Ecological Applications, 21(3), 2011, pp. 678–694
Ó 2011 by the Ecological Society of America
Water subsidies from mountains to deserts:
their role in sustaining groundwater-fed oases in a sandy landscape
E. G. JOBBA
´
GY,
1,2,6
M. D. NOSETTO,
1,2,3
P. E. VILLAGRA,
4
AND R. B. JACKSON
5
1
Grupo de Estudios Ambientales—IMASL, CONICET, San Luis, Argentina
2
Departamento de Agronomı
´
a—FICES, Universidad Nacional de San Luis, San Luis, Argentina
3
Ca
´
tedra de Climatologı
´
a Agrı
´
cola, Facultad de Ciencias Agropecuarias, Universidad Nacional de Entre Rı
´
os, Argentina
4
Instituto Argentino de Nivologı
´
a, Glaciologı
´
a y Ciencias Ambientales, CCT-CONICET, Mendoza, Argentina
5
Department of Biology and Nicholas School of the Environment and Earth Sciences, Duke University, Durham, North Carolina, USA
Abstract. In arid regions throughout the world, shallow phreatic aquifers feed natural
oases of much higher productivity than would be expected solely from local rainfall. In South
America, the presence of well-developed Prosopis flexuosa woodlands in the Monte Desert
region east of the Andes has puzzled scientists for decades. Today these woodlands provide
crucial subsistence to local populations, including descendants of the indigenous Huarpes. We
explore the vulnerability and importance of phreatic groundwater for the productivity of the
region, comparing the contributions of local rainfall to that of remote mountain recharge that
is increasingly being diverted for irrigated agriculture before it reaches the desert. We
combined deep soil coring, plant measurements, direct water-table observations, and stable-
isotopic analyses (
2
H and
18
O) of meteoric, surface, and ground waters at three study sites
across the region, comparing woodland stands, bare dunes, and surrounding shrublands. The
isotopic composition of phreatic groundwaters (d
2
H: 137% 6 5%) closely matched the
signature of water brought to the region by the Mendoza River (137% 6 6%), suggesting
that mountain-river infiltration rather than in situ rainfall deep drainage (39% 6 19%) was
the dominant mechanism of recharge. Similarly, chloride mass balances determined from deep
soil profiles (.6 m) suggested very low recharge rates. Vegetation in woodland ecosystems,
where significant groundwater discharge losses, likely . 100 mm/yr occurred, relied on
regionally derived groundwater located from 6.5 to 9.5 m underground. At these locations,
daily water-table fluctuations of ;10 mm, and stable-isotopic measurements of plant water,
indicated groundwater uptake rates of 200–300 mm/yr. Regional scaling suggests that
groundwater evapotranspiration reaches 18–42 mm/yr across the landscape, accounting for 7–
17% of the Mendoza River flow regionally. Our study highlights the reliance of ecosystem
productivity in natural oases on Andean snowmelt, which is increasingly being diverted to one
of the largest irrigated regions of the continent. Understanding the ecohydrological coupling
of mountain and desert ecosystems here and elsewhere should help managers balance
production agriculture and conservation of unique woodland ecosystems and the rural
communities that rely on them.
Key words: algarrobo woodlands; arid oasis; central Monte Desert, western Argentina; Cordillera de los
Andes; groundwater recharge/discharge; Larrea shrublands; Mendoza River; Monte Desert; phreatophytes;
Prosopis flexuosa woodlands; Telteca Provincial Reserve (Argentina).
INTRODUCTION
Shallow groundwater can sustain the productivity of
arid ecosystems at much higher levels than would be
expected solely from local rainfall (Freeze and Cherry
1979, Eamus et al. 2006, Jackson et al. 2009).
Groundwater-fed upland ecosystems are important
hotspots of biodiversity and economic activity, including
those in the deserts of Australia, North America, and
Asia (e.g., Devitt et al. 2002, Eamus et al. 2006,
Sanderson and Cooper 2008). Understanding the
sources of recharge that maintain such ecosystems is of
critical importance to the diverse people that depend on
them, typically rural communities adapted to these
particularly harsh environments. For instance, the
relatively high-productivity woodlands of the Monte
Desert have provided local peoples with crucial subsis-
tence for centuries (Monta
˜
na et al. 2005, Torres 2008).
