Life form-specific variations in leaf water oxygen-18 enrichment in Amazonian vegetation

14 Pages·2008·0.78 MB·English

Oecologia(2008)157:197–210
DOI10.1007/s00442-008-1071-5
PHYSIOLOGICAL ECOLOGY - ORIGINAL PAPER
Life form-specific variations in leaf water oxygen-18 enrichment
in Amazonian vegetation
Chun-Ta Lai Æ Jean P. H. B. Ometto Æ
Joseph A. Berry Æ Luiz A. Martinelli Æ
Tomas F. Domingues Æ James R. Ehleringer
Received:21November2006/Accepted:14May2008/Publishedonline:10June2008
(cid:1)Springer-Verlag2008
Abstract Leaf water 18O enrichment (D ) influences the were detected in 18O leaf water values. Initial modeling
o
isotopic composition of both gas exchange and organic efforts to explain these observations over-predicted night-
matter, with D values responding to changes in atmo- time D values by as much as 10%. Across all species,
o o
sphericparameters.Inordertoexaminepossibleinfluences errors associated with measured values of the d18O of
of plant parameters on D dynamics, we measured oxygen atmospheric water vapor (d ) appeared to be largely
o v
isotope ratios (d18O) of leaf and stem water on plant spe- responsible for the over-predictions of nighttime D
o
cies representing different life forms in Amazonia forest observations. We could not eliminate collection or storage
and pasture ecosystems. We conducted two field experi- of water vapor samples as a possible error and therefore
ments: one in March (wet season) and another in developed an alternative, plant-based method for estimat-
September(dryseason)2004.Ineachexperiment,leafand ing the daily average d value in the absence of direct
v
stem samples were collected at 2-h intervals at night and (reliable) measurements. This approach differs from the
hourlyduringthedayfor50 hfromeightspeciesincluding common assumption that isotopic equilibrium exists
upper-canopy forest trees, upper-canopy forest lianas, and between water vapor and precipitation water, by including
lower-canopy forest trees, a C pasture grass and a C transpiration-based contributions from local vegetation
4 3
pasture shrub. Significant life form-related differences through18Omeasurements ofbulk leaf water.Inclusion of
both modified d and non-steady state features resulted in
v
model predictions that more reliably predicted both the
CommunicatedbyToddDawson.
magnitudeandtemporalpatternsobservedinthedata.The
C.-T.Lai(&) influence of life form-specific patterns of Do was incor-
DepartmentofBiology,SanDiegoStateUniversity, porated through changes in the effective path length, an
5500CampanileDrive,SanDiego,CA92182,USA
important but little known parameter associated with the
e-mail:[email protected]
Pe´clet effect.
J.P.H.B.Ometto(cid:1)L.A.Martinelli
CentrodeEnergiaNuclearnaAgricultura,Av.Centena´rio303, Keywords Craig–Gordon model (cid:1) Water vapor (cid:1)
Piracicaba,SPCep13416-000,Brazil
Isotopic non-steady state (cid:1) Transpiration (cid:1) Tropics
J.A.Berry
DepartmentofGlobalEcology,CarnegieInstitution
ofWashington,Stanford,CA,USA Introduction
T.F.Domingues
Quantifyingtemporalvariationsoftheoxygenisotoperatios
InstituteofGeography,SchoolofGeoSciences,
UniversityofEdinburgh,DrummondStreet, (d18O)ofleafwaterisofecologicalinterestbecauseofthe
EdinburghEH89XP,Scotland,UK role this parameter plays in its influences on the isotopic
compositionofCO andH Ofluxesingasexchangeandon
J.R.Ehleringer 2 2
organic matter (Farquhar and Lloyd 1993; Yakir 1998;
DepartmentofBiology,UniversityofUtah,257S1400E,
SaltLakeCity,UT84112-0840,USA HellikerandEhleringer2000;Dawsonetal.2002;Ometto
123
198 Oecologia(2008)157:197–210
et al. 2005; Lai et al. 2006a, b; West et al. 2006; Barbour inspectroscopictechniquearelikelytosomewhatalleviate
2007).Evaporation, driven byvapor pressure deficit ofthe thisdifficulty(Leeetal.2007).Whendirectobservationsare
air,discriminates againstH 18Ofluxes. Leaf waterthereby notavailable,afirstapproximationford valuesistoassume
2 v
becomes18O-enrichedrelativetothesourcewaterasratesof that precipitation and water vapor are in isotopic equilib-
transpirationincreasesduringtheday.Meanwhile,CO that rium. This d estimate is then assumed invariable over the
2 v
diffusesintoleafintercellularspaceandlateroutofstomata courseofadayformodelingD onhourlytimescales.The
o
become labeled by this enriched 18O signature. Improving assumptionthatanisotopicequilibriumexistsbetweenwater
ourunderstandingofleafwater18Oenrichmentwouldallow vapor and precipitation water and its influences on D cal-
o
for the development of better process-based models to culationhasnotbeenadequatelyevaluated.Second,littleis
investigatebiosphere–atmospherewaterandCO exchange known about potential life form-specific variations of leaf
2
processes spanning from ecosystem to global scales (Far- water 18O enrichment from plants grown in a common
quhar et al. 1993; Flanagan et al. 1997; Wang and Yakir environment. The Wang et al. (1998) survey of 90 plant
2000;Cuntzetal.2003;Laietal.2006a,b;Welpetal.2006). speciesgrowninacommongardenrepresentstheonlystudy
Leaf water 18O enrichment, a labile signal, is also an that examined non-climatic effects on the evaporative
importantdeterminantoftheultimateoxygenisotopecom- enrichmentofbulkleafwater.
