Content uploaded by Hugo Romero-Saltos
Author content
All content in this area was uploaded by Hugo Romero-Saltos
Content may be subject to copyright.
443
American Journal of Botany 92(3): 443–455. 2005.
R
AINFALL EXCLUSION IN AN EASTERN
A
MAZONIAN
FOREST ALTERS SOIL WATER MOVEMENT AND DEPTH
OF WATER UPTAKE
1
H
UGO
R
OMERO
-S
ALTOS
,
2,5
L
EONEL DA
S. L. S
TERNBERG
,
2,5
M
ARCELO
Z. M
OREIRA
,
3
AND
D
ANIEL
C. N
EPSTAD
4
2
Department of Biology, University of Miami, Coral Gables, Florida 33124 USA;
3
Centro de Energia Nuclear na Agricultura,
Universidade de Sa˜o Paulo, CP 96, CEP 13400-970, Piracicaba, Sa˜o Paulo, Brazil; and
4
The Woods Hole Research Center,
Woods Hole, Massachusetts 02543 USA
Deuterium-labeled water was used to study the effect of the Tapajo´s Throughfall Exclusion Experiment (TTEE) on soil moisture
movement and on depth of water uptake by trees of Coussarea racemosa,Sclerolobium chrysophyllum, and Eschweilera pedicellata.
The TTEE simulates an extended dry season in an eastern Amazonian rainforest, a plausible scenario if the El Nin˜o phenomenon
changes with climate change. The TTEE excludes 60% of the wet season throughfall from a 1-ha plot (treatment), while the control
1-ha plot receives precipitation year-round. Mean percolation rate of the label peak in the control plot was greater than in the treatment
plot during the wet season (0.75 vs. 0.07 m/mo). The rate was similar for both plots during the dry season (ca. 0.15 m/mo), indicative
that both plots have similar topsoil structure. Interestingly, the label peak in the control plot during the dry season migrated upward
an average distance of 64 cm. We show that water probably moved upward through soil pores—i.e., it did not involve roots (hydraulic
lift)—most likely because of a favorable gradient of total (matric 1gravitational) potential coupled with sufficient unsaturated hydraulic
conductivity. Water probably also moved upward in the treatment plot, but was not detectable; the label in this plot did not percolate
below 1 m or beyond the depth of plant water uptake. During the dry season, trees in the rainfall exclusion plot, regardless of species,
consistently absorbed water significantly deeper, but never below 1.5–2 m, than trees in the control plot, and therefore may represent
expected root function of this understory/subcanopy tree community during extended dry periods.
Key words: Coussarea racemosa; deuterium; drought; eastern Amazonia; El Nin˜o; Eschweilera pedicellata; global climate
change; Sclerolobium chrysophyllum.
Numerous physiological studies have described the different
strategies an individual plant uses to maintain homeostasis un-
der dry conditions. These strategies involve intricately syn-
chronized physiological mechanisms that include changes in
stomatal conductance, photosynthesis rate, sap flow, water/os-
motic potential, stem hydraulic conductivity (i.e., xylem vul-
nerability to cavitation), stem capacitance (waterstorage), veg-
etative/reproductive phenology, biomass accumulation
(growth), biomass allocation (above- and belowground), leaf/
stem tissue properties (e.g., elasticity), and chemical signaling/
regulation (e.g., accumulation of proline, abscisic acid, dehy-
drins, ubiquitins, aquaporins) (see e.g., Taiz and Zeiger, 2002).
While these mechanisms are relatively well understood in
some species, the number of studies focusing on the response
of whole communities to dry soil conditions is limited, be-
cause the feasibility of such studies depends on seasonal rain-
fall regimes or unpredictable droughts (e.g., Mulkey and
Wright, 1996; Wright, 1996; Cao, 2000; Yavitt and Wright,
1
Manuscript received 16 April 2004; revision accepted 9 November 2004.
This study was supported by grant DEB-0213011 from the U.S. National
Science Foundation to D. N., a grant from Andrew M. Foundation to L. S.,
and a Fulbright scholarship to H. R. administered by LASPAU(Amazon Basin
Scholarship program). The Instituto Brasileiro de Meio Ambiente e Recursos
Renova´veis (IBAMA) provided housing and other commodities in the field.
We thank the innumerable scientists of the TTEE, particularly David Ray and
Marisa Tohver for providing background data. Joa˜o Farias and other field
workers diligently helped to set up the experiment and collect the samples.
We immensely thank Dave Janos, Jack Fisher, John Cozza, Tara Greaver,
Nathan Muchhala, Fernando Miralles-Wilhelm, Martin Hodnett, Maria Fer-
rari, Guillermo Goldstein and two anonymous reviewers for revising earlier
versions of this manuscript.
5
Authors for correspondence (e-mail: hugo.romero@bio.miami.edu,
l.sternberg@miami.edu).
2001; Borchert et al., 2002; Engelbrecht et al., 2002; Potts,
2003). Experimental manipulations at the ecosystem level to
study the effects of prolonged drought are even scarcer (e.g.,
Engelbrecht and Kursar, 2003), in part because the logistic
effort required is immense but mainly because a significant
reason must exist to justify such an experiment. Global climate
change is one of them.
An extended soil drought in eastern Amazonia may be
caused by El Nin˜o episodes if global climatic conditions
change in the future (Nepstad et al., 2002). In 1997–1998, for
example, the strongest El Nin˜o phenomenon of the past cen-
tury was estimated, by a modeling approach, to have depleted
soil water content in the soil to a depth of at least 5 m in an
area of 1.5 million km
2
(Nepstad et al., 1999), a quarter of the
whole Amazon basin. The drought dried the soil so intensively
that surface fires—which in 50% of the cases occur by acci-
dent—extended over an area of at least 20000 km
2
, severely
damaging pastures and mature Amazonian forests (Nepstad et
al., 1999). The ecological response of the forest to these dry
periods is not yet fully understood.
The Tapajo´s Throughfall Exclusion Experiment (TTEE), de-
signed to simulate an extended drought period, is a large-scale
manipulation working toward understanding the effect of dry
conditions in the Amazon. The TTEE consists of two 1-ha
plots in evergreen tropical seasonal forest in eastern Brazil, in
which the understory of one of the plots (treatment plot) is
covered with plastic panels (installed at 1–2 m height)
throughout the wet season (Nepstad et al., 2002). This exper-
iment is elucidating ecological, physiological and geochemical
patterns at the community/ecosystem level that could occur
under persistent dry conditions (e.g., patterns in vegetative and
444 [Vol. 92A
MERICAN
J
OURNAL OF
B
OTANY
Fig. 1. Location of the Tapajo´s Throughfall Exclusion Experiment (black
star) in Brazil (2853948.480S, 5485796.840W; Nepstad et al., 2002). Map
modified from Parrotta et al. (1995) with permission.
reproductive phenology, leaf water potential, stem growth and
respiration, litterfall, leaf-area index and canopy openness,
photosynthesis, litter decomposition, biogeochemistry, and soil
moisture dynamics, among others; Nepstad et al., 2002). The
TTEE builds upon a pilot throughfall exclusion experience in
the area of Paragominas, state of Para´, Brazil, which lasted
from 1993 to 1996 (Cattaˆnio et al., 2002).
In the eastern Amazon, as elsewhere, soil water depletion
begins in the topsoil when evapotranspiration surpasses pre-
cipitation. In rainforests, the depletion is seasonal and the re-
maining water stored in deep soil layers is enough to support
trees throughout the dry season (e.g., Nepstad et al., 1994;
Jackson et al., 1995; Hodnett et al., 1996; Jipp et al., 1998;
Meinzer et al., 1999). In fact, several studies have shown that
many tropical tree species, including seedlings/saplings, phys-
iologically adapt very well to seasonal dry conditions and that
there is no significant mortality (e.g., Cao, 2000; Williamson
et al., 2000; Engelbrecht et al., 2002; Engelbrecht and Kursar,
2003; Potts, 2003). The main objective of the TTEE, however,
is to simulate a scenario of soil drought that elicits dramatic
physiological responses. The ultimate cause of such expected
water stress is the steady reduction of soil moisture, a phe-
nomenon that should initially affect patterns of water uptake.
The uptake of water by different plant species or functional
groups and the movement of pore water in the soil can be
effectively studied using isotope technology. The techniques
to study soil water movement in the unsaturated zone with
isotope tracers were developed in the 1960s when water, in
well structured soils, was demonstrated to diffuse or disperse
little vertically or horizontally, and to infiltrate the soil by lay-
ers (Zimmermann et al., 1966, 1967a, b; Blume et al., 1967;
Kline and Jordan, 1968). Concurrently with these studies, a
technique to distinguish potential sources of soil water for
plants using tritium or deuterium concentrations in the soil/
plant continuum was also developed (Gonfiantini et al., 1965;
Woods and O’Neal, 1965; Wershaw et al., 1970). However,
the usefulness of this technique was not recognized fully until
the late 1980s and early 1990s (White et al., 1985; Sternberg
and Swart, 1987; Dawson and Ehleringer, 1991; Flanagan and
Ehleringer, 1991; Walker and Richardson, 1991).