However, the source of the water that maintains these
ecosystems is uncertain. Here we explore the reliance of
these woodlands on their underlying aquifers, identify-
ing their sources of recharge and quantifying their
overall contribution to the regional water balance of the
desert.
Manuscript received 5 August 2009; revised 9 June 2010;
accepted 9 June 2010. Corresponding Editor: D. S. Schimel.
6
Present address: Grupo de Estudios Ambientales,
IMASL—UNSL, Eje
´
rcito de los Andes 950 -5700, San Luis,
Argentina. E-mail: jobbagy@unsl.edu.ar
678
Shallow aquifers can be sustained by local or distant
recharge sources. In most arid regions, evapotranspira-
tion recycles essentially all precipitation inputs back to
the atmosphere, resulting in negligible recharge locally
(Scanlon et al. 2006). Important exceptions occur in
locations that have sandy or rocky soils, such as sand
dunes and fractured rock outcrops, low or degraded
vegetation cover, highly seasonal and intense precipita-
tion regimes, or extensive lateral flow or run-on; in such
cases at least some deep drainage into the saturated zone
will eventually occur (Scanlon and Goldsmith 1997,
Athavale et al. 1998, Seyfried et al. 2005, Small 2005,
Gates et al. 2008). More-distant sources of recharge are
particularly significant in arid regions located down-
stream of water-yielding mountains (Wilson and Guan
2004). For example, the occurrence of shallow water
tables, wetlands, and lakes in the lowest positions of
many sand dune landscapes has been attributed to
mountain snowmelt in locations such as the Great Sand
Dunes of Colorado (USA) (Wurster et al. 2003) and the
Barain Jaram and Taklamakan deserts in China
(Thomas et al. 2000, Chen et al. 2004, Gates et al. 2008).
From a water-balance perspective, locally derived
groundwater that becomes accessible to plants does not
represent a new net water contribution but rather an
opportunity to consume unused rainfall redistributed in
space (such as from dune tops to interdunes) or time
(from wet years to dry years). In contrast, more-distant
sources of recharge can increase the overall productivity
of a landscape as well as its vulnerability to hydrologic,
climatic, and ecological change upstream. In sandy
deserts flanked by high mountain ranges, local rainfall
inputs and remote water subsidies can both be
significant sources of groundwater recharge, but remote
water sources are more vulnerable to human diversion
and use through irrigation (Milner et al. 2009).
Groundwat er use by plants typically declines as
water-table levels drop, both in space along topographic
gradients (Zencich et al. 2002, Gries et al. 2003, Nosetto
et al. 2009), and through time, such as seasonal shifts in
water-table depth (Stromberg et al. 1992, Naumburg et
al. 2005, Cooper et al. 2006). Although shrubs and tree
species can have maximum rooting depth of many
meters (Canadell et al. 1996, Schenk and Jackson 2002),
and some observations suggest groundwater uptake
below 20 m of depth (Haase et al. 1996, Gries et al.
2003), we are unaware of significant groundwater supply
to desert ecosystems with water tables that are .10 m
deep (Nichols 1994, Zencich et al. 2002). Most plants
show a dynamic and facultative reliance on groundwater
according to rainfall variability, with soil moisture
preferred over groundwater when available (Engel et
al. 2005). Knowledge of the groundwater reliance of
plant species and, more importantly, whole ecosystems,
should be useful in shaping the development of
management strategies for arid oases.
The central Monte Desert occupies a vast territory
east of the Andes Cordillera in Argentina. Although
sparsely covered Larrea shrublands dominate the
widespread sand dune landscapes, interdune areas often
host Prosopis woodland ecosystems of high biological
and economic value (Rundel et al. 2007). The presence
of these woodlands appears to be spatially associated
with shallow (,10 m) water tables (Gonzalez Loyarte et
al. 2000), yet their degree of reliance on groundwater
and its recharge sources is unknown. Local recharge
may occur because soils are sandy, rainfall inputs are
intense and are concentrated in summer, and the
vegetation has been subject to overgrazing and logging
for almost a century (Abraham et al. 2009). Remote
recharge sources, however, could also be important,
since several rivers carry snowmelt water to the region
from high-elevation watersheds in the Andes.