position recorded in plant cellulose (Epstein et al. 1977; Ourobjectivesinthisstudyaretwofold:toelucidatethe
RodenandEhleringer1999a,b;Barbour2007),providinga variabilityintheobservedD valuesamongplantlifeform
o
valuable tool for environmental reconstructions (Epstein thatcannotbeexplainedbyclimaticvariablesalone,andto
etal.1977;Andersonetal.1998;Rodenetal.2000;Roden investigate a plant-based method for estimating the daily
andEhleringer2000). averaged intheabsenceofdirect(reliable)measurements.
v
Diurnal patterns of leaf water 18O enrichment above We conducted two intensive field experiments during
source water (D ), are primarily driven by climatic vari- the rainy and dry seasons in pasture and forest ecosystems
o
ables,suchasrelativehumidity(RH),temperature,andthe of central Amazonia. We selected eight species associated
d18O values of atmospheric vapor (Craig and Gordon with the six functional groups described by Domingues
1965). The Craig–Gordon model, assuming D values are (2005) at the top, the middle and bottom portion of the
o
dependanton instantaneousenvironmentalconditions(i.e., canopy;eachrepresentedadistinctmicroclimate.Contrasts
steady state), simulates the diurnal D pattern, but often in biotic factors are often related to life form-specific dif-
o
over-predicts the magnitude of D in field conditions ferences in physiological properties such as stomatal
o
(Dongmann et al. 1974; Bariac et al. 1989; Flanagan and conductance (g) and leaf anatomy. The latter affects the
s
Ehleringer 1991; Yakir 1992; Roden and Ehleringer pathway of water movement and turnover within leaves
1999a). This modeling discrepancy has been investigated (Gan et al. 2002). In addition to the collection of leaf and
with a focus on two aspects: evaluation of the steady-state plant source waters for stable isotope analyses, we also
assumption in contrasting environmental conditions (Ro- measured air temperature (T ), RH, d , g and leaf water
a v s
den and Ehleringer 1999a; Cernusak et al. 2002; Lai et al. contents. These measurements were used to model 18O
2006b; Seibt et al. 2006; Barnard et al. 2007), and quan- enrichmentofleafwater,fromwhichwegainedadditional
tification of the spatial heterogeneity of D within a leaf information to assess life form-specific patterns.
o
(Farquhar and Lloyd 1993; Farquhar and Gan 2003; Gan
et al. 2002, 2003; Barbour et al. 2004). New models that
considerthecombinedeffectofnon-steadystate(NSS)and
Materials and methods
progressive enrichment within a leaf (Helliker and Ehle-
ringer 2000; Gan et al. 2002, 2003) have been developed
Prediction of leaf water enrichment
andtested(FarquharandCernusak2005;Oge´eetal.2007).
Despite these recent advances, two critical aspects con-
A modified Craig–Gordon formulation (Craig and Gordon
cerningNSSinfluencesonvariationsinD arestillnotwell
o 1965) was used to predict isotopic enrichment of leaf
understood:theinfluenceoflifeformthroughitsimpactsof
water. Steady-state leaf water enrichment at the sites of
leaf and hydraulic properties and the influence of water
evaporation(D )wasmodeledaccordingtoFarquharetal.
vapor.First,d18Ovaluesofwatervapor(d )areinfrequently es
v (1989), given by:
measuredonshorttimescalesbecauseofthetimeneededto
e
acquireasampleofsufficientsize.Investigationsassociated D ¼e(cid:2)þe þðD (cid:3)e Þ a ð1Þ
es k v k e
with d variability and its interactions with the isotope i
v
composition of leaf water have long been limited by the where e* is the temperature-dependent equilibrium frac-
scarcityofmeasurements,exceptforsteadystatemeasure- tionation factor between liquid and vapor (Majoube 1971)
ments(e.g.,RodenandEhleringer1999a).Recentadvances and e is the kinetic fractionation factor that occurs during
k
123
Oecologia(2008)157:197–210 199
diffusion through the stomatal pore. The latter can be cal-
whereWisthelaminaleafwaterconcentration(mol m-2),
culatedforoxygenisotopesase ¼ð32r þ21r Þ=ðr þr Þ;
k s b s b a = 1 + e , a* = 1 + e*, and a a+ & 1. g is the total
where r and r are the stomatal and boundary layer resis- k k k
s b conductancetowatervaporofstomataplusboundarylayer
tances,respectively(Barbouretal.2004).D istheisotopic
v (mol m-2 s-1)andw ismolefractionofwatervaporinthe
discrimination between water vapor and the source water i
leaf intercellular air spaces (mol mol-1).