In Neotropical ecosystems, stable isotopes have proven use-
ful to understand different natural processes. For example, tri-
tium-enriched water was used near Manaus, Brazil, to study
soil water balance (Aragua´s-Aragua´s et al., 1995), and natural
abundances of deuterium were used in Central America and
the Brazilian cerrado to study spatial and temporal partitioning
of soil water by plants, probably as a result of competition
(Jackson et al., 1995, 1999; Meinzer et al., 1999, 2001). These
studies have been complemented by experiments using deu-
terium-enriched water (Moreira et al., 2000; Sternberg et al.,
2002), which in fact have become increasingly common, par-
ticularly in temperate and subtropical zones (e.g., Bishop and
Dambrine, 1995; Plamboeck et al., 1999; Turner et al., 2001;
Schwinning et al., 2002; Pen˜uelas and Filella, 2003). The ad-
dition of an isotope label is particularly useful when the nat-
ural deuterium profile in the soil is convoluted and repetitive
(i.e., with the same deuterium signature at different depths),
thereby making a clear interpretation of plant water sources
difficult (Moreira et al., 2000; Meinzer et al., 2001).
In this study, we investigated how the throughfall (i.e., rain-
fall) exclusion treatment affected (1) the percolation rate of
soil moisture movement and (2) the depth of water uptake by
understory to midcanopy trees. Exclusion of water inevitably
diminishes percolation rate, but the intensity of this reduction
in percolation was unknown. We knew nevertheless that the
top 200 cm of soil of the treatment plot contained at least 100–
150 mm less water than that of the control plot, and that deep
water reserves in the treatment plot were apparently not being
replenished (Nepstad et al., 2002; D. Nepstad et al., The
Woods Hole Research Center, unpublished data). Thesefactors
led us to postulate that the depth of water uptake by trees
shifted deeper in the treatment than in the control plots, es-
pecially in the dry season, but also in the wet season, when
the panels are installed in the treatment plot simulating a dry
season. The percolation rate of water and the depth of water
uptake was investigated by irrigating a defined area of soil
around various trees in the treatment and control plots with
deuterium-enriched water at the start and end of the 2002 wet
season. The deuterium label was followed in the soil profile
and in wood cores/stems during the wet and dry seasons of
the same year. The depth of water uptake was inferred using
a conceptual model which estimates a mean depth of water
uptake at a given time, based on the distribution of deuterium
in the soil and the deuterium concentration in a plant.
MATERIALS AND METHODS
Study site—The TTEE is located in Tapajo´s National Forest, state of Para´,
Brazil (2853948.50S, 5485796.80W, 150–200 m altitude; Nepstad et al., 2002;
Williams et al., 2002; Fig. 1). The study area is flat terrain of old-growth
upland forest, accessible by the Santare´m-Cuiaba´ road. Accordingto historical
meteorological records from the nearby city of Santare´m (1914–1981;
www.worldclimate.com), mean annual rainfall is 2061 mm. Nepstad et al.
(2002) reports an average rainfall of 2000 mm/yr, with a minimum of 600
mm and a maximum of 3000 mm, while Parrotta et al. (1995) reports an
average rainfall of 1920 mm/yr. Dry season occurs from June/July to Decem-
ber, but is more pronounced (,100 mm monthly rainfall) from August to
November (www.worldclimate.com). The mean monthly temperature is
around 25–268C (Parrotta et al., 1995; www.worldclimate.com).
March 2005] 445R
OMERO
-S
ALTOS ET AL
.—R
AINFALL EXCLUSION IN AN
A
MAZONIAN FOREST
The TTEE consists of two 1-ha plots with similar structure, diversity, flo-
ristics, biomass, topography, and soil characteristics (Nepstad et al., 2002).
The plots are 25 m apart at their closest point (Nepstad et al., 2002). The soil
in the area is an Oxisol (Haplustox) with 60% (Nepstad et al., 2002) to 90%
clay content (Williams et al., 2002). The water table is 100 m deep (Nepstad
et al., 2002). The treatment plot is covered with plastic panels, installed 1–2
m above the ground, during the wet season. This prevents 50% of the rainfall
(60% of the throughfall) from reaching the soil (Nepstad et al., 2002). The
other plot serves as the control and is never covered with plastic panels. The
first throughfall exclusion period started in early February 2000, after a 1-yr
calibration period (1999) in which both plots were closely monitored to define
pretreatment conditions (Nepstad et al., 2002). To avoid lateral infiltration of
soil water, a trench 1 m wide 31.7 m deep was excavated around each of
the plots and lined with plastic (Nepstad et al., 2002). In the treatment plot,
the water collected by the plastic panels evacuates into the trench, and then
into a deeper drainage ditch that ends in a small valley (Nepstad et al., 2002).
The trenches prevent water infiltration from the outside to a depth of at least
2–3 m (Nepstad et al., 2002; L. Sternberg, University of Miami, personal
observation).
Irrigation design and study species—The first irrigation experiment started
on 7 January 2002, and the second on 16 May 2002. These months correspond
to the start and end of the wet season, respectively. We selected three species
of understory to midcanopy trees shared by both plots: Coussarea racemosa
A. Rich. (‘‘caferana,’’ Rubiaceae), Sclerolobium chrysophyllum Poepp. (‘‘taxi
vermelho,’’ Leguminosae-Caesalpinioideae), and Eschweilera pedicellata
(Rich.) S. A. Mori (‘‘mata-mata´ liso,’’ Lecythidaceae). The ecophysiology
(e.g., sap flow, photosynthesis rate) of these and other focal species in the
plots is being monitored by the TTEE scientific team. Coussarea racemosa,
typically an understory tree of height #10 m, is the most common tree species
in both plots; S. chrysophyllum, usually 15–20 m tall, is the third most com-
mon shared species; and E. pedicellata, generally 10–15 m tall, is the seventh
most common shared species (D. Nepstad et al., The Woods Hole Research
Center, unpublished data on vegetation structure from the plots).
In each of the two experiments, five trees of each species were irrigated in
each plot (Appendix 1, see Supplemental Data accompanying online version
of this article). The same individual trees in the treatment plot were used for
both the first (7 January 2002) and second irrigation (16 May 2002) experi-
ments because we predicted that the label in the treatment plot would remain
close to the surface and any further irrigation with deuterium-labeled water
would only reinforce it. We randomly selected adult trees having a diameter
at breast height (DBH) $10 cm, but, in order to sample the same number of
individuals of each species, we occasionally had to select individuals ,10 cm
DBH when there were no adult trees available (Appendix 1, see Supplemental
Data accompanying online version of this article). If a tree in the treatment
plot grew within 5 m of a very large tree (DBH .1 m), it was not selected
because stemflow from the large tree could overwhelm the throughfall exclu-
sion treatment.
In each irrigation experiment,8Lofasolution made of 99% D
2
O (Icon
Services, Summit, New Jersey, USA) and local water was evenly sprinkled
around the trunk of each tree on a circular area with a radius of 1.5 m, where
most of the water-absorbing roots of adult small trees are apparently found
(Sanford and Cuevas, 1996; Sternberg et al., 2002). The dDof the solution
irrigated was around 130000‰. The litter around each tree was removed
before the irrigation to ensure percolation of the deuterium label into the soil
profile. The litter was replaced immediately after irrigation. No extra water
was added to push the label into the soil because this would have affected
the throughfall exclusion treatment. The 8 L of deuterium-labeled water rep-
resents 1 mm of rain. In the control plot, this amount of water is equivalent
to 23% of the average daily throughfall (0.06% of the average annual through-
fall), while for the treatment plot it represents 32% of the average daily
throughfall (0.08% of the average annual throughfall).
Sample collection—The label of the first irrigation (7 January 2002, day
0) and the second irrigation (16 May 2002, day 0) experiments was traced by
collecting soil and plant samples at periodic time intervals during the wet and
dry seasons of 2002.
Soil samples—Samples for the first irrigation experiment were collected in
both plots on day 2, day 8, and day 17. In the control plot, soil samples were
also taken on day 128, day 156, and day 281. Samples for the second irri-
gation experiment were taken on day 6, day 27, and day 152. Background
deuterium levels were assessed in soil cores sampled to a depth of at least 80
cm in a nearby area, outside the plots, on 8 January 2002 (N53 cores) and
24 January 2002 (N51). At each sampling date, the soil from two to six
irrigated areas around the trees, randomly selected, was sampled using a man-
ual soil auger (Forestry Suppliers, Inc., Jackson, Massachusetts, USA). A soil
core (one per irrigated area) consisted of soil samples (approximately 7–26 g
of wet soil) taken at the surface and at various sampling depths. Sampling
depths were intended to include the whole deuterium pulse in the soil (i.e.,
samples were taken down to a depth below the expected depth of the deute-
rium peak at the time of collection). We collected the soil samples in clear
25 3150 mm screw-cap culture glass tubes and sealed them immediately
around the cap with Parafilm M (Thomas Scientific, Inc., Swedesboro, New
Jersey, USA) to avoid any isotope fractionation due to evaporation.