In this paper we explore the interaction of ecosystems
and groundwater in the central Monte Desert, examin-
ing the source of phreatic groundwater and its uptake by
plants. Six questions guided our work at regional,
landscape, and local scales. At regional scales we asked,
(1) What are the contributions of local rainfall and
mountain snowmelt to recharge? and (2) How much
groundwater is discharged through evapotranspiration?
At the landscape scale, we asked (3) What are the
patterns of vertical water exchange between ecosystems
and phreatic aquifers across the sand dunes landscapes?
and (4) How do these patterns change across topo-
graphic positions and plant cover? Finally, more locally
we asked (5) How stable are water-table depths across
seasons and years? and (6) How does groundwater
reliance vary among plant species and seasons? To
address these questions we combined stable-isotopic
analyses, plant measurements, soil and sediment coring,
and direct measurements of water-table depth and
fluctuations. At the regional scale we investigated the
sources of groundwater recharge by characterizing the
stable-isotopic composition of water from rainfall,
snowmelt-fed rivers, phreatic groundwater, and unsatu-
rated soil, achieving a regional estimate of groundwater
evapotranspiration by scaling up a suite of detailed
observations. At the landscape scale we evaluated the
direction and magnitude of vertical water exchange by
measuring moist ure contents, isot opic compositions,
and chloride concentrations in soil and sediment profiles
across multiple topographic positions. Finally, at the
local stand scale we monitored water-table levels and
used stable isotopes to assess the reliance of different
plant species on groundwater uptake throughout the
year.
M
ATERIALS AND METHODS
Study region and sites
The central Monte is a temperate desert that spans
;120 000 km
2
of western Argentina. Its vast sandy
aeolian plains host shrublands dominated by Larrea
divaricata. In s ome areas, however, low la ndscape
positions are occupied by larger trees dominated by
Prosopis flexuosa. Located next to one of the highest
April 2011 679MOUNTAIN WATER SUBSIDIES TO DESERTS
sections of the Andean Cordillera, the central Monte
receives large inflows of snowmelt from the San Juan,
Mendoza, and Tunuya
´
n Rivers (mean annual flow for
1954–2004: 58, 45, and 29 m
3
/s, respectively;
Subsecretaria de Recursos Hı
´
dricos 2004). These three
rivers also sustain one of the largest irrigated agricul-
tural regions in South America, ;7000 km
2
of vineyards
and olive orchards (Fig. 1).
Our study was carried out in the Telteca Provincial
Reserve and surrounding areas, an aeolian sandy plain
near the lower Mendoza river and downstream of its
major irrigated area (Fig. 1a). Transverse dunes oriented
NNW–SSE and discontinuous valleys create elevation
gradients of 8–20 m between interdune lowlands and
dune crests. Dating of sands close to the study area
indicate a late Quaternary origin for these dune fields
with several stabilization and remobilization periods in
the last five millennia (Tripaldi and Forman 2007). The
water table in the lowlands is located 6–15 m below the
surface and no permanent surface water is present in the
dune territory. Soils are poorly developed Entisols with
.95% sand. In spite of the current water-table depths,
some lowlands show an accumulation of evaporites
(calcite and gypsum) and, in a few cases, small shell
fragments that suggest periods of higher water tables or
even surface-water availability in the past. At the
bottom of some lowlands, an accumulation of fine-
textured materials, mostly evaporites, is observed at the
surface but not at depth. These fine materials restrict
infiltration and support the formation of ephemeral
ponds in the lowest positions after large rain events.
Current precipitation at the head quarters of the
Telteca Provincial Reserve is 156 mm/yr (1972–2007
average), .80% of which takes place in the austral
summer and fall between October and March. Penman-
Monteith potential evapotranspiration approaches 1300
mm/yr, and the aridity index (precipitation/potential
evapotranspiration) is 0.12 (1960–1990, Climate
Research Unit database; New et al. 2002). Mean annual
temperature is 18.58C, with minimum and maximum
temperatures reaching a few degrees below zero in
winter to at least 478C in summer.