(D = R /R-1,whereRrepresentsthemolarratioofheavy
v v s
tolightisotopes,andsubscriptsvandsrepresentvaporand
Study sites and precipitation patterns
sourcewater,respectively).e /e istheratioofvaporpressure
a i
intheairtothatoftheleafintercellularspace.
The study sites were located south of Santare´m, Para´,
A modified Craig–Gordon model (Eq. 1) often overes-
Brazil at primary sites associated with the Large-Scale
timates observed leaf water enrichment (Dongmann et al.
Biosphere-Atmosphere Experiment Amazonia (LBA)-
1974; Bariac et al. 1989; Flanagan and Ehleringer 1991;
Ecology Project (ECO). The forest site was located at a
Yakir1992).FarquharandLloyd(1993)suggestedthatthis
primary forest in the Tapajo´s National Forest (2.85(cid:2)S,
discrepancy may be partly explained by the Pe´clet effect.
54.05(cid:2)W) with a mean canopy height of 35 m and emer-
The Pe´clet effect describes the convection of non-frac-
gent trees reaching up to 50 m. The pasture was *40 km
tionated water to the site of evaporation opposed by the
distant (2.77(cid:2)S, 54.58(cid:2)W) and was chosen because these
diffusionoftheenrichedwater.AccordingtoFarquharand
sites represent the major land-use change impacting the
Lloyd (1993), the average lamina water enrichment above
primary forests in Amazonia; pastures are dominated by a
source water at steady state (D ) can be calculated by
Ls non-native C grass (Brachiaria spp.). Information on soil
incorporating the Pe´clet ‘‘correction,’’ and is given by: 4
characteristics in the region can be found in Telles et al.
ð1(cid:3)e(cid:3)PÞ (2003). Briefly, soils are deeply weathered oxisols (Ha-
D ¼ D ð2Þ
Ls P es pludox)withhighclaycontent(60–80%),lowpH(4.0–4.3)
and low nutrient contents. Additional details are available
where P is the Pe´clet number, defined by P = LE/CD,
where E is leaf transpiration rate (mol m-2 s-1), C is the online at http://beija-flor.ornl.gov/lba/. The precipitation
molar concentration of water (5.55 9 104 mol m-3), D is regime in the region defines two distinct seasons: a rainy
the diffusivity of H 18O in water (2.66 9 10-9 m2 s-1), season from December up to and including July and a dry
2
season(\100 mmprecipitationpermonth)fromAugustto
andLisaneffectivepathlength(m).Lisaspecies-specific
November. The two 3-day experiments occurred between
parameter that may vary widely due to the tortuous nature
12–14 March and 17–19 September 2004, when leaf and
of water movement within a leaf (Barbour and Farquhar
stemsampleswerecollected10–15timesperdayforstable
2004). To our knowledge, L has never been directly mea-
isotope ratio analyses.
sured. The uncertainty associated with L and its impact on
the calculation of D has not been frequently tested. Few
Ls
Water sample collection and isotope analysis
studieshaveestimatedL,withvaluesvaryingfrom0.004to
0.166 (Flanagan et al. 1993, 1994; Wang et al. 1998).
Leafandstemwatersampleswerecollectedat2-hintervals
These authors used a common approach to calculate L by
atnightandhourlyduringthedayforatotalof50 hduring
matchingpredictedD withobservedvalues.Inthisstudy,
Ls
each experiment. In the forest, samples were taken from
we use a similar approach to estimate L.
vegetation adjacent to a 45-m tower where two overstory
NSS models have been proposed to simulate leaf water
trees (Manilkara huberi and Copaifera duckei) and two
enrichment(Dongmannetal.1974;White1983;Cernusak
lianas (Prionostemma aspera and Abuta rufescens) were
etal.2002;FarquharandCernusak2005).Dongmannetal.