Plant samples—Wood cores (approximately 5 cm long and 5 mm diameter)
at breast height (130 cm) were collected from C. racemosa and S. chryso-
phyllum trees using a Haglo¨f Increment Borer (Forestry Suppliers, Inc., Jack-
son, Massachusetts, USA). Wood cores from E. pedicellata could not be col-
lected because of its very dense hardwood; instead, we collected stems (ap-
proximately 8 cm long 38 mm diameter) with a complete covering of cork
(complete periderm) to avoid fractionation due to evaporation. Samples for
the first irrigation experiment were collected on day 8 and day 17 in both
plots. In the control plot, samples were also taken on day 130, day 156, and
day 281. Samples for the second irrigation experiment were collected on day
6, day 27, and day 152 in both plots. Background deuterium levels were
measured in plant stems from a nearby area outside the plots on 9 January
2002 (N57 stems), 12 January 2002 (N53), 24 January 2002 (N510)
and 22 May 2002 (N56). Wood cores and stems were collected in vacu-
tainers (13 3100 mm; Becton Dickinson, Franklin Lakes, New Jersey, USA)
that were immediately sealed with the vacutainer’s rubber cap and Parafilm.
Because the samples were transported via airplane, air in the vacutainers was
evacuated with a syringe to reduce pressure.
Water extraction and deuterium measurements—Soil and plant samples
were shipped to the Laboratory of Stable Isotope Ecology in Tropical Eco-
systems (University of Miami, Miami, Florida, USA) or to the Centro de
Energia Nuclear na Agricultura (Universidade de Sa˜o Paulo, Piracicaba, Sa˜o
Paulo, Brazil). Once they arrived, the samples were stored in a freezer at
2108C.
Procedures for water extraction and measurement of deuterium concentra-
tion followed standard procedures (see e.g., Moreira et al., 2000; Sternberg
et al., 2002). Samples were first thawed overnight to ambient temperature
(208C) and then placed in a cryogenic vacuum distillation system. Water col-
lected in this manner was stored in small scintillation vials with waterproof
caps (Fisher Scientific Co., Suwanee, Georgia, USA). The hydrogen from the
water was extracted using a modified version of the Coleman et al. (1982)
method: (1) Three mL of water were mixed in an ampoule with 150 mg of
Zn reagent (Biogeochemical Laboratories, Indiana University, Bloomington,
Indiana, USA) previously outgassed at 3508C for 10 min, (2) the ampoule
was frozen in liquid nitrogen and sealed under high vacuum, and (3) hydrogen
gas was produced by placing the ampoule in an oven for 2 h at 5008C. The
deuterium content of the hydrogen gas sample was measured with a Micro-
mass Prism II dual-inlet isotope ratio mass spectrometer (Micromass Inc.,
Waters Corporation, Milford, Massachusetts, USA) and/or a Finnigan Mat
Delta E mass spectrometer (Thermo Electron Co., Bremen, Germany). The
average precision of the spectrometers was 60.14 dDunits.
Deuterium content is expressed as dDvalues, which represent the relative
difference per mil (‰) between the deuterium isotopic composition of a sam-
ple and that of Vienna Standard Mean Ocean Water (VSMOW; Craig, 1961).
446 [Vol. 92A
MERICAN
J
OURNAL OF
B
OTANY
The isotopic composition is calculated using the atomic ratio (R) of ‘‘heavy’’
isotopes to ‘‘light’’ isotopes. In this case, Ris the atomic ratio of deuterium
hydrogen (
2
Hor simply D) to protium hydrogen (
1
Hor simply H):
(D/H)
sample
dD52131000 (1)
[]
(D/H)
SMOW
Data analysis—Data from the first irrigation experiment covered the period
from January, the start of the wet season, to October, mid-dry season. Data
from the second irrigation experiment covered the period from May, near the
end of the wet season, to October, mid-dry season. To simplify the interpre-
tation, all data from the second irrigation experiment were considered dry
season data.
Soil samples—Soil dDvalues at each sampling date were averaged and
plotted against depth to show the migration of the label in the soil. We de-
veloped a MATLAB (The MathWorks Inc., Natick, Massachusetts, USA) pro-
gram to calculate the expected dDvalues for every cm in the soil profile via
a cubic spline interpolation using the observed mean dDvalues. We used this
program to calculate the mean depth of peak deuterium concentration in the
soil for each sampling date. The spline interpolation was represented in the
figures by a curve, which was graphed using SigmaPlot (Systat Software Inc.,
Point Richmond, California, USA). Data from 9 January 2002 sampling (day
2 after the first irrigation) was not graphed because the soil cores on that day
were taken only 20 cm deep. Soil water in the irrigated areas at any depth
was considered enriched when it was above the highest background value (dD
5212‰).
In addition, to understand how the amount of rain reaching the soil affects
the vertical movement of water in the soil, we did a correlation analysis
between the depths of peak deuterium concentration and the amount of
throughfall accumulated over time, in the presence and in the absence of
panels. The treatment plot was assumed to have no panels installed on 22
May and 12 June 2002, even though the panels in reality started to be removed
on 1 July 2002. This approach is justified because there was no significant
difference between the amount of throughfall under the panels in the treatment
plot and the amount of throughfall in the control plot from 22 May to 1 July
2002 (two-tailed ttest: t51.56, P50.12, N540 d).
Plant samples—We used a one-tailed ttest to evaluate if dDvalues of plant
samples were significantly higher than background deuterium levels at each
sampling date. Because no independent plant sampling of background deu-
terium levels was done for the 12 June and 15 October 2002 sampling dates,
data from these dates were compared to the 22 May 2002 background data.
Model to determine mean depth of water uptake by plants—To estimate the
mean depth of water uptake by a tree at a given time, the dDvalues in the
soil profile and the dDvalue of sap water were coupled in a model. Sap water
data used in the model met the following criteria: (1) deuterium concentration
of water sampled at a given date was significantly greater than background
levels (at least in one of the plots), (2) sufficient time had already passed for
the label to migrate to the sample point (breast height in case of wood cores,
or lowest accessible branch in case of stems), so that equilibration of the
deuterium concentration in the soil–plant continuum had probably been at-
tained, and (3) data for a given date was available for both plots. Data from
the following sampling dates complied with the three criteria: 24 January 2002
(day 17 after first irrigation), 12 June 2002 (day 27 after second irrigation),
and 15 October 2002 (day 152 after second irrigation). Data from day 8 of
the first irrigation experiment and day 6 of the second irrigation experiment
did not meet the second criterion, and therefore were not used in this analysis.
Furthermore, the time water (i.e., the deuterium label) takes to migrate to the
sample point varies between treatment and control trees during the throughfall
exclusion months (wet season) because average sap flow velocity of the treat-
ment trees is usually 50% slower than that of the control trees, taking at least
3–5 d for water to migrate from upper soil layers to the sample point (D.
Nepstad et al., The Woods Hole Research Center, unpublished data).
The model, written in MATLAB (Appendix 2, see Data Supplement ac-
companying the online version of this article), was constructed with two as-
sumptions and one axiom.
Assumption 1—We assumed that, at any given time, a tree can take up
water from a 50-cm vertical segment of soil at any depth in the soil profile.
The following reasoning was used to support this assumption. During 3 yr of
time-domain-reflectometry (TDR) measurements of volumetric water content
in the soil (January 2000–November 2002), the maximum daily average at
depths of 0.3, 0.5, 1, 2 and 3 m, for both plots, was 0.35 cm
3
/cm
3
, while the
minimum average was 0.25 (Nepstad et al., 2002; D. Nepstad et al., The
Woods Hole Research Center, unpublished data). Thus, assuming that the
difference between the maximum TDR and the minimum TDR is due solely
to evapotranspiration, each cm of soil provides an average maximum of 1
mm of water for evapotranspiration (0.35 20.25 50.1; i.e., 1 mm per cm
of soil). Because daily evapotranspiration in these forests is on the average
around 4 mm/d (Leopoldo et al., 1995; Hodnett et al., 1996; Jipp et al., 1998;
Costa and Foley, 1999), plants must acquire water every day from a vertical
segment of soil approximately 4 cm long.
There may be differences in the length of soil segment used by different
species of trees to harvest water, especially between dry and wet seasons. In
a typical dry season, the mean number of consecutive days without rainfall
(according to 2000, 2001 and 2002 rainfall measurements; D. Nepstad et al.,
The Woods Hole Research Center, unpublished data) is 7 d. To sustain the
transpiration demand during those 7 d when no precipitation falls and the soil
is not recharged with rain, plants must be able to extract water from at least
28 cm of soil (54 cm soil/d 37 d). Furthermore, the maximum number of
continuous days with no rain from January 2000 to December 2002 was 25.
During these extended dry periods, plants must be able to extract water from
at least 100 cm of soil (54 cm soil/d 325 d) to sustain transpiration. The
length of the vertical soil segment (50 cm) used in this study falls within the
range of these two estimates and is a conservative approximation. Any error
in our assumption regarding the length of soil segment used by plants will
lead to only slight changes in the conclusions because the model output is
not very sensitive to this parameter. The mean depths of water uptake cal-
culated with a 50-cm segment were linearly correlated with those calculated
with a 20-cm segment (r50.898, slope 50.91, P,0.001) and with those
calculated with a 80-cm segment (r50.877, slope 50.90, P,0.001). The
slopes of the correlations suggest that for a 60% change in the parameter (i.e.,
length of soil segment), there is only a 10% change in model output (i.e.,
calculated mean depths of water uptake).
