publications/Census%20ATLAS.pdf.
Accessed November 2018
Statistics Botswana (2015b) Population & housing census 2011 selected
indicators: cities & towns. Statistics Botswana, Gaborone, Botswana
http://www.statsbots.org.bw. Accessed November 2018
Statistics Botswana (2015c) Population & housing census 2011 selected
indicators: South East District, Gaborone, Botswana. http://www.
statsbots.org.bw. Accessed November 2018
Statistics Botswana (2017) Botswana: causes of mortality 2013. Statistics
Botswana, Gaborone, Botswana. http://www.statsbots.org.bw/sites/
default/files/publications/Botswana%20Causes%20of%
20Mortality%202013_0.pdf. Accessed November 2018
Staudt M (2003) Environmental hydrogeology of Ramotswa, South East
District, Republic of Botswana. Botswana Department of
Geological Survey, Gabronne, Botswana. https://services.geodan.
nl/public/document/AGRC0001XXXX/api/data/
AGRC0001XXXX/mim/Staudt_2003.pdf_6m5981m9s. Accessed
November 2018
Stevenson RJ (2011) A revised framework for coupled human and natural
systems, propagating thresholds, and managing environmental
problems. Phys Chem Earth A/B/C. 36:342–351. https://doi.org/
10.1016/j.pce.2010.05.001
Stone AEC, Edmunds WM (2014) Naturally-high nitrate in unsat-
urated zone sand dunes above the Stampriet Basin, Namibia.
J Arid Environ 105:41–51. https://doi.org/10.1016/j.jaridenv.
2014.02.015
Stuart ME, Lapworth DJ, Thomas J, Edwards L (2014) Fingerprinting
groundwater pollution in catchments with contrasting contaminant
sources using microorganic compounds. Sci Total Environ 468–
469:564–577. https://doi.org/10.1016/j.scitotenv.2013.08.042
Taylor RG, Koussis AD, Tindimugaya C (2009) Groundwater and cli-
mate in Africa: a review. Hydrol Sci J 54:655–664. https://doi.org/
10.1623/hysj.54.4.655
Taylor RG, Scanlon B, Doll P, Rodell M, van Beek R, Wada Y,
Longuevergne L, Leblanc M, Famiglietti JS, Edmunds M,
Konikow L, Green TR, Chen J, Taniguchi M, Bierkens MFP,
Hydrogeol J

MacDonald A, Fan Y, Maxwell RM, Yechieli Y, Gurdak JJ, Allen
DM, Shamsudduha M, Hiscock K, Yeh PJF, Holman I, Treidel H
(2013) Ground water and climate change. Nature Clim Change 3:
322–329
Tompkins JA, Smith SR, Cartmell E, Wheater HS (2001) In-situ biore-
mediation is a viable option for denitrificatin of chalk groundwaters.
Q J Eng Geol Hydrogeol 34:111–125. https://doi.org/10.1144/qjegh.
34.1.111
Tredoux G, Talma AS (2006) Nitrate pollution of groundwater in south-
ern Africa. In: Xu Y, Usher BH (eds) Groundwater pollution in
Africa. CRC, Boca Raton, FL
USEPA (2013) Introduction to in situ bioremediation of groundwater.
USEPA, Washington, DC. https://www.epa.gov/sites/production/
files/2015-04/documents/introductiontoinsitubioremediation
ofgroundwater_dec2013.pdf. Accessed November 2018
USGS (2006) Collection of water samples. In: National field manual for
the collection of water-quality data book 9, 2 edn. US Geological
Survey, Reston, VA
Villholth KG (2013) Groundwater irrigation for smallholders in sub-
Saharan Africa: a synthesis of current knowledge to guide sustain-
able outcomes. Water Int 38:369–391. https://doi.org/10.1080/
02508060.2013.821644
Walmsley B, Patel S (2011) Botswana handbook on environmental as-
sessment legislation in the SADC region, 3 edn., chap 4.
Development Bank of Southern Africa andSouthern African
Institute for Environmental Assessment (SAIEA)Midrand, South
Africa
Wanda EMM, Nyoni H, Mamba BB, Msagati TAM (2017)
Occurrence of emerging micropollutants in water systems in
Gauteng, Mpumalanga, and North West Provinces, South
Africa. Int J Environ Res Publ Heal 14. https://doi.org/10.3390/
ijerph14010079
Waso M, Ndlovu T, Dobrowsky PH, Khan S, Khan W (2016) Presence of
microbial and chemical source tracking markers in roof-harvested
rainwater and catchment systems for the detection of fecal contam-
ination. Environ Sci Pollut Res 23:16987–17001. https://doi.org/10.
1007/s11356-016-6895-7
Weiss RF (1974) Carbon dioxide in water and seawater: the solubility of a
non-ideal gas. Mar Chem 2:203–215. https://doi.org/10.1016/0304-
4203(74)90015-2
Weiss RF, Price BA (1980) Nitrous oxide solubility in water and seawa-
ter. Mar Chem 8:347–359
Wickham H (2009) ggplot2: elegant graphics for data analysis. Springer,
New York. http://ggplot2.org. Accessed November 2018
Hydrogeol J