sampledat38 m,amid-canopyspecies(Derrisamazonica)
(1974) proposed a transient model that considers NSS leaf
wassampledat14 m,andanunderstorytree(Ingasp.)was
water enrichment, given by:
sampledbetween0.5and1 m.Inthepasturetwospecies,a
DtL ¼DLsþðDtL(cid:3)1(cid:3)DLsÞ(cid:1)e(cid:3)dt=s ð3Þ C4 grass (Brachiaria brizantha) and a C3 shrub (Vismia
whereDt istheNSSleafwaterenrichmentattimet,D is sp.),weresampledataheightof1 m.Foreachspecies,we
predictedL steady-state leaf water enrichment, Dt(cid:3)1 isLtshe collected and pooled three to five stem samples from non-
L greentissuesandthenthreetofiveleavesfromthesesame
NSS leaf water enrichment at time t-1, dt is time interval
branches. After collection, leaf and stem samples were
(s), and s represents the turnover time of leaf water,
immediately placed inside separate individual glass vials
calculated by:
sealed with a screw cap and the vial-cap junction was
Wa a(cid:2)
s¼ k ð4Þ wrappedwithParafilm.Sampleswerekeptcold(0–5(cid:2)C)in
gw
i thefieldandlaterfrozeninthelaboratory.Waterwasthen
123
200 Oecologia(2008)157:197–210
extracted from these plant samples using cryogenic vac- heights in the forest and one height in the pasture for the
uumdistillationford18Oanalysesofstem(d)andleaf(d ) two study periods. There was a considerable vertical gra-
s L
waters (West et al. 2006). We express the measured 18O dient of T and RH within the forest. RH near the forest
a
enrichment of leaf water (D ) relative to stem water, cal- floor was considerably lower in the dry than in the wet
o
culated as: D ¼ðd (cid:3)d Þ=ð1þd =1000Þ: Hydrogen and season,aresultofdecreasedsurfacesoilmoisture.Diurnal
o L s s
oxygenisotoperatiosofallwatersamplesweredetermined temperature fluctuations were greater in the pasture than
using an online thermal conversion elemental analyzer– the forest ecosystem, with the pasture having higher day-
isotope ratio mass spectrometer process (Finnigan MAT, time T in March and lower nighttime T in September
a a
Bremen, Germany). whencomparedtotheforest.RHwasconsistentlylowerin
Atmosphericwatervaporwascryogenicallycapturedby thepastureduringthedaythanintheforest.NighttimeRH
pumping air through dry ice–ethanol cold traps at a flow values approached 100% in both ecosystems. These pat-
rateof1 ml s-1.Thisflowratewasatthelowerendofthe ternsofT andRHwereconsistentwithobservationsfrom
a
reported range when sampling apparatus was developed in micrometeorological towers at these sites.
the laboratory (Helliker et al. 2002). Water vapor samples
werecollectedfromthreeheightswithintheforestandtwo g and W
heights in the pasture. Sample collection time usually
ranges between 10 and 12 min. g and W values, expressed on a per leaf area basis, were
Wereportd18Omeasurementswithanoverallprecision seasondependent(Fig. 2a–h;Table 1).ThehighergandW
of ±0.2%. All observations are reported in the d notation values measured in all species during the wet season was
on the Vienna standard mean ocean water scale (Coplen anticipated, as were the reductions in both parameters
1996). We only present d18O measurements in this study. duringthedryseason.Largelifeform-specificvariationsof
Hydrogen isotope ratio data can be found at the NASA gandWwereobservedduringbothseasons.Upper-canopy
LBA-ECO website (http://beija-flor.ornl.gov/lba/). treesshowedhighergvaluesatmiddaycomparedtoupper-
canopy lianas and understory trees in the forest. Highest g
Meteorological and physiological measurements values were also observed in Vismia sp., a pasture shrub,
during the wet season (Fig. 2d). In contrast to Vismia sp.,
A hand-held probe sensor was used to measure T and RH the C grass, B. brizantha, had lower g values. As water
a 4
at heights where plant tissue samples were collected. became relatively more limiting, leaves of Vismia sp.
Values of e /e were estimated from T and RH measure- reduced g values by nearly fivefold while leaves of B.
a i a
ments by assuming leaf temperature equals T . Total leaf brizantha reduced g values by less than twofold (Fig. 2h),
a
conductance to water vapor (stomatal plus boundary-layer suggestingthatC shrubsweremoredrought-sensitivethan
3
conductance, g) was measured every 2 h from 0700 to C grasses in this pasture ecosystem.
4
1700 hours for all the species using a Li-Cor 6400 photo- W values varied from 4.4 to 13.3 mol m-2 in the wet
synthesis system (Li-Cor, Lincoln, Neb.) (Domingues season, and from 1.2 to 10.2 mol m-2 in the dry season
2005). The boundary-layer conductance was set to a con- (Table 1).Amongthecontrastinglifeforms,theunderstory
stant (1 mol m-2 s-1) in these measurements. We did not tree Inga sp. had lowest W values, which was expected
measuregatnight.Themiddayleafwatercontent(W)was sincethisspeciesappearedtohavethethinnestleavesofall
determined gravimetrically by measuring the difference the species measured in the study. All species had higher
between leaf fresh and dry weights, usually on several middayWvaluesduringthewetseason,butthemagnitude
leavescombinedperspecies.Thisway,weobtainedoneW ofchanges differed among species.All species maintained
valueforeachspeciesbutnowithin-speciesvariabilitycan higher midday W values even though conductances were
be determined. Measurements of g and W were used to higher in the wet season than in the dry season (Fig. 2).
evaluate leaf water enrichment with modeling. These physiological adjustments, along with changes in
environmental conditions, were likely to have contributed
to individual and seasonal variations of 18O enrichment in
Results bulk leaf water, because leaf water will reflect both turn-
over- and flux-related components.