Assumption 2—We assumed that the amount of water taken up by a tree is
not the same at all depths throughout the 50-cm segment, but instead is taken
up according to a normal distribution (Sokal and Rohlf, 1995):
1
22
2(Y2m)/2s
n5e(2)
i
sÏ2p
where n
i
is the proportion of water taken up at a depth Y, and mis the mean
depth of water uptake. The proportions (n
i
) always add up to one, except
when the normal curve approaches the surface or the lower depth limit and
is truncated. In those cases, the proportions were corrected by weighing them
against the area under the curve so they always sum to one. A normal dis-
tribution of the depth of water uptake means that 99.7% of the water comes
from a segment of soil that is approximately m63scm long. Because we
assumed that a tree takes up water from a 50-cm segment of soil, the standard
deviation (s) of this normal distribution is equal to 8.33 cm. We also assumed
that trees do not acquire water from two distinct regions of the soil profile
because this phenomenon has been demonstrated only in arid regions where
there are two distinct water sources available for plant uptake, either the water
table or rain water (e.g., Dawson and Ehleringer, 1991; Schwinning et al.,
2002).
Axiom 1—The axiom is supported by mass balance principles and states
that the deuterium signature in the plant stem/trunk water can be interpreted
as the sum of the deuterium signatures of the soil water absorbed at different
depths (Moreira et al., 2000). In the model, therefore, the deuterium signature
March 2005] 447R
OMERO
-S
ALTOS ET AL
.—R
AINFALL EXCLUSION IN AN
A
MAZONIAN FOREST
in the plant sample is equal to the sum of the deuterium signatures at all the
1-cm interval depths within a given 50-cm segment, weighed by the propor-
tion of water uptake at each cm as described by the normal distribution (Eq.
1). This can be expressed as (Moreira et al., 2000):
m
plant soil
dD5(n3dD) (3)
O
cii
i51
where dD
cplant
is the calculated hydrogen isotopic composition of plant trunk/
stem water, mis the maximum depth analyzed, and dD
isoil
is the average
isotopic composition of soil water at the i
th
depth. For every position of the
50-cm segment of water uptake, as it moves deeper and deeper in the soil
profile by 1-cm increments until depth m, there is a calculated mean depth of
water uptake (m). Therefore, for a given deuterium profile in the soil, there
are mpossible dD
cplant
values, each with a corresponding mean depth of water
uptake (m). Those mean depths at which dD
cplant
is equal to the observed dD
value of the actual sample measured in the lab (dD
oplant
) are considered so-
lutions.
In some cases, assuming a soil segment of water uptake of 50 cm, no
solution was possible, i.e., dD
oplant
did not match any dD
cplant
at any depth. We
excluded these trees from the data analysis. The effect of their exclusion is
probably minimal because they represent only 11% of the total number of
samples analyzed (N590 samples). On the other hand, in eight of the 90
samples analyzed, two or three solutions for the mean depth of water uptake
were found. Even though plants are likely taking water from only one of these
depths, we averaged these mean depths and treated them as a single obser-
vation for data analysis. The mean depths of water uptake (m) of the trees in
each species were averaged and compared between the two plots using an
independent two-tailed ttest.
RESULTS
Soil water movement—First irrigation experiment: wet and
dry seasons 2002—During the wet season of 2002 (January–
May), the deuterium-labeled water percolated at a mean rate
of 0.75 m/mo in the control plot, and 0.07 m/mo in the treat-
ment plot (Fig. 2). On 15 May (128 d after the first irrigation),
the deuterium peak in the control plot had reached an average
depth of 254 cm while in the treatment plot the peak had only
reached an average depth of 29 cm (Fig. 2). During the dry
season (which starts around June), the deuterium-labeled water
in the control plot not only stopped percolating but actually
moved back upward (Table 1, Figs. 2 and 3): by 15 October
(N52 soil cores) the deuterium peak rose to an average depth
of 190 cm, i.e., 64 cm shallower than the 15 May mean depth
(N54 soil cores), a difference that is statistically significant
(P50.046, one-tailed ttest). In the treatment plot, we were
not able to observe upward movement of the deuterium label
at 2 m depth because the deuterium did not percolate deeply
(Fig. 2).
Two days after the irrigation, on 9 January, the peak deu-
terium concentration in the treatment plot (mean dD61SE
510348 62568; N57 soil cores) was significantly greater
(P50.03, two-tailed ttest) than that in the control plot (mean
dD61SE53427 6696; N56). As the deuterium per-
colated down the soil profile, the concentration in both plots
kept decreasing. However, by 15 May, the difference in the
peak deuterium concentrations between the plots had become
much less obvious than that of 15 January (Fig. 2).
Second irrigation experiment: dry season 2002—During the
dry season of 2002, the deuterium label in both plots perco-
lated downwards at a similar mean rate (approximately 0.15
m/mo; Fig. 4). Thus, by 15 October (152 days after the May
irrigation) the label peak had percolated to similar depths in
both plots (30 cm in the control, and 41 cm in the treatment;
Fig. 4, Table 1). The depths of these peaks were similar to the
depth reached by the peak in the treatment plot during the wet
season (29 cm; Fig. 2), while the panels were in place simu-
lating a dry season. Upward movement of the second label
during the dry season in deeper soil layers was not observed
in either plot because the deuterium did not percolate deeply
(Fig. 4). As in the first irrigation, the deuterium concentration
in both plots decreased over time (Fig. 4).
Effect of precipitation on the percolation of deuterium—
Downward movement of the deuterium peak was linearly cor-
related with throughfall, regardless of whether or not the pan-
els were present (both P#0.01; Fig. 5). Under natural (no
panels) conditions (i.e., during the dry season in the treatment
plot and all year long in the control plot), every cm of rain
‘‘pushed’’ the soil water a distance of about 2 cm (1 : 2 re-
lationship; slope 51.88, r50.98, P,0.0001; Fig. 5). On
the other hand, depths of deuterium peaks in the treatment plot
when the panels were in place were approximately half the
amount of accumulated throughfall (slope 50.45, r50.99,
P50.01; Fig. 5). The correlation under natural conditions
was strengthened by the inclusion of four data points (repre-
sented as triangles in Fig. 5) obtained from another deuterium
irrigation experiment conducted in the same site, next to the
TTEE plots, from March 1999 to January 2000 (Sternberg et
al., 2002). The three extreme points (upper right corner of Fig.
5) do not represent variation of the data, but only an artifact
of sampling frequency. The statistical significance of the linear
correlation does not depend on these extreme points because
if they are excluded from the analysis the correlation is still
significant and the approximate 1 : 2 relationship is maintained
(slope 52.28, r50.928, P,0.001).
Patterns of plant enrichment and mean depths of water
uptake—After 17 d of the first irrigation, average dDvalues
of sap water from all tree species in both plots were signifi-
cantly above background (Table 1). A similar result was found
27 d after the second irrigation (Table 1). This initial deute-
rium enrichment decreased over time, but not at a constant
rate across species or individuals. For example, E. pedicellata
was still significantly enriched 130 d after the first irrigation
in the control plot, and three C. racemosa control trees (9–
164, 9–189, 9–201; Appendix 1, see Supplemental Data ac-
companying online version of this article) became significantly
enriched again in October after being at background levels at
least since 17 May (PK0.001, Ztest using 241.2, the mean
background dDvalue of 12 June, as the population mean).
Although the enrichment of these three trees increased the
mean dDof all five C. racemosa trees to 215.2, this mean
was not significantly higher than background levels (Table 1).
The mean depths of water uptake, as calculated by the mod-
el, follow similar patterns across the three species studied. Fur-
ther, the differences in the mean depth of water uptake between
control and treatment plots, at a given date, were significant
or almost significant for all three species (two-tailed ttest; Fig.
6). On 24 January, 17 d after the first irrigation, at the onset
of the wet season, trees in the control plot were on average
taking up water at 63 cm deeper than the trees in the treatment
plot, which at that time was covered by the panels (Fig. 6).
This pattern reversed on 12 June (the start of the dry season,
27 d after the second irrigation) and 15 October (mid-dry sea-
son, 152 d after the second irrigation), when the treatment
448 [Vol. 92A
MERICAN
J
OURNAL OF
B
OTANY
Fig. 2. Movement of deuterium-labeled water in the soil profile (depth) in the first irrigation experiment at the Tapajo´s Throughfall Exclusion Experiment,
in Brazil. Trees were irrigated at the start of the wet season on 7 January 2002. In 2002, the treatment plot was covered with panels from 1 January to 1 July.
Note the scale differences of the dDaxis as the deuterium concentration in the soil diminishes over time. dDunits represent the per mil (‰) relative difference
between the deuterium isotopic composition of a sample and that of an international standard (V-SMOW, with dD50). Error bars represent 61 SE.
trees were taking up water at greater depths than the control
trees (Fig. 6). On average, on 12 June and 15 October, treat-
ment trees harvested water at a depth 26 cm and 37 cm deeper
than control trees, respectively (Fig. 6). As the dry season
progressed from June to October, trees in both plots tended to
take up water at greater depths (Fig. 6).