High and intermediate landscape positions are dom-
inated by Larrea divaricata, Tricomaria usillo, Bulnesia
FIG. 1. The study region. (a) Regional map showing the watersheds of the San Juan, Mendoza, and Tunuya
´
n Rivers from their
origin in the high Andes to their convergence into the Desaguadero. Three of the largest irrigated areas of South America are fed by
these rivers on the foot slope of the Andes (gray areas). Our research focused on the sand dune territory located downstream of the
lower Mendoza River (white line). (b) A detailed view of our study region based on high-resolution Google Earth imagery indicates
the location of study sites. Hand-dug wells are built by local people and are usually associated with homesteads. Precipitation
amount and chemistry were recorded at the headquarters of the Telteca Reserve. ‘‘Altos Limpios’’ is a bare-dune area with no
vegetation. A currently dry riverbed with SW–NE direction, corresponding to the ancient and abandoned trajectory of the
Mendoza River can be seen in the image.
E. G. JOBBA
´
GY ET AL.680
Ecological Applications
Vol. 21, No. 3
retama,andAtriplex lampa accompanied by small
individuals (height ,3m)ofProsopis flexuosa and a
sparse layer of the grass Panicum urvilleanum. Lowlands
with a typical area of 1–10 ha are generally covered by
denser and taller formations of the tree species Prosopis
flexuosa, Bulnesia retama,andGeoffrea decorticans
(height, 5–10 m) (Villagra et al. 2005, Alvarez et al.
2006) accompanied by most of the species found in
higher topographic positions (Gonzalez Loyarte et al.
2000).
Local settlers live in homesteads or ‘‘puestos’’ located
in interdune positions and based around hand-dug wells
that reach the water table and supply both people and
livestock with water. The area around a given puesto is
grazed by goats and, to a lesser extent, cattle. More than
80 locations with single or clustered puestos are found
across the sand dune territory of the study region
(Torres 2008). Large Prosopis trees growing in the
lowlands not only supply goats with a major forage
source, including seed pods and foliage, but also supply
people with timber for construction and well casings as
well as food for people, including flour from seeds, and
syrup and beverages from seedpod carpels. Although the
sandy landscape is stable and the dunes are stabilized by
the current vegetation, a few areas of bare dunes reach
up to 50 ha in size, as in the Altos Limpios area (Fig.
1b).
Our observations were focused on three study sites of
vegetated dunes (Fig. 1b). Sites A and B have had less
grazing pressure since the creation of the reserve in 1986,
whereas site C has been subject to more intensive
grazing that is typical for the region. Site A also included
a bare dune zone (Altos Limpios) without plants to
compare to the adjacent stable, vegetated landscape. At
each site we sampled soil profiles along transects
covering full toposequences. We sampled soil and
sediments extensively to the water table at lowland
positions, using the resulting boreholes for longer-term
monitoring of groundwater depth and chemistry at each
location. We complemented our regional assessment of
the stable-isotopic composition of rainfall and surface
waters at each site with a more detailed subset of data
for soils at study sites A and C and for plants at study
site C.
Stable-isotope survey
To explore the different contributions of local rainfall
and distant Andean snowmelt to groundwater recharge,
and to trace its consumption by natural vegetation, we
analyzed the stable-isotopic composition of rainfall,
rivers, phreatic groundwater, ephemeral rain-fed ponds,
soil moisture, and plant xylem water between May 2005
and May 2008. The lowest sections of the Mendoza and
San Juan Rivers were sampled on five occasions of
contrasting flow conditions, just upstream of the study
region (Fig. 1a). Phreatic groundwater was sampled
once at four hand-dug wells 6.5 to 7.5 m below the
surface and repeatedly at the four boreholes that we
established for this study. Water from 10 rainfall events
was collected at the headquarters of the Telteca
Provincial Reserve, and these data were supplemented
with 54 additional rainfall events (November 1982 to
November 1999) from the GNIP database (IAEA/
WMO 2006) for the city of Mendoza, 90 km southwest
of our study region. We sampled two ephemeral ponds
in lowlands one day after a large rainfall event in
January 2007. All water samples were filtered (0.45 lm)
and sealed in vials for analysis. Rainfall collected at the
Telteca Reserve was also analyzed for chloride concen-
trations using ion chromatography to help determine Cl
inputs for our recharge estimates using soil and sediment
chloride profiles.