Environmental conditions in the forest and the pasture
Measured d18O of plant source waters
The total monthly rainfall in March and September 2004
was 158 and 48 mm, respectively. This difference in pre- The d18O values of stem water varied from -5.8 ± 0.2%
cipitation was typical for the wet and dry seasons in (B. brizantha) to -1.9 ± 0.4% (Inga sp.) in March and
Amazonia. Figure 1 compares T and RHmeasured at two from -3.3 ± 0.3% (Vismia sp.) to 1.1 ± 0.1% (Inga sp.)
a
123
Oecologia(2008)157:197–210 201
Fig.1 Airtemperature(Ta)and March 12-14 September 17-19
40
relativehumidity(RH)
measuredwithintheforestand Forest 38m
inthepasturefortwostudy Forest 1m
periodsinMarch(wetseason) 36
pasture 1m
andSeptember(dryseason)
2004
32
C )
oa ( 28
T
24
20
100
90
% ) 80
H (
R
70
60
50
12 20 04 12 20 04 12 20 12 20 04 12 20 04 12 20
Time (hour)
Fig.2 Measuredleaf 0.3 0.8
a b c d
conductancefortheeightstudy 0.25 0.7
specieson13March(a–d)and 0.6
18September(e–h)2004 1) 0.2
-s 0.5
-2m 0.15 0.4
ol 0.1 0.3
m
0.2
ce ( 0.05 0.1
n
a 0.3 0.3
duct 0.25 e CMoapnailkifaerraa f APbriountaostemma g DInegraris 0.25 h BVrisamchiaiaria
n
o
c 0.2 0.2
af
e 0.15 0.15
L

0.1 0.1
0.05 0.05
0 0
4 8 12 16 20 8 12 16 20 8 12 16 20 4 8 12 16 20
Time (hour)
inSeptember(Table 1).Overall,theaveraged18Ovalueof weighted d18O values for Manaus (3.12(cid:2)S, 60.02(cid:2)W) of -
the plant source water was 3.0% more enriched in the dry 6.5% in March and -2.0% in September (source:
season than in the wet season. The International Atomic http://www-naweb.iaea.org/napc/ih/GNIP/IHS_GNIP.html
Energy Agency–Global Network of Isotopes in Precipita- ).Thisseasonaldifferenceof4.5%issimilarinmagnitude
tion (IAEA-GNIP) has reported long-term, precipitation- tothed18Odifferenceinplantsources.Withineachseason,
123
202 Oecologia(2008)157:197–210
Table1 Averageleafwatercontentandoxygenisotoperatios(d18O)ofstemwatermeasuredfortheeightstudyspecies
Species Functional Averageleaf Stemwater
group watercontenta d18Oaverage
(molm-2leaf) (SE)(%)
March September March September
Copaiferaduckei Overstorytree 9.6 7.8 -4.07(0.31) -2.04(0.28)
Manilkarahuberi Overstorytree 12.7 10.2 -5.14(0.30) -2.90(0.20)
Abutarufescens Overstoryliana 11.8 6.8 -3.32(0.38) 0.27(0.38)
Prionostemmaaspera Overstoryliana 13.3 2.4 -4.24(0.39) -0.74(0.17)
Derrisamazonica Mid-canopyliana 12.8 4.6 -2.98(0.31) 0.88(0.13)
Ingasp. Understorytree 4.4 1.2 -1.92(0.42) 1.10(0.06)
Brachiariabrizantha Pasturegrass 9.0 4.0 -5.78(0.23) -1.36(0.25)
Vismiasp. Pastureshrub 10 7.5 -4.86(0.16) -3.29(0.28)
a Nowithin-speciesvariabilitywasdeterminedforleafwatercontent(seetext)
xylem sap of upper-canopy trees had the lowest source Measured d
v
water d18O values, followed by upper-canopy lianas and
lastly understory species. The pasture grass had lower Atmospheric water samples collected throughout each of
source water d18O values than the shrub in the wet period the field campaign periods were at the minimal size nec-
and higher values in the dry period. essary for stable isotope analyses. Measured d values
v
rangedfrom-11.7to-7.7%intheforestandfrom-10.0
Measured 18O enrichment in leaf waters to -3.0% in the pasture in March, and from -7.9 to -
3.6%intheforestandfrom-8.8to-3.7%inthepasture
Figures 3 and 4 show diel patterns of D observed in the in September. These measured d values showed consid-
o v
wet and dry seasons, respectively. Midday leaf water erable hour-to-hour variation (data not shown). These
enrichment was observed for all the species except Inga apparent large changes contrasted with previous d mea-
v
sp.,anunderstorytreeinthedenseforest.ThesemiddayD surements in the Amazonia forests, which were based on
o
values were higher in the dry season, which can partly be larger sample volumes (Moreira et al. 1997). From a the-
explained by the lower RH encountered during the mea- oretical perspective, the d values seemed too high and
v
surement period (Fig. 1). Forest species, except one liana possibly reflected contaminated or partially enriched sam-
A.rufescens,exhibitedpre-dawnD valuesclosetozeroin ples. The IAEA/GNIP Manaus station reported weighted
o
thewetseason.Incontrast,forestspecieshadpre-dawnD d18O precipitation values as -6.5% in March and -2.0%
o
values that were elevated by 2–6% above zero in the dry in September. Using equilibrium fractionation factors (e )
k
season. The two pasture species showed diel D patterns of 9.3% in March and 9.2% in September, corresponding
o
similar to the forest species in both seasons, although the withaverageT valuesof26.0and27.7(cid:2)Crespectivelyfor
a
magnitudes of the patterns differed. the2 months,providedthebasisoftheexpectedd values.