DISCUSSION
Rain regulates the percolation of water in the soil profile,
but water may also move upward—During the 2002 wet sea-
son, the percolation rate in the control plot (0.75 m/mo) was
an order of magnitude higher than that in the treatment plot
March 2005] 449R
OMERO
-S
ALTOS ET AL
.—R
AINFALL EXCLUSION IN AN
A
MAZONIAN FOREST
T
ABLE
1. Deuterium isotopic composition (mean
d
D
6
1 SD) of the plant samples from the trees irrigated with deuterium-labeled water (N
5
5 trees per species in each irrigation
experiment) at the Tapajo´s Throughfall Exclusion Experiment, in Brazil. Asterisks (*
5
P
#
0.05, **
5
P
#
0.01) denote significantly higher
d
Dvalues (one-tailed ttest) with
respect to background control
d
Dvalues of non-irrigated trees. The trees in the treatment plot were used for both irrigations, while a second set of trees in the control plot was
selected for the second irrigation (see Materials and Methods, Irrigation design and study species; Appendix 1). The depth range at which the soil was enriched with deuterium (i.e.,
soil
d
Dat that depth range was above the maximum background soil
d
Dvalue measured:
2
12‰), and the mean depth of peak deuterium concentrations are also shown (see Materials
and Methods, Data analysis for further explanation).
Date
Control plot
Coussarea
racemosa
(cores)
Sclerolobium
chrysophyllum
(cores)
Eschweilera
pedicellata
(stems)
D-enriched
depth range
(cm)
Mean depth
of peak
(cm)
Treatment plot
Coussarea
racemosa
(cores)
Sclerolobium
chrysophyllum
(cores)
Eschweilera
pedicellata
(stems)
D-enriched
depth range
(cm)
Mean
depth of
peak (cm)
First irrigation experiment (7 January 2002)
24 January 207.8
6
139.3** 39.6
6
75**
2
1.8
6
31.6** 0–150 50 608.0
6
216.5** 105.9
6
95** 41.3
6
68.6** 0–55 0
(day 17)
15/17 May
2
39.1
6
5.9
2
31.2
6
10.9
2
10.5
6
6.3** 84–300 254
(day 128/130)
12 June
2
41.2
6
2.6
2
33.9
6
9.5
2
21.2
6
3.6 142–319 233
(day 156)
15 October
2
15.2
6
25.9
2
37.3
6
9.2
2
20.8
6
9.9 97–300 190
(day 281) Second irrigation experiment (16 May 2002)
12 June 150.0
6
81.8**
2
1.1
6
23.9* 0.5
6
17.6** 0–50 19 176.6
6
80.5** 6.1
6
36.3* 68.9
6
56.3** 0–92 20
(day 27)
15 October
2
6.2
6
18.9*
2
35.8
6
8.3
2
22.2
6
3.6 0–59 30 29
6
22.2**
2
16.3
6
17.2 3.1
6
15** 0–126 41
(day 152)
(0.07 m/mo; Fig. 2), therefore demonstrating that the panels
had significantly reduced the percolation of water in the soil
profile. The percolation rate in the treatment plot in the wet
season (0.07 m/mo; Fig. 2), when the panels were in place,
was even lower than the rate in the dry season (around 0.15
m/mo; Fig. 4). In a physically homogeneous soil profile, per-
colation is mainly regulated by amount and frequency of rain-
fall and initial soil moisture conditions. On 16 May 2002, the
day of the second irrigation at the onset of the dry season,
volumetric soil water content at a depth of 30 cm differed by
an average of 0.13 cm
3
/cm
3
between the control and treatment
plots (0.39 vs. 0.26, respectively; D. Nepstad et al., The Woods
Hole Research Center, unpublished data). This difference in
volumetric soil water content, however, had little effect on the
percolation rate because both plots showed similar rates (ca.
0.15 m/mo; Fig. 4). This confirms that physical characteristics
of the topsoil are alike in both plots and that the main factor
regulating the downward movement of water in the plots is
rain input (Aragua´s-Aragua´s et al., 1995; Moreira et al., 2000;
Sternberg et al., 2002). The rates reported in this study are
within the range published in the literature for humid tropical
forest soils (Kline and Jordan, 1968; Aragua´s-Aragua´s et al.,
1995; Moreira et al., 2000; Sternberg et al., 2002), although
the 2002 wet season percolation rate in the control plot may
be the highest ever reported (0.75 m/mo).
The 1 : 2 relationship between cumulative throughfall and
depth of deuterium peak under natural conditions (Fig. 5) can
be explained by considering that saturated volumetric water
content of this forest soil is ca. 0.5 cm
3
/cm
3
(Nepstad et al.,
2002; D. Nepstad et al., The Woods Hole Research Center,
unpublished data). This means that around 50% of a volume
unit of clay soil is water, while the other 50% is either solid
matter or entrapped air, which explains why 1 cm of rain can
only fit in a soil column if the water from 2 cm of soil is
displaced. This 1 : 2 relationship between the amount of cu-
mulative rainfall and the depth of the deuterium peak does not
apply in the presence of panels (Fig. 5) probably because of
rapid loss of water by plant uptake and/or evaporation at or
near the soil surface.
Water in the soil profile at depths .2 m can move upward
several decimeters during a dry season (Fig. 3). A similar,
although less conspicuous, pattern was also found in a sea-
sonal forest and abandoned pasture in the eastern Amazon
(Moreira et al., 2000), confirming this phenomenon. From the
end of the wet season (15 May) to the middle of the dry season
(15 October), the deuterium peak on average moved up from
254 cm to 190 cm (Figs. 2 and 3, Table 1). This means, as-
suming that the 1 : 2 relationship described previously (Fig.
5) is applicable in this situation as well, that 32 cm of water
moved up during this period, an amount comparable to the
rainfall falling during the wettest months (March and April;
climate data on Santare´m in www.worldclimate.com).
Soil physics theory states that upward movement of water
can be explained by an upward gradient of total potential (ma-
tric 1gravitational) coupled with sufficiently high unsaturated
hydraulic conductivity (see e.g., Marshall et al., 1996). Roots
may create this gradient potential because of greater water ab-
sorption at or near the soil surface, which creates a lower po-
tential than in the soil layers below. Complementarily, thegra-
dient can also be caused by soil evaporation. If water is re-
distributed from deep to shallow soil layers through roots, the
process is commonly known as ‘‘hydraulic lift’’ (Richards and
Caldwell, 1987; Dawson, 1996). In this study, however, water
450 [Vol. 92A
MERICAN
J
OURNAL OF
B
OTANY
Fig. 3. Upward movement of the deuterium-labeled water during the 2002 dry season in the control plot at the Tapajo´s Throughfall Exclusion Experiment,
in Brazil. Irrigation took place on 7 January 2002 (first irrigation experiment). The dotted line represents the 12 June sampling (day 156; N54 soil cores) and
the continuous line the 15 October sampling (day 281; N52). dDunits represent the per mil (‰) relative difference between the deuterium isotopiccomposition
of a sample and that of an international standard (V-SMOW, with dD50). Error bars represent 61 SE.
most probably was not transported through roots. If roots were
absorbing water from moist (deep) soil layers and depositing
it in drier (shallow) soil layers, the deuterium label in the soil
should clearly reflect this process because in general there is
no isotopic fractionation by plant roots (Gonfiantini et al.,
1965; Wershaw et al., 1970). If roots harvested up water at
2.5 m depth and then deposited it around 2 m (or less) night
after night, the 15 October deuterium profile should show a
second peak around 2 m (Fig. 7). Alternatively, if hydraulic
lift occurred gradually throughout the soil profile, as expected
if the fine root distribution is continuous, then the deuterium
profile curve should have an extended tail (Fig. 7). Neither of
these possible deuterium profiles was observed in this study,
and therefore we suggest that, in tropical forests, when little
or no rain falls for a relatively long period of time, significant
amounts of water can move upward from depth not through
roots (hydraulic lift), but through soil pores. In fact, given that
micropores (5–30 mm) and ultramicropores (0.1–5 mm) occur
in clay soils, water in capillaries of this size could be driven
upward several meters if a significant water potential gradient
in the soil profile is present. The extent of this upward move-
ment and its effect on the water relations of eastern Amazonian
forests certainly merits further study.
Water uptake at depths .2 m is minimal—The rapid loss
of deuterium from the surface of the soil is typical of irrigation
studies with a hydrogen-isotope tracer (e.g., Aragua´s-Aragua´s
et al., 1995; Moreira et al., 2000; Sternberg et al., 2002) and
is probably related to evaporation, equilibration with ambient
vapor, or dilution of the label in the water already present at
the soil surface (Zimmermann et al., 1966, 1967a, b; Blume
et al., 1967; Mu¨nnich, 1983). At depths .50 cm, evaporation
is minimal and therefore loss of the label must be related to
dilution caused by mixing of deuterium-labeled water with in-
coming precipitation and/or local soil water, coupled with
plant water uptake (Woods and O’Neal, 1965; Zimmermann
et al., 1966, 1967a, b; Blume et al., 1967; Mu¨nnich, 1983;
Aragua´s-Aragua´s et al., 1995; Moreira et al., 2000; Sternberg
et al., 2002). However, no decrease in deuterium concentration
was observed below 2 m depth (e.g., the dDvalue of the 15
May peak from the first irrigation in the control plot is very
similar to the dDvalues of the 12 June and 15 October peaks;
Figs. 2 and 3). This leads to the conclusion that plant water
uptake at depths .2 m must be minimal (see also Sternberg
et al., 2002), which, regarding the discussion in the previous
section, is further evidence against the possibility that hydrau-
lic lift was the cause of upward movement of water in the
control plot from approximately 2.5 to 2 m depth during the
2002 dry season.
dDof sap water from all tree species in the control plot
were significantly enriched as long as soil water close to the
surface had dDvalues above background (Table 1). By 17
May, after downward percolation of the deuterium label
caused the soil to be above background from 0.84 to 3.0 m
deep, only one species (E. pedicellata) had sap water with dD
values significantly above background levels (Table 1). None
of the species was enriched after the label had percolated fur-
ther downward, and the soil water was above background only
at depths .1.42 m (by 12 June; Table 1). Therefore, the in-
dividuals studied here—some of which had a DBH as large
as 20–26 cm (Appendix 1, see Supplemental Data accompa-
nying online version of this article) and buttresses extending
0.5 m or more from the trunk (particularly S. chrysophyl-
lum)—were not accessing soil water to a substantial degree
much beyond 1 m. This is puzzling when considered in the
context of the observed decrease in soil moisture in the control
plot down to at least 3 m during the dry seasons (Nepstad et
al., 2002), and the presence of fine roots at that depth, although
at a much lower density than at the surface (D. Nepstad et al.,
The Woods Hole Research Center, unpublished data). A pos-
sible solution to this problem is the observation of the upward
movement of the deuterium label during the dry season (Fig.