Soil and sediment samples from lowland and upland
boreholes at study sites A and C were saved for isotopic
analysis. At study site C we sampled plant xylem sap
water in two seasons. Early sampling took place shortly
before Prosopis trees started to shed their leaves at the
end of a relatively wet growing season ca. 25 May 2005
(hereafter called the ‘‘wet season’’). The second sampling
took place at the end of the dry season before the onset
of rain on 3 November 2006; although the rainy period
had not started, the first new leaves of Prosopis trees
were already fully expanded. At each sampling date and
topographic position we sampled five individuals of the
target species by cutting stems 2–10 mm wide, 2 to 3
stems per individual. Both soil samples and plant stems
were immediately sealed in 10-mL vials. Water from
these samples was extracted using an azeotropic
distillation procedure (Ehleringer et al. 1991). All water
samples were analyzed at the Duke University
Environmental Stable Isotope labor atory (Durha m,
North Carolina, USA) using a Finnigan MAT Delta
Plus XL continuous flow mass spectrometer. All isotopic
values are reported as per mil delta ( d%) relative to V-
SMOW (Vienna standard mean ocean water).
Calibration against NIST / IAEA reference materials
V-SMOW and SLAP (standard light Antarctic precip-
itation) and two or more internal standards w ere
performed. Typical precision is approximately 61.5%
and 0.1% at one SD for
2
H and
18
O, respectively.
Soil and sediment sampling and estimates
of water transport
To evaluate ecosystem–aquifer water exchange at our
three study sites, we obtained soil moisture and chloride
profiles. We established transects across two contiguous
dune crests covering a full toposequence of seven
positions (e ast crest, east mids lope, east footslope,
bottom, west footslope, west midslope, west crest)
sampled down to 2.5 m of depth. For a more complete
vertical description of soil and sediment profiles, we also
sampled lowlands down to the water table at 6–9.5 m of
depth and dune tops down to 5 m of depth. In all cases
we used hand augers (10 cm in diameter) and applied a
PVC casing when needed to avoid borehole clogging by
collapsed sand. Full samples from 50-cm-depth intervals
April 2011 681MOUNTAIN WATER SUBSIDIES TO DESERTS
were immediately mixed in the field, subsampled and
stored in double plastic bags for moisture and chloride
analysis as well as in sealed vials for isotopic analysis.
Moisture content was determined gravimetrically one to
four days after sampling (oven drying method, Gardner
1986). Chloride concentrations were measured on 1:2
soil-to-water extracts using a solid-state ion-selective
electrode (Frankenberger et al. 1996) and a five-point
calibration scheme that included additional references
and spikes. Standards calibrated with more precise ion-
chromatography equipment suggested a detection limit
of 0.3 ppm and typical precision ranging from ;5% to
2% of the readings as concentrations became higher. Soil
texture was characterized for five 50-cm-deep intervals
distributed throughout each profile using the Bouyocus
method (Gee and Bauder 1986), complemented with
sieving for the separation of sand particles. Soils were
relatively homogeneous within and among soil profiles
with the exception of topsoil samples (0–50 cm deep) in
vegetated lowlands, which had clay þsilt contents of 7–
10%; all the rest of the analyzed samples had clay þ silt
contents ,5%, with the sand fraction being dominant
(80–99%, primarily fine sand particles 0.05–0.125 mm in
size).