v
We observed large differences in the diel pattern of D If we assumed that water vapor was in equilibrium with
o
among upper-canopy forest species. Considerable D dif- precipitation,thed18Ovalueofwatervapor(d )should
o v,equil
ferences were noted between liana and overstory tree have been ca. -15.8% in March and -11.2% in Sep-
functional groups. Because leaf samples were collected tember, respectively. Our measured d values were
v
from co-located branches from the same height, little/no considerably more enriched than predicted d values.
v,equil
changes occurred in general microclimatic conditions that Toexaminehowtherapidtransitionofd valuesinteracted
v
surround these leaves. Instead these observations suggest with leaf water18Oenrichmentat steady state, weinitially
biotic, life form-specific parameters were likely to have used the measured d values to model 18O enrichment of
v
contributed to differences of D among these Amazonian bulk leaf water.
o
forest species. Leaf water in the understory tree (Inga sp.)
showed no or little D enrichment (Fig. 3c). These D Modeling leaf water 18O enrichment with measured d
o o v
patterns were consistent with previous observations in the
Amazonian forest (Ometto et al. 2005). In contrast to the To model steady-state leaf water 18O enrichment at the
forest species, the pasture species showed nearly identical sites of evaporation (D ), we assumed nighttime val-
es
D patterns in both wet and dry seasons. ues of g = 0.01 mol m-2 s-1 for all C species and
o 3
123
Oecologia(2008)157:197–210 203
Fig.3 Measuredleafwater 12
d18Oenrichmentinthewet Copaifera a Abuta b
seasonforaupper-canopytrees, 10 Manilkara Prionostemma
bupper-canopylianas,andc
mid-canopyandunderstory 8
speciesintheforest,anddaC
4
grassandaC shrubinthe 6
3
pasture.Forclarity,only
averagevaluesofleafwater18O 4
enrichmentareshown
2
)
‰
( ο 0
∆
nt, -2
e
m
h -4
c
nri
e 12
18erO 10 DInegraris c BVrisamchiaiaria d
at
w 8
af
e 6
L
4
2
0
-2
-4
12 20 4 12 20 4 12 20 12 20 4 12 20 4 12 20
Time (hour)
0.1 mol m-2 s-1 for the C grass when observations were 2005). Thus, the measured d values were then
4 v
not available. These nighttime g values were in close reconsidered.
agreement with literature values assumed for C and C Inhumidenvironments,errorsassociatedwithmeasured
3 4
plants(Sellersetal.1996).Foreachspecies,wecalculated d values have largest impacts on D . Figure 6 shows the
v es
D to compare with observed D values (Fig. 5). sensitivity of D to changes of d under various RH con-
es o es v
Our calculations over-predicted observed D in all ditions. In this example, when RH = 0.1 (10%), a d
o v
instances; differences were most pronounced in the wet change from -5 to -15% results in a 1% decrease in the
season, especially in the pasture ecosystem where the value of D . When RH = 0.9 (90%), the same d change
es v
discrepancy reached 10% at night and was [10% at resultsinadecreaseoftheD valuebynearly9%.Hence,
es
midday (Fig. 5). The Pe´clet effect has been suggested as a we further examined whether potential errors associated
mechanism to partially explain differences in predicted withmeasuredd couldleadtothediscrepancyinpredicted
v
versus observed D values during the day (Farquhar and versus modeled values shown in Fig. 5.
o
Lloyd 1993), but its effect is expected to be negligible at
night because of low transpiration rates. Errors associated Estimating d
v
with RH measurements could potentially introduce large
discrepancies in the prediction of D (Roden et al. 2000; We estimated d values using a plant-based approach,
es v
Barbour et al. 2004). However, this was not the case here considering the equilibrium between leaf water and atmo-
becausetheprobe-basedRHmeasurementswereconfirmed spheric water vapor. The 18O abundance in the vapor is
bytower-basedobservations.Weassumedthatd18Ovalues usually less than that in the leaf water, because 18O has a
of leaf and stem water were correct, because replicates lower saturation vapor pressure than 16O. As RH
yielded similar values and because these values were approaches unity in the late afternoon, the equilibrium
similartopreviouslypublishedobservations(Omettoetal. process becomes dominant, which consequently decreases
123
204 Oecologia(2008)157:197–210
Fig.4 Measuredleafwater 16
d18Oenrichmentinthedry
Copaifera a Abuta b
seasonforthesamespeciesas 14 Manilkara Prionostemma
showninFig.3
12
10
8
6
)
‰
( 4
ο
∆
nt, 2
e
m
h 0
c
nri 16
e
O Derris c Brachiaria d
18 14 Inga Vismia
er
at 12
w
af
e 10
L
8
6
4
2
0
12 20 4 12 20 4 12 20 12 20 4 12 20 4 12 20
Time (hour)
the 18O content of leaf water. This exchange process can whether bulk leaf d values can be used as a reasonable
L
continue throughout the night if stomata remain partially proxy for the estimates of D . This requires that d be
es L
open. measured at a time when leaf water enrichment is at
On this basis, leaf water and atmospheric vapor may steady-state and that the Pe´clet effect is negligible. The
eventually approach a complete equilibrium sometime latter criterion implies low rates of leaf transpiration, most
duringthenight.Thisequilibriumprocesscanconveniently likely to be met during nighttime conditions.