3). Plants may access a substantial amount of deep soil mois-
ture through this route rather than by direct water uptake at
March 2005] 451R
OMERO
-S
ALTOS ET AL
.—R
AINFALL EXCLUSION IN AN
A
MAZONIAN FOREST
Fig. 4. Movement of deuterium-labeled water in the soil profile (depth) in the second irrigation experiment at the Tapajo´s Throughfall Exclusion Experiment,
in Brazil. Trees were irrigated at the end of the wet season on 16 May 2002. Panels in the treatment plot were removed on 1 July, at the onset of the dry
season. Note the scale differences of the dDaxis as the deuterium concentration in the soil diminishes over time. dDunits represent the per mil (‰) relative
difference between the deuterium isotopic composition of a sample and that of an international standard (V-SMOW, with dD50). Error bars represent 61 SE.
depth. Evidence supporting this hypothesis can be seen in the
changes of the dDvalues of sap water from C. racemosa con-
trol trees in the first irrigation experiment: by October 15,
when the deuterium label had moved upward in the soil pro-
file, three of the five individuals of C. racemosa became un-
expectedly enriched in deuterium (Ztest, PK0.001; see Re-
sults). It is also possible that deep water is accessed only by
larger trees (DBH $50 cm), but this pattern has not been
supported by other studies in neotropical forests. For example,
Meinzer et al. (1999), in Barro Colorado Island (Panama´), ob-
served that, during the dry season, relatively small trees tend
to get water from deeper layers of the soil profile compared
452 [Vol. 92A
MERICAN
J
OURNAL OF
B
OTANY
Fig. 5. Relationship between depth of deuterium peak in the soil profile and cumulative throughfall reaching the soil under natural conditions (three data
sources: treatment plot during the dry season [mostly without panels], control plot all year long, and forest site nearby the plots reported by Sternberg et al.,
2002) and when it is covered by panels (one data source only: treatment plot during the wet season). The 12 June and 15 October 2002 samplings for the first
irrigation experiment in the control plot were excluded from this analysis because the deuterium peak was moving upward. The 22 May and 12 June 2002 data
for the second irrigation experiment in the treatment plot were analyzed as if the panels already had been removed, although they were actually removedon1
July 2002 (see Materials and Methods, Data analysis for further explanation).
Fig. 6. Mean depths of water uptake (m)byCoussarea racemosa,Sclerolobium chrysophyllum, and Eschweilera pedicellata trees, as inferred from the
deuterium signatures of the plants (see Table 1) and the deuterium profiles of the soil (see Figs. 2 and 4). The mean depths were calculated using a conceptual
model (see Materials and Methods, Data analysis). The number below each column is the sample size (N), i.e., the number of trees (of five in each species)
that had a solution in the model. The probabilities (P) of a two-tailed ttest performed between control and treatment plots are also shown. Error bars represent
61 SE.
March 2005] 453R
OMERO
-S
ALTOS ET AL
.—R
AINFALL EXCLUSION IN AN
A
MAZONIAN FOREST
Fig. 7. Expected profiles of deuterium concentration in the soil as a con-
sequence of hydraulic lift, and possible root distributions explaining such
profiles. Full functionality of all roots shown is assumed. The y-axis represents
the approximate depths at which upward movement of water was observed
in this study (from 2.5 m to 2 m; Table 1, Figs. 2 and 3). The original
deuterium profiles are indicated by a dotted line which overlaps the expected
deuterium profile at certain depths. (A) Water is absorbed by the roots at 2.5
m and deposited in the soil at 2 m; as a result, two dDpeaks should be
observed. This is the classic model of hydraulic lift, except that plants may
also deposit the absorbed water closer to the surface. (B) Water is absorbed
by the roots at 2.5 m and gradually deposited in the soil profile above this
depth; as a result, the dDvalue of the soil layers above 2.5 m should steadily
increase over time.
to larger trees. In addition, Sternberg et al. (1998), in Fazenda
Victo´ria (eastern Brazil), showed that there is no vertical dis-
tinct segregation of roots in the soil horizon with respect to
tree size.
Treatment trees tap water at significantly greater depths
than control trees, but also respond to seasonal stimuli—
Because water content is much lower in the surface of the
treatment plot and therefore is more quickly depleted (Nepstad
et al., 2002), it was expected that treatment trees would access
water from deeper layers of the soil profile. Indeed, mean
depths of water uptake by treatment trees were deeper than
those of control trees in June (end of 2002 wet season) and
October (mid 2002 dry season) (Fig. 6). All three species had
this pattern, with some species in the treatment plot accessing
water as much as 0.5 m below that accessed in the control
plot (Fig. 6). Superimposed on the difference in depth of water
uptake between treatment and control plants is a seasonal ef-
fect: during the end of the wet season (June) both treatment
and control trees were on average harvesting water from shal-
lower layers of the soil profile (0.3–0.8 m) than during the
mid-dry season (0.6–1.25 m; Fig. 6). The similarity in depth
of water uptake response implies that fine roots apparently
function and are distributed in a similar way among these three
unrelated species of trees.
The January sampling (at the start of the 2002 wet season)
did not conform to our expectations because, on average, trees
in the treatment plot were actually accessing soil water from
shallower, not deeper, layers of the soil profile than control
trees (0.13–0.57 m vs. 0.79–1.11 m, respectively; Fig. 6). The
simplest explanation of this phenomenon is that the irrigation
with 8 L of deuterium-labeled water (1 mm of rain) around
each tree in the treatment plot moistened the soil surface
enough to induce fine root production and the subsequent ab-
sorption of moisture at that layer of the soil profile. The en-
riched water was sprinkled on a relatively litter-free soil (see
Methods), which may have reinforced this effect. A comple-
mentary explanation is that substantial throughfall and stem-
flow leaked between the panels at the very beginning of Jan-
uary 2002 (e.g., 162 mm from 1 January to 24 January; D.
Nepstad et al., The Woods Hole Research Center, unpublished
data), which induced superficial fine root growth around the
trunks from root primordia that were latent during the strong
2001 dry season. As a result, treatment trees may have quickly
switched their strategy from deep to shallow water uptake at
the start of the 2002 wet season, independent of the addition
of the deuterium-labeled water (7 January). In fact, a major
increase in fine root production at 50 cm depth in the treatment
plot has been observed at the start of a wet season (D. Nepstad
et al., The Woods Hole Research Center, unpublished data).
Rapid superficial fine root production in response to rain has
also been observed in other studies in the tropics (e.g., Sanford
and Cuevas, 1996; Cao, 2000; Yavitt and Wright, 2001).
Shallow fine root production in response to the first wet
season rains has also been observed in the control plot (D.
Nepstad et al., The Woods Hole Research Center, unpublished
data) but, at least at the very beginning of the 2002 wet season,
the bulk of water uptake seemed to still occur at depths around
1 m (0.79–1.11 m, on average; Fig. 6). The water at these
depths was probably enough to sustain transpiration demand
because it had not been depleted by the throughfall exclusion
treatment and was being quickly recharged by infiltrating pre-
cipitation (see de Souza et al., 1996; Nepstad et al., 2002).
This could be the situation throughout the wet season, al-
though we lack data to support this speculation. When the
frequency and intensity of rains began to decrease in June (a
transitional month between wet and dry seasons), the control
as well as the treatment trees eventually conformed to the ex-
pected trend: control trees tapped water at shallower depths
than treatment trees (Fig. 6) because of higher depletion of
surface water content in the treatment plot than in the control
plot (D. Nepstad et al., The Woods Hole Research Center,
unpublished data).
Conclusions—In 2002, we applied deuterium-enriched wa-
ter around selected understory/subcanopy trees in the treat-
ment and control plots of the Tapajo´s Throughfall Exclusion
Experiment (see Nepstad et al., 2002) to describe differences
between plots in regards to (1) seasonal movement of soil pore
water in the soil profile, and (2) depth at which such moving
water is taken up by the trees.
During the wet season, the percolation rate in the treatment
plot was lower than that in the control plot because the plastic
panels effectively simulated a dry season. During the dry sea-
son, when the panels were removed and/or both plots received
the same rainfall, the percolation rates of treatment and control
plots were very similar, indicating similar physical character-
istics between both plots. Interestingly, upward water move-
ment was detected with the deuterium label in the control plot
during the dry season. This phenomenon was probably caused
by an upward gradient of total potential (matric 1gravita-
tional) and sufficient unsaturated hydraulic conductivity in the
soil. We argue that this water mostly moved through soil pores,
not through roots (hydraulic lift), but further studies should
confirm or reject this postulate. Upward water movement most
likely occurred in the treatment plot as well, but was not de-
tected because the label did not percolate deeply in this plot.