Chloride profiles provide qualitative information on
the dominant vertical direction of ecosystem–ground-
water exchange. In addition more quantitative estimates
of recharge and discharge rates can be attempted if
atmospheric chloride inputs can be constrained. For this
purpose we approximated atmospheric chloride inputs
based on our rainfall concentration measurements (n ¼
10, mean ¼ 0.35 mg/L), long-term average rainfall rates
for the site (150 L
m
2
yr
1
¼ mm/yr) and a twofold
magnification factor applied to accounted for dry
deposition inputs (Scanlon et al. 2005), which yielded
a figure of 105 mg Cl
m
2
yr
1
. Although our Cl
deposition estimates have large uncertainty, they were
useful to approximate the order of magnitude of
recharge rates and, more i mportantly, the relative
differences between landscape positions.
Recharge fluxes in upland areas were approximated
using chloride concentrations in the vadose zone based
on the assumption that (a) chloride transport is well
approximated by piston flow, (b) all inputs to the
ecosystem originate from atmospheric deposition, with
rock-weathering supply being negligible, and (c) plant
uptake and storage in biomass and organic matter were
negligible components of the chloride balance (Allison et
al. 1985). This conservati ve behavior enables the
calculation of the residual moisture flux (Phillips 1994):
J
R
¼ D
cl
3 C
1
cl
3 1000 ðmm=mÞð1Þ
where J
R
is the net downward residual flux at the depth
of measurement (mm/yr), D
cl
is the Cl
deposition rate
(g
m
2
yr
1
), and C
cl
is the measured Cl
concentration
in the soil water (g/m
3
). The value of C
cl
was determined
by plotting cumulative Cl
content (mass Cl
per unit
volume of soil) with depth against cumulative water
content (volume water per unit volume soil) at the same
depths. In all cases straight-line segments were found
once the top meter was discarded. The slope of these
plots was used as the C
cl
value of Eq. 1 (Phillips 1994).
We also calculated the residence time of the soil water
Cl
by dividing Cl
storage down to the depth of interest
by the annual Cl
deposition rates according to Phillips
(1994):
t
z
¼
Z
ð0zÞ
hC
cl
3 dz=D
cl
ð2Þ
where t
z
is the transport time of the soil-water Cl
to
depth z (yr), and h is the volumetric water content (m
3
/
m), calculated at our sites based on gravimetric moisture
concentrations (g/g) adjusted by bulk density (g/m
3
)
measurements at each sampling interval. In lowland
sites, where the water table is closer to the surface and
evapotranspirative discharge can take place, chloride
accumulationinthevadosezonecanexceedthe
amounts attributable to atmospheric deposition, thanks
to the contributions from groundwater. In these
situations Eq. 2 can yield extremely long turnover rates
that exceed the age of sediments, providing evidence of
net groundwater discharge (Jobba
´
gy and Jackson 2007).
Groundwater monitoring
The boreholes that we maintained after our deep soil
and sediment sampling at vegetated lowlands in study
sites A, B, and C and at the bare lowland position in
study site A were permanently cased with PVC pipes and
used to monitor groundwater depth and chemistry. Well
casing was introduced ;0.5 m below the water table into
the saturated zone. Water-table depth from the surface
in each well was monitored manually at 1–4 month
intervals for at least two full growing seasons. In
addition, we installed automated pressure transducers
with built-in dataloggers (HOBO water level logger;
Onset Computer Corporation, Bourne, Massachusetts,
USA) to record groundwater level with higher frequency
(every 30 minutes) and vertical resolution (less than
60.5 cm) for several periods in different wells. In all
cases water depth was corrected after discounting
barometric pressure, which was measured with similar
sensors installed in the surface. Estimates of groundwa-
ter discharge were obtained from daily fluctuations of
water-table levels (Engel et al. 2005, Loheide et al. 2005,
Nosetto et al. 2007), often observed during the growing
season in the vegetated lowlands. To obtain daily
discharge values we used the estimates of specific yield
(groundwater discharged per unit of water-table level
decline) under fluctuating conditions proposed by
Loheide et al. (2005) for sandy soils (0.3 mm/mm). We
also sampled groundwater on five occasions with bailer
samplers after purging water stored in the well casing.
Samples were analyzed for chloride, electrical conduc-
tivity, and stable-isotope composition.
E. G. JOBBA
´
GY ET AL.682
Ecological Applications
Vol. 21, No. 3