beexplainedbyconsideringEq. 1.WhenRHequals100%, We assumed measured D values would be closest to a
o
the ratio e /e = 1, which reduces Eq. 1 to complete equilibrium at pre-dawn (0400–0600 hours),
a i
providing a reasonable proxy for D . We then calculated
D ¼e(cid:2)þD : ð5Þ es
es v
the expected d value at this time using Eqs. 5 and 6. We
v
UsingEq. 5,onecancalculateD ifD isknown.When obtained average d values of -13.4 ± 0.8% in the forest
v es v
preferred, d can be calculated instead of D using the and -14.9 ± 1.1% in the pasture for the wet season; -
v v
following relationship: 7.1 ± 1.1%intheforestand-11.2 ± 0.2%inthepasture
for the dry season. These estimates were similar to the
d ¼d (cid:3)ðd þ1Þ(cid:1)e(cid:2): ð6Þ
v L s d values (-15.8 and -11.2% for the two periods,
v,equil
From amodeling perspective,D representsteady-state respectively), assuming that atmospheric water vapor was
es
leaf water enrichment at sites of evaporation. From a in equilibrium with precipitation. The estimated forest d
v
measurement point of view, this variable is rarely directly value was more positive than d in the dry season (-
v,equil
measured. What was normally measured in the field is the 7.1vs.-11.2%).Interestingly,thisestimatewassimilarto
enrichment of bulk leaf water (D ). measured d values (-6.7 ± 1.5%, n = 9).
o v
A plant-based approach to estimate d relies on mea- Using estimated d and assuming that d remained
v v v
sured d and d values, two quantities that are relatively constantthroughouteachstudyperiod,were-calculatedD
L s es
easy to measure in field experiments. The question is and compared it to observed D in Fig. 5. These revised
o
123
Oecologia(2008)157:197–210 205
Fig.5 Comparisonsbetween March 12-14 September 17-19
modeledandmeasuredleaf 24
waterd18Oenrichmentfora predicted ∆es with measured δv a b
forestspeciesCopaiferaduckei 20 predicted ∆es with estimated δv
(a,b)andapasturespecies
Brachiariabrizantha(c,d).The measured ∆o
modifiedCraig–Gordonmodel 16
(Eq.1)wasusedhereto
demonstratetheinfluenceof 12
errorsassociatedwithmeasured
d18Oofwatervapor(d )on
v
modelpredictions.D 18O 8
es
enrichmentatsitesof ‰)
evaporation (nt 4
e
m
h 0
c
nri
O e c d
18er 20
at
w
af 16
e
L
12
8
4
0
12 20 4 12 20 4 12 20 12 20 4 12 20 4 12 20
Time (hour)
modeled D values showed greater agreement with the D and modeled D values for the two growing seasons
es Ls L
observed nighttime measurements, suggesting that errors (Figs. 7, 8). The NSS model appeared to explain observed
associated with measured d may have been the basis for D values better than thesteadystate model for M.huberi,
v o
why model and observation did not initially agree. How- A. rufescens, and P. aspera, while the steady state model
ever, the re-calculations still over-predicted daytime D explained observed D values better than the NSS model
o o
values, suggesting a second factor not incorporated that for D. amazonica, Vismia sp., C. duckei and B. brizantha
was responsible for the midday enrichment (the Pe´clet (seecomparisonsinFig. 5forthelattertwospecies).Both
effect). Hereafter, we use estimated d and include the modelspredictedrelativelysmall18Oenrichmentinleaves
v
Pe´clet effect in the model to examine biotic influences on of Inga sp., consistent with observed D values. We cal-
o
daytime patterns of D . culated L values for all the forest and pasture species
o
except Inga sp., which showed no/little enrichment during
Biotic effects on leaf water 18O enrichment our experiment. The value of L was determined by mini-
mizing root mean square errors between predicted and
To evaluate life form-specific variations of and biotic observed D at midday (1200–1600 hours).
o
effects on leaf water enrichment, we consider the Pe´clet
effect when predicting D and D for each species. To
Ls L
contrast individual differences, we focused on two com- Discussion
parisons among life forms: steady-state versus NSS
patternsofleafwaterenrichment,anddifferencesintheL. Life form-specific variations of leaf water 18O
We used different L values for each species when calcu- enrichment
lating D and D (discussed later).
Ls L
Two distinct groupings of enrichment patterns can be Biotic influences on leaf water 18O enrichment depend on
seen when measured D values are compared to modeled lifeform-specificfactorssuchasLandleaf-waterturnover
o
123
206 Oecologia(2008)157:197–210
40 over-simplifiedparameterizationasmodeledDLonlypartly
explains nighttime D . Dawson et al. (2007) showed that
35 RH = 0.1 nighttime transpiratioon generally occurred when nighttime
30 RH = 0.3 atmosphericpressuredeficit(VPD)exceeded*0.2 kPafor
plants inhabiting ecosystem types without soil water limi-
)
(‰ 25 R H = 0.5 teaxtcioened.eDdur0i.n2gkoPuartiwnothsetsuedympoeisritotdros,pnicigalhtetcimoseyVstePmDsn.eTvheer
s
e 20
∆ absence of a strong VPD likely prohibits significant
15 R H = 0.7 nighttimetranspirationinthestandsstudied,assumingthat
leaf temperature approximately equals T .
10 H = 0.9 The calculated L values should be intaerpreted as initial
R
estimates,becauseoftheassumptionofaconstantd value.
v
5
Given the uncertainty in these estimates, only C. duckei
and D. amazonica showed significant differences of L
0
-40 -35 -30 -25 -20 -15 -10 -5 0 between wet and dry periods. The high L value for C.
δ (‰) duckei in the wet period was perhaps the calculation with
v
lowest confidence,withuncertainties1orderofmagnitude
Fig.6 ThesensitivityofCraig–GordonmodeledD tochangingd greater than other estimates. Excluding L values for these
es v
indifferentRHconditions.Inthisexample,wekeptDes=37.74%at two species, average L values fall within a small range
d =0% by using a value of 0.0316 for the kinetic fractionation
v (0.04–0.15 m) for forest species. The average L values for
factor (e ) and a value of 0.0093 for the equilibrium fractionation
k the pasture species were similar in both seasons, ranging
factor between liquid and vapor (e*) when RH=10%. These
fractionation factors were average values representing of the condi- from 0.01 to 0.04 m. These results fall within the range
tionsduringourexperiments.Valuesofe*wereallowedtoincrease reportedfor 90 different plantspecies grownin acommon
by 0.1e for every 10% increase of RH to keep a constant D at
k es garden by Wang et al. (1998).
d =0%.Forotherabbreviations,seeFigs.5and 1
v We used 1 SD of the daily-average midday D values,
o
ranging from 1 to 3% depending on species, to estimate
rate(afunctionofleafg andleafwatercontent).Together, uncertainties with respect to the calculated L values. This
s
leaf g and L are determinants of the Pe´clet effect. rangeofuncertaintyisequivalenttoa2–5%changeinthe
s
NSS leaf water enrichment has been shown to explain d value, if diurnal variations were to be accounted for. In
v
under-estimates of predicted nighttime D values in leaves other words, the uncertainty estimates shown in Table 2
o
of Lupinus angustifolius (Cernusak et al. 2002; Farquhar also address the sensitivity of L calculation to the
and Cernusak 2005) and over-predictions of daytime D assumption of a constant d value in the model.
o v
values in needles of Pseudotsuga menziesii (Douglas-fir)
under water-stress conditions (Lai et al. 2006b). Based on An alternative approach to estimate water vapor d18O
model comparisons (Figs. 5, 7, 8), we found that: (1) the
steady state modelwas suitable to explain D observations Weassumethatanequilibriumestablishesatnightbetween
o
made in the pasture ecosystem, likely because of the high leaf water and atmospheric water vapor. This assumption
rates of water turnover in leaves; (2) the NSS model was may be supported by a growing number of studies indi-
suitable to explain D observations made in the two lianas cating that in many woody plants stomata remain partially
o
which had relatively lower leaf-level water turnover rates; open at night (Donovan et al. 1999; Burgess and Dawson
and (3) the two overstory forest trees had contrasting pat- 2004; Barbour et al. 2005; Dawson et al. 2007). Dawson
terns. The model predictions also suggest that leaf water etal.(2007)usedsapflowanddeuteriumtracertechniques
enrichmentatmiddaywasatanapproximatesteadystatein to measure nighttime g and transpiration from woody
s
all the species. The difference between modeled D and plants grown in an array of environments, including
Ls
D values became relatively indistinguishable at midday, Amazoniaspecies.Theyconcludedthatnighttimestomatal
L
when leaf water turns over more quickly. conductance to water vapor is widespread among woody
The NSS influence was most pronounced at night when plant species that inhibit a broad range of environments.
leaf conductances were reduced; inclusion of NSS effects Althoughwedidnotdirectlymeasurenighttimeg inthe
s
explained nighttime differences in the observed D values current study, flask measurements made at 21 and 45 min
o
among the studied species. We cannot attribute the night- theforestshowedenricheddC18OOvalueswhencompared
time difference to the Pe´clet effect because we assume tothoseneartheground(J.Berry,unpublisheddata).These
nighttime values of g = 0.01 mol m-2 s-1 for all C enriched dC18OO values are most likely a result of an
3
species in the NSS calculation. This is obviously an equilibration process at night between CO molecules and
2
123