The three different tree species responded in a similar man-
ner to the throughfall exclusion treatment in regards to the
depth at which they harvested water and therefore may be
454 [Vol. 92A
MERICAN
J
OURNAL OF
B
OTANY
representative of other understory/subcanopy trees in the plots.
In the dry season, treatment trees obtained most water at deep-
er soil layers than control trees (although never below 1.5–2
m). This was expected because the wet-season throughfall-
exclusion treatment, which started in February 2000, signifi-
cantly reduced water content of the top soil layers in the treat-
ment plot (Nepstad et al., 2002), probably reducing fine root
biomass and function.
How the depth of water uptake by different life forms will
change as the water content of the soil in the treatment plot
continues to diminish is unknown. The understory/subcanopy
trees, especially during the dry season, may continue to har-
vest water even deeper, but is unknown if other life forms
(e.g., lianas) or life stages (e.g., seedlings or large trees with
DBH $50 cm) will respond in similar ways. This kind of
ecophysiological information is important to understand how
eastern Amazonian forests would respond in the event of ex-
tended droughts, which are expected during El Nin˜o years be-
cause of global climate change (Nepstad et al., 2002).
LITERATURE CITED
A
RAGUA
´S
-A
RAGUA
´S
, L., K. R
OZANSKI
,A.P
LATA
B
EDMAR
,A.R.T
UNDIS
V
ITAL
,A.C.T
ANCREDI
,
AND
W. F
RANKEN
. 1995. Changes of soil water
balance due to forest clearing in the central Amazon region. In Solutions
’95, Managing the effects of man’s activities on groundwater, IAH In-
ternational Congress XXVI, 4–10 June 1995, Edmonton, Canada. CD-
ROM, Mow-Technology Ltd., Edmonton, Alberta, Canada.
B
ISHOP
, K.,
AND
E. D
AMBRINE
. 1995. Localization of tree water uptake in
Scots pine and Norway spruce with hydrological tracers. Canadian Jour-
nal of Forest Research 25: 286–297.
B
LUME
, H. P., U. Z
IMMERMANN
,
AND
K. O. M
U
¨NNICH
. 1967. Tritium tagging
of soil moisture: the water balance of forest soils. In Isotope and radiation
techniques in soil physics and irrigation studies, proceedings of a sym-
posium, Istanbul, Turkey, 12–16 June 1967, 315–332. International
Atomic Energy Agency, Vienna, Austria.
B
ORCHERT
, R., G. R
IVERA
,
AND
W. H
AGNAUER
. 2002. Modification of veg-
etative phenology in a tropical semi-deciduous forest by abnormal
drought and rain. Biotropica 34: 27–39.
C
AO
, K. 2000. Water relations and gas exchange of tropical saplings during
a prolonged drought in a Bornean heath forest, with reference to root
architecture. Journal of Tropical Ecology 16: 101–116.
C
ATTA
ˆNIO
, J. H., E. A. D
AVIDSON
,D.C.N
EPSTAD
,L.V.V
ERCHOT
,
AND
I.
L. A
CKERMAN
. 2002. Unexpected results of a pilot throughfall exclusion
experiment on soil emissions of CO
2
,CH
4
,N
2
O and NO in eastern Ama-
zonia. Biology and Fertility of Soils 36: 102–108.
C
OLEMAN
, M. L., T. J. S
HEPHERD
,J.J.D
URHAM
,J.E.R
OUSE
,
AND
G. R.
M
OORE
. 1982. Reduction of water with zinc for hydrogen isotope anal-
ysis. Analytical Chemistry 54: 993–995.
C
OSTA
,M.H.,
AND
J. A. F
OLEY
. 1999. Trends in the hydrological cycle of
the Amazon basin. Journal of Geophysical Research—Atmospheres
104(D12): 14189–14198.
C
RAIG
, H. 1961. Standard for reporting concentrations of deuterium and ox-
ygen-18 in natural waters. Science 133: 1833–1834.
D
AWSON
, T. E. 1996. Determining water use by trees and forests from iso-
topic, energy balance and transpiration analyses: the roles of tree size
and hydraulic lift. Tree Physiology 16: 263–272.
D
AWSON
,T.E.,
AND
J. R. E
HLERINGER
. 1991. Streamside trees that do not
use stream water. Nature 350: 335–337.
D
AWSON
,T.E.,
AND
J. S. P
ATE
. 1996. Seasonal water uptake and movement
in root systems of Australian phraeatophytic plants of dimorphic root
morphology: a stable isotope investigation. Oecologia 107: 13–20.
D
E
S
OUZA
, J. R. S., F. M. A. P
INHEIRO
,R.L.C.
DE
A
RAUJO
,H.S.J.P
IN
-
HEIRO
,
AND
M. G. H
ODNETT
. 1996. Temperature and moisture profiles
in soil beneath forest and pasture areas in eastern Amazonia. In J. H. C.
Gash, C. A. Nobre, J. M. Roberts, and R. L. Victoria [eds.], Amazonian
deforestation and climate, 125–137. John Wiley and Sons, Chichester,
UK.
E
NGELBRECHT
,B.M.J.,
AND
T. A. K
URSAR
. 2003. Comparative drought-
resistance of seedlings of 28 species of co-occurring tropical woody
plants. Oecologia 136: 383–393.
E
NGELBRECHT
, B. M. J., S. J. W
RIGHT
,
AND
D. D
E
S
TEVEN
. 2002. Survival
and ecophysiology of tree seedlings during El Nin˜o drought in a tropical
moist forest in Panama´. Journal of Tropical Ecology 18: 569–579.
F
LANAGAN
,L.B.,
AND
J. R. E
HLERINGER
. 1991. Stable isotope composition
of stem and leaf water: applications to the study of plant water use.
Functional Ecology 5: 270–277.
G
ONFIANTINI
, R., S. G
RATZIU
,
AND
E. T
ONGIORGI
. 1965. Oxygen isotopic
composition of water in leaves. In Isotopes and radiation in soil–plant
nutrition studies, proceedings of a symposium, 405–410. Ankara, Turkey,
1965. International Atomic Energy Agency, Vienna, Austria.
H
ODNETT
, M. G., J. T
OMASELLA
,A.
DE
O. M
ARQUES
F
ILHO
,
AND
M. D.
O
YAMA
. 1996. Deep soil water uptake by forest and pasture in central
Amazonia: predictions from long-term daily rainfall data using a simple
water balance model. In J. H. C. Gash, C. A. Nobre, J. M. Roberts, and
R. L. Victoria [eds.], Amazonian deforestation and climate, 79–99. John
Wiley, Chichester, UK.
J
ACKSON
, P. C., J. C
AVELIER
,G.G
OLDSTEIN
,F.C.M
EINZER
,
AND
N. M.
H
OLBROOK
. 1995. Partitioning of water resources among plants of a
lowland tropical forest. Oecologia 101: 197–203.
J
ACKSON
, P. C., F. C. M
EINZER
,M.B
USTAMANTE
,G.G
OLDSTEIN
,A.F
RANCO
,
P. W. R
UNDEL
,L.C
ALDAS
,E.I
GLER
,
AND
F. C
AUSIN
. 1999. Partitioning
of soil water among tree species in a Brazilian cerrado ecosystem. Tree
Physiology 19: 717–724.
J
IPP
, P. H., D. C. N
EPSTAD
,D.K.C
ASSEL
,
AND
C. R.
DE
C
ARVALHO
. 1998.
Deep soil moisture storage and transportation in forests and pastures of
seasonally-dry Amazonia. Climatic Change 39: 395–412.
K
LINE
,J.R.,
AND
C. F. J
ORDAN
. 1968. Tritium movement in soil of tropical
rain forest. Science 160: 550–551.
L
EOPOLDO
, P. R., W. K. F
RANKEN
,
AND
N. A. V
ILLA
N
OVA
. 1995. Real
evapotranspiration and transpiration through a tropical rain forest in cen-
tral Amazonia as estimated by the water balance method. Forest Ecology
and Management 73: 185–195.
M
ARSHALL
, T. J., J. W. H
OLMES
,
AND
C. W. R
OSE
. 1996. Soil physics. Cam-
bridge University Press, New York, New York, USA.
M
EINZER
, F. C., J. L. A
NDRADE
,G.G
OLDSTEIN
,N.M.H
OLBROOK
,J.C
AV
-
ELIER
,
AND
S. J. W
RIGHT
. 1999. Partitioning of soil water among canopy
trees in a seasonally dry tropical forest. Oecologia 121: 293–301.
M
EINZER
, F. C., M. J. C
LEARWATER
,
AND
G. G
OLDSTEIN
. 2001. Water trans-
port in trees: current perspectives, new insights and some controversies.
Environmental and Experimental Botany 45: 239–262.
M
OREIRA
, M. Z., L.
DA
S. L. S
TERNBERG
,
AND
D. C. N
EPSTAD
. 2000. Vertical
patterns of soil water uptake by plants in a primary forest and an aban-
doned pasture in the eastern Amazon: an isotopic approach. Plant and
Soil 222: 95–107.
