1 23
Oecologia ISSN 0029-8549 OecologiaDOI 10.1007/s00442-012-2263-6
Variation in foliar nitrogen and albedoin response to nitrogen fertilization andelevated CO2
Haley F. Wicklein, Scott V. Ollinger,Mary E. Martin, David Y. Hollinger,Lucie C. Lepine, Michelle C. Day, MeganK. Bartlett, Andrew D. Richardson, et al.
1 23
Your article is protected by copyright and
all rights are held exclusively by Springer-
Verlag. This e-offprint is for personal use only
and shall not be self-archived in electronic
repositories. If you wish to self-archive your
work, please use the accepted author’s
version for posting to your own website or
your institution’s repository. You may further
deposit the accepted author’s version on a
funder’s repository at a funder’s request,
provided it is not made publicly available until
12 months after publication.
Oecologia
Author's personal copy
DOI 10.1007/s00442-012-2263-6
PHYSIOLOGICAL ECOLOGY - ORIGINAL RESEARCH
Variation in foliar nitrogen and albedo in response to nitrogen fertilization and elevated CO2
Haley F. Wicklein · Scott V. Ollinger · Mary E. Martin · David Y. Hollinger · Lucie C. Lepine · Michelle C. Day · Megan K. Bartlett · Andrew D. Richardson · Richard J. Norby
Received: 21 January 2011 / Accepted: 15 January 2012© Springer-Verlag 2012
Abstract Foliar nitrogen has been shown to be positivelycorrelated with midsummer canopy albedo and canopy nearinfrared (NIR) reXectance over a broad range of plant func-tional types (e.g., forests, grasslands, and agriculturallands). To date, the mechanism(s) driving the nitrogen–albedo relationship have not been established, and it isunknown whether factors aVecting nitrogen availabilitywill also inXuence albedo. To address these questions, weexamined variation in foliar nitrogen in relation to leafspectral properties, leaf mass per unit area, and leaf watercontent for three deciduous species subjected to eithernitrogen (Harvard Forest, MA, and Oak Ridge, TN) or CO2
fertilization (Oak Ridge, TN). At Oak Ridge, we alsoobtained canopy reXectance data from the airborne visible/infrared imaging spectrometer (AVIRIS) to examinewhether canopy-level spectral responses were consistentwith leaf-level results. At the leaf level, results showed nodiVerences in reXectance or transmittance between CO2 ornitrogen treatments, despite signiWcant changes in foliarnