Nature Climate Change (2012) doi:10.1038/nclimate1731
1
Evidence of the dependence of groundwater resources on extreme rainfall in
East Africa
Richard G. Taylor
1
, Martin C. Todd
2
, Lister Kongola
3
, Louise Maurice
4
, Emmanuel
Nahozya
3
,Hosea Sanga
3
and Alan M. MacDonald
4
1) Department of Geography, University College London, UK
2) Department of Geography, University of Sussex, UK
3) Ministry of Water and Irrigation, Tanzania
4) British Geological Survey, UK
Groundwater recharge sustains the groundwater resources on which there is
global dependence for drinking water and irrigated agriculture
1
. For many
communities, groundwater is the only perennial source of water. Here, we
present a newly compiled 55-year record of groundwater-level observations in
an aquifer of central Tanzania that reveals the highly episodic occurrence of
recharge resulting from anomalously intense seasonal rainfall. Episodic
recharge interrupts multiannual recessions in groundwater levels, maintaining
the water security of the groundwater-dependent communities in this region.
This long-term record of groundwater storage changes in the semi-arid tropics
demonstrates a nonlinear relationship between rainfall and recharge wherein
intense seasonal rainfall associated with the El Niño Southern Oscillation and
the Indian Ocean Dipole mode of climate variability
2,3
contributes
disproportionately to recharge. Analysis of the Intergovernmental Panel on
Climate Change AR4 and AR5 multi-model ensembles for the twenty-first
century indicates that projected increases in extreme monthly rainfall,
responsible for observed recharge, are of much greater magnitude than
changes to mean rainfall. Increased use of groundwater may therefore prove a
potentially viable adaptation to enhanced variability in surface-water resources
and soil moisture resulting from climate change
4–7
. Uncertainty in the
projected behaviour of the El Niño Southern Oscillation and associated
teleconnections remains, however, high
8
.
Groundwater is the world's largest accessible store of fresh water and supplies 36%
of the world's drinking water and ~42% of the water used for irrigation
1
. Groundwater
is the only reliable source of fresh water in many semi-arid and arid regions where
surface waters are seasonally or perennially absent
9
. The long-term viability of
groundwater resources as well as the ecosystems and livelihoods that they sustain,
depends on replenishment of groundwater by recharge. Over the past 50 years,
groundwater depletion has been estimated and observed in several aquifers
throughout the tropics and sub-tropics
10-13
. Such depletion not only threatens
ecosystem function and the livelihoods of groundwater-dependent communities in
some of the world's poorest regions but is also estimated to contribute to sea-level
rise
12,13
. A conceptual understanding of the relationship between rainfall and
recharge is fundamental to the development of robust estimates and projections of
not only groundwater recharge and depletion but of all components of the terrestrial
water balance under changing climates and increasing freshwater demand.
Nature Climate Change (2012) doi:10.1038/nclimate1731
2
Recharge results from effective precipitation (that is, precipitation minus losses from
evapotranspiration) infiltrating the subsurface where hydraulic gradients are
downward. Diffuse groundwater recharge occurs directly through the soil matrix in
saturated soils and through soil macropores and fractures that bypass the soil
matrix. Focused groundwater recharge takes place indirectly by way of leakage from
runoff and surface-water sources including ephemeral streams and is often a critical
source of recharge in semi-arid environments
14-16
. The magnitude of effective
precipitation is highly sensitive to changes in precipitation and evapotranspiration,
particularly in semi-arid environments where differences between these fluxes are
small
16,17
. Soil-moisture balance modelling studies in the tropics
18-20
suggest a
nonlinear relationship between rainfall and recharge in which recharge is biased to
heavy rainfall events (>10mmd
-1
) that temporarily exceed high rates of prevailing
evapotranspiration. A key uncertainty is whether soil infiltration capacities are able to
transmit, in practice, modelled increases in recharge generated by heavy rainfall.
Indeed, the relationship between precipitation and groundwater recharge remains
poorly resolved in many regions owing to a lack of long-term observational data.
Here we present empirical evidence of the relationship between rainfall and
groundwater recharge in semi-arid tropical East Africa from a recently compiled
near-continuous, 55-year (1955-2010) record of coincidental, in situ groundwater-
level observations (variable time step with gaps) and monthly rainfall (Fig. 1a,b).
Observations derive from the Makutapora Wellfield (38º 45’ E, 5º, 55’ S) in central
Tanzania where groundwater is abstracted from an aquifer comprising deeply
weathered granite overlain by alluvium. This unique time series, the longest
observed record yet published for any location in the tropics, reveals the highly
episodic nature of recharge events indicated by positive deflections in groundwater
levels that result from anomalously intense rainfall during the austral summer
monsoon (November-April). These recharge events interrupt multiannual recessions
in groundwater levels. Rates of groundwater level decline have increased
substantially from ~0:5 m yr
-1
(1955-1979) to ~1:7 m yr
-1
since 1990. This change is
a response to pronounced increases in monthly groundwater abstraction from 0.1 to
0.9 million m
3
to supply potable water to the national capital, Dodoma (Fig. 1c).
Intensive groundwater abstraction is sustained by natural, inter-annual groundwater
storage that is replenished on a decadal timescale by episodic recharge.
The observed relationship between seasonal rainfall and groundwater recharge is
nonlinear (Fig. 2a) as recharge is largely restricted to anomalously intense seasonal
rainfall. The cumulative recharge distribution (Fig. 2b) shows that the top 7 (11)
seasons of rainfall account for 60% (75%) of the total recharge observed over 55
years from 1955 to 2010; remaining recharge is confined to seasons that feature
individual months of statistically extreme (>95th percentile) rainfall (Fig. 2a). For
nearly two-thirds of the 55-year record, no recharge is observed. Indeed, the
Makutapora record suggests that unless monthly rainfall exceeds 200mm (>95th
percentile) or seasonal rainfall is greater than 670mm (third quartile), little or no
recharge occurs. These observed thresholds reflect the requirement of intense
rainfall to overcome the high rates of potential evapotranspiration that prevail in the
tropics, estimated locally to be 160 mm month
-1
during the monsoon season, to
generate recharge. Recharge pathways to the Makutapora Wellfield are both diffuse,
through surficial sediments within the wellfield depression, and focused by way of
Nature Climate Change (2012) doi:10.1038/nclimate1731
3
ephemeral streams flowing over the coarse-grained soils within alluvial fans at the
margins of the depression
21
(see Supplementary Information).
We examine anomalously intense seasonal rainfall that generates recharge in
central Tanzania in terms of the wider regional and global climate system.
Composite analysis of regional-scale rainfall anomalies associated with the seven
largest episodic groundwater recharge events indicates a marked north-south dipole
pattern of precipitation over tropical southeast Africa (Fig. 3a) with opposing positive
(negative) precipitation anomalies north (south) of ~10 S. This dipole pattern is
congruent with the most important structure of rainfall variability across southeast
Africa as defined by the leading empirical orthogonal function (EOF) of monsoon
season rainfall
2
(see Supplementary Information). The leading EOF is itself strongly
correlated with tropical sea-surface temperature anomalies (SSTAs) indicative of
both the Indian Ocean Dipole (IOD) and the El Niño Southern Oscillation (ENSO),
the dominant modes of coupled ocean-atmosphere interaction in the tropical global
and Indian oceans, respectively
2
. Figure 3b clearly shows the close association
among the time series of the leading EOF, ENSO and groundwater recharge events.
Of the seven largest recharge events, all but one are in the top eight events of the
EOF time series and five coincide with El Niño events. The other groundwater
recharge event (1959-1960) is associated with locally high rainfall (Fig. 1b) that does
not have a strong regional expression.
The complex interaction of ENSO teleconnections and IOD variability is known to be
the key driver of climate variability over southeast Africa
2,3
. Major ENSO warm (El
Niño) events and the positive phase of the IOD lead to wet extremes in the East
African sector and our study region. The most striking example is that the greatest
recharge event (521mm) observed in the Makutapora record (Fig. 1a) resulted from
the heaviest season of monsoonal rainfall (1997-1998) recorded. This event is
associated not only with the strongest ENSO warm event of the past century but also
a positive IOD event
22
. From the above analysis, we conclude that the infrequent and
episodic groundwater recharge events at Makutapora are primarily driven by
regional-scale extreme precipitation anomalies associated with major events of the
dominant modes of tropical climate variability in the region. Some of the more minor
recharge events are associated with more localized rainfall anomalies (see
Supplementary Information). It is unclear at present whether climate change will
strengthen or weaken the influence of ENSO on East African rainfall
8
. An increase in
the probability of positive IOD modes associated with heavy monsoonal rainfall in
East Africa has recently been suggested as a response to anthropogenic warming
from a review of AR4 models
23
. At present, the complex interactions of ENSO and
IOD and their teleconnections preclude, however, firm conclusions on the impact of
global warming on ENSO and IOD modes of variability and their influence on heavy
rainfall in central Tanzania. This uncertainty represents a key question to be
investigated using new output from AR5 models.
A robust signal of projected global warming is an increase in the intensity of heavy
rainfall events. This intensification is expected to be especially pronounced in tropical
wet seasons as a result of the ~6:5% K
-1
increase in atmospheric humidity defined
by the Clausius-Clapeyron relation and the sensitivity of tropical convective rainfall to
total moisture content
4,5,23
, verified by observational studies
6
. Analysis of general
circulation model (GCM) projections for the twenty-first century over the region
Nature Climate Change (2012) doi:10.1038/nclimate1731
4
surrounding the study site (Fig. 4) suggests an increase in mean precipitation over
the study region, associated with projected increases over equatorial East Africa
more widely (see Supplementary Information). This result is similar in the analysis of
GCMs contributing to the Intergovernmental Panel on Climate Change AR4 and
forthcoming AR5 (+9.7 and +5.2 mm d
-1
by the end of the twenty-first century,
respectively), although the uncertainty is higher in the latter case, reflecting some
important regional differences (see Supplementary Information). At the broader
scale, this is part of a wider quasi-global rich-get-richer pattern in which regions of
moisture convergence (divergence) are expected to experience increased
(decreased) precipitation
6
, consistent between the AR4 and AR5 models. However,
of particular importance to this study is that projected changes to extreme monthly
rainfall driving groundwater recharge observed in the Makutapora record are of much
greater magnitude (+22.5 and +25.4 mm month
-1
for AR4 and AR5 GCMs,
respectively) than changes projected for mean monthly rainfall. These changes in
the higher moments of the rainfall distribution are an important dimension to non-
stationarity in future climate and, as shown here, have important implications for
groundwater processes.
Anomalously intense seasonal and monthly rainfall has been associated with
negative socio-economic consequences
3
that include the loss of crops and livestock,
and the destruction of homes, yet the Makutapora record shows that these episodic
events sustain groundwater resources on which there is often complete dependence
for freshwater in tropical semi-arid environments. The observed dependence of
episodic groundwater recharge on intense rainfall is consistent with evidence from
semi-arid areas of Australia
24
, southwestern USA
14
and West Africa
15
. The projected
shift towards more intensive monthly rainfall favouring groundwater recharge
suggests that greater use of groundwater may form a viable adaptation to increased
variability in surface-water resources and soil moisture resulting from climate
change. In light of the observed dependence of groundwater recharge on ENSO and
IOD, the limited ability of GCMs to represent these modes of climate variability and
their teleconnections remains a key impediment to understanding climate-change
impacts on freshwater supplies in East Africa and regional climate change scenarios
more widely.
METHOD SUMMARY
The near-continuous time series of groundwater-level measurements drawn from 6
monitoring wells over a variable time step (daily to monthly) and monthly pumping
volumes from the Makutapora Wellfield was constructed from observations collected
by the Ministry of Water and Irrigation and the Dodoma Urban Water Supply and
Sanitation Agency. Data were assembled from computer files, hardcopy plots, and
notebooks stored in Wamaruvu Basin Office of the Ministry of Water and Irrigation.
Monthly rainfall at the Makutapora Wellfield (35°45’E, 5°55’S) was monitored by the
Tanzanian Meteorological Agency; daily records are unavailable. Groundwater
recharge (q) was estimated from changes in groundwater levels (∂h) through time
(∂t) assuming changes in groundwater storage are controlled by the balance of
recharge and net groundwater drainage (D) from a monitoring well where specific
yield (S
y
) is the storage co-efficient through the equation
25
: q = S
y
(∂h/∂t)+D. The
Makutapora Wellfield resides within the large, local depression wherein recharge
occurs both directly, through the direct infiltration of rainfall and indirectly through
ephemeral streams (see Supplementary Information). D occurs both as a result of
Nature Climate Change (2012) doi:10.1038/nclimate1731
5
intensive groundwater abstraction for the city of Dodoma (Fig. 1c) and natural
discharges. D was estimated from recessionary trends in groundwater levels during
extended periods of absent recharge (q = 0). S
y
was estimated from the statistically
significant (r
2
= 0.94, p = 0.001) correlation that was observed (see Supplementary
Information) between cumulative wellfield abstraction (-Q) and groundwater-level
recession (-δh/δt) within the wellfield (A = 59 km
2
) wherein S
y
= (∂t/-∂h)(-Q/A). This
relationship assumes that groundwater is drawn from pore storage evenly from
multiple boreholes over the wellfield area. The derived value of S
y
(0.064±0.004)
applies to groundwater-level fluctuations over a depth interval (1046 to 1058 m
above mean sea level) comprising in situ weathered granite
26
and is consistent with
that recently estimated from tracing experiments in weathered crystalline rock in
Uganda
27
. Estimates of recharge during two gaps in the Makutapora Record (1960-
1965, 1980-1984) were imputed empirically from the statistically significant (R
2
=
0.82, p <10
-4
) relationship between heavy (>580 mm) seasonal rainfall and observed
recharge; aggregate values imputed for each period were validated against the
observed gap in the record (∂h/∂t) and estimated D (see Supplementary
Information).
Analysis of historical climate over the wider region employed: (i) gridded monthly
precipitation at 0.5° resolution from the Global Precipitation Climatology Centre
(GPCC) product from 1955 to 2009
28
; and (ii) gridded global sea surface temperature
anomalies (SSTAs) on a 1.0° global grid from the Hadley Centre Sea Ice and Sea
Surface Temperature data set
29
. In both cases our results are insensitive to the
choice of other available observed gridded data products (see Supplementary
Information). We apply the following statistical analyses: (i) composite analysis of
gridded rainfall (Fig. 3a) and SSTA fields (see Supplementary Information) based on
sample years of major groundwater recharge events; (ii) Empirical Orthogonal
Function analysis of monthly rainfall over the region to determine objectively the
space/time structures of rainfall variability (see Supplementary Information); (iii) time-
correlation of the resulting Empirical Orthogonal Function time series with SSTA
gridded fields to determine the association with global and regional modes of climate
variability (see Supplementary Information). Climate change projections were
obtained from the multi-model ensemble (MME) compiled under the 3rd (CMIP3) and
5th (CMIP5) Coupled Model Intercomparison Projects contributing to the
Intergovernmental Panel on Climate Change (IPCC) 4th and (forthcoming) 5th
Assessment Reports, AR4 and AR5, respectively. In total, the MME contained data
from 23 GCMs for the CMIP3 data and 21 GCMs for currently incomplete CMIP5
archive, of which 8 are Earth System Models (see Supplementary Information). We
use data from a single greenhouse gas emission scenario (SRES A1B) from the
CMIP3 collection, and two emission scenarios from the CMIP5 collection (RCP45
and RCP85). Changes in climate were calculated over a 10° box centred on the
study site for three future epochs representing the early (2021-2050), mid (2035-
2065) and late (2070-2099) twenty-first century. Here, we present only results from
analysis RCP8.5 scenario, as this is the trajectory closest to recent greenhouse gas
emissions, and for the late 21st century alone (Fig 4). The basic structure, if not the
magnitude, of the projected changes to mean and the 90th percentiles of monthly
rainfall is essentially insensitive to both the epoch and choice of RCP (see
Supplementary Information).