M
ULKEY
,S.S.,
AND
J. S. W
RIGHT
. 1996. Influence of seasonal drought on
the carbon balance of tropical forest plants. In S. S. Mulkey, R. L. Chaz-
don, and A. P. Smith [eds.], Tropical forest plant ecophysiology, 187–
216. Chapman and Hall, New York, New York, USA.
M
U
¨NNICH
, K. O. 1983. Moisture movement in the unsaturated zone. In
Guidebook of nuclear techniques in hydrology, Technical Report Series
no. 91, 203–222. International Atomic Energy Agency, Vienna, Austria.
N
EPSTAD
, D. C., C. R.
DE
C
ARVALHO
,E.A.D
AVIDSON
,P.H.J
IPP
,P.A.
L
EFEBVRE
,G.H.N
EGREIROS
,E.D.
DA
S
ILVA
,T.A.S
TONE
,S.E.T
RUM
-
BORE
,
AND
S. V
IEIRA
. 1994. The role of deep roots in the hydrological
and carbon cycles of Amazonian forests and pastures. Nature 372: 666–
669.
N
EPSTAD
, D. C., A. G. M
OREIRA
,
AND
A. A
LENCAR
. 1999. Flames in the
rain forest: origins, impacts and alternatives to Amazonian fire. The Pilot
Program to Conserve the Brazilian Rain Forest, World Bank, Brasilia,
Brazil.
N
EPSTAD
, D. C., P. M
OUTINHO
,M.B.D
IAS
–F
ILHO
,E.A.D
AVIDSON
,G.
C
ARDINOT
,D.M
ARKEWITZ
,R.F
IGUEIREDO
,N.V
IANNA
,J.Q.C
HAM
-
BERS
,D.G.R
AY
,J.B.G
UERREIROS
,P.A.L
EFEBVRE
,L.
DA
S. L. S
TERN
-
BERG
,M.Z.M
OREIRA
,L.B
ARROS
,F.Y.I
SHIDA
,I.T
OHLVER
,E.B
ELK
,
K. K
ALIF
,
AND
K. S
CHWALBE
. 2002. The effects of partial throughfall
exclusion on canopy processes, aboveground production, and biogeo-
chemistry of an Amazon forest. Journal of Geophysical Research—At-
mospheres 107 (D20), 8085, doi: 10.1029/2001JD000360.
P
ARROTTA
, J. A., J. K. F
RANCIS
,
AND
R. R.
DE
A
LMEIDA
. 1995. Trees of the
Tapajo´s: a photographic field guide. General Technical Report IITF-1.
March 2005] 455R
OMERO
-S
ALTOS ET AL
.—R
AINFALL EXCLUSION IN AN
A
MAZONIAN FOREST
United States Department of Agriculture, Forest Service, International
Institute of Tropical Forestry, Rı´o Piedras, Puerto Rico.
P
EN
˜UELAS
, J.,
AND
I. F
ILELLA
. 2003. Deuterium labelling of roots provides
evidence of deep water access and hydraulic lift by Pinus nigra in a
Mediterranean forest of NE Spain. Environmental and Experimental Bot-
any 49: 201–208.
P
LAMBOECK
, A. H., H. G
RIP
,
AND
U. N
YGREN
. 1999. A hydrological tracer
study of water uptake depth in a Scots pine forest under two different
water regimes. Oecologia 119: 452–460.
P
OTTS
, M. D. 2003. Drought in a Bornean everwet rain forest. Journal of
Ecology 91: 467–474.
R
ICHARDS
,J.H.,
AND
M. M. C
ALDWELL
. 1987. Hydraulic lift: substantial
nocturnal water transport between soil layers by Artemisia tridentata
roots. Oecologia 73: 486–489.
S
ANFORD
,R.L.J
R
.,
AND
E. C
UEVAS
. 1996. Root growth and rhizosphere
interactions in tropical forest. In S. S. Mulkey, R. L. Chazdon, and A. P.
Smith [eds.], Tropical forest plant ecophysiology, 268–300. Chapman
and Hall, New York, New York, USA.
S
CHWINNING
, S., K. D
AVIS
,L.R
ICHARDSON
,
AND
J. R. E
HLERINGER
. 2002.
Deuterium enriched irrigation indicates different forms of rain use in
shrub/grass species of the Colorado Plateau. Oecologia 130: 345–355.
S
OKAL
, R. R.,
AND
F. J. R
OHLF
. 1995. Biometry: the principles and practice
of statistics in biological research. W. H. Freeman and Company, New
York, New York, USA.
S
TERNBERG
,L.
DA
S. L., L
A
V. G
REEN
,M.Z.M
OREIRA
,D.C.N
EPSTAD
,L.
A. M
ARTINELLI
,
AND
R. V
ICTO
´RIA
. 1998. Root distribution in an Am-
azonian seasonal forest as derived from d
13
C profiles. Plant and Soil 205:
45–50.
S
TERNBERG
,L.
DA
S. L., M. Z. M
OREIRA
,
AND
D. C. N
EPSTAD
. 2002. Uptake
of water by lateral roots of small trees in an Amazonian tropical forest.
Plant and Soil 238: 151–158.
S
TERNBERG
,L.
DA
S. L.,
AND
P. K. S
WART
. 1987. Utilization of freshwater
and ocean water by coastal plants of southern Florida. Ecology 68: 1898–
1905.
T
AIZ
, L.,
AND
E. Z
EIGER
. 2002. Plant physiology, 3rd ed. Sinauer,Sunderland,
Massachusetts, USA.
T
URNER
,J.V.,P.F
ARRINGTON
,
AND
V. G
AILITIS
. 2001. Extraction and anal-
ysis of plant water for deuterium isotope measurement and application
to field experiments. In M. Unkovich, J. Pate, A. McNeill, and D. J.
Gibbs [eds.], Stable isotope techniques in the study of biological pro-
cesses and functioning of ecosystems, 37–55. Kluwer, Dordrecht, Neth-
erlands.
W
ALKER
,C.D.,
AND
S. B. R
ICHARDSON
. 1991. The use of stable isotopes
of water in characterizing the source of water in vegetation. Chemical
Geology (Isotope Geoscience Section) 94: 145–158.
W
ERSHAW
, R. L., I. F
RIEDMAN
,S.J.H
ELLER
,
AND
P. A. F
RANK
. 1970. Hy-
drogen isotopic fractionation of water passing through trees. In G. D.
Hobson and G. C. Speers [eds.], Advances in organic geochemistry, pro-
ceedings of the Third International Congress on Organic Geochemistry
1966, 55–67. Pergamon Press, New York, New York, USA.
W
HITE
, J. W. C., E. R. C
OOK
,J.R.L
AWRENCE
,
AND
W. S. B
ROECKER
. 1985.
The D/H ratios of sap in trees: implications for water sources and tree
ring D/H ratios. Geochimica et Cosmochimica Acta 49: 237–246.
W
ILLIAMS
, M., Y. E. S
HIMABUKURO
,D.A.H
ERBERT
,P.S.L
ACRUZ
,C.R
EN
-
NO
,
AND
E. B. R
ASTETTER
. 2002. Heterogeneity of soils and vegetation
in an eastern Amazonian rain forest: implications for scaling up biomass
and production. Ecosystems 5: 692–704.
W
ILLIAMSON
, G. B., W. F. L
AURANCE
,A.A.O
LIVEIRA
,P.D
ELAMO
ˆNICA
,C.
G
ASCON
,T.E.L
OVEJOY
,
AND
L. P
OHL
. 2000. Amazonian tree mortality
during the 1997 El Nin˜o drought. Conservation Biology 14: 1538–1542.
W
OODS
,F.W.,
AND
D. O’
NEAL
. 1965. Tritiated water as a tool for ecological
field studies. Science 147: 148–149.
W
RIGHT
, S. J. 1996. Phenological responses to seasonality in tropical forest
plants. In S. S. Mulkey, R. L. Chazdon, and A. P. Smith [eds.], Tropical
forest plant ecophysiology, 440–460. Chapman and Hall, New York,
New York, USA.
Y
AVITT
,J.B.,
AND
S. J. W
RIGHT
. 2001. Drought and irrigation effects on
fine root dynamics in a tropical moist forest, Panama´. Biotropica 33:
421–434.
Z
IMMERMANN
, U., D. E
HHALT
,
AND
K. O. M
U
¨NNICH
. 1967a. Soil-water
movement and evapotranspiration: changes in the isotopic composition
of the water. In Isotopes in hydrology symposium, proceedings of a sym-
posium in Vienna, Austria, 14–18 November 1966, 567–585. Interna-
tional Atomic Energy Agency, Vienna, Austria.
Z
IMMERMANN
, U., K. O. M
U
¨NNICH
,
AND
W. R
OETHER
. 1967b. Downward
movement of soil moisture traced by means of hydrogen isotopes. In G.
E. Stout [ed.], Isotope techniques in the hydrologic cycle: a symposium
at the University of Illinois, 10–12 November 1965, 28–36. Geophysical
Monograph Series 11, American Geophysical Union, Washington, D.C.,
USA.
Z
IMMERMANN
, U., K. O. M
U
¨NNICH
,W.R
OETHER
,W.K
REUTZ
,K.S
CHUBACH
,
AND
O. S
IEGEL
. 1966. Tracers determine movement of soil moisture and
evapotranspiration. Science 152: 346–347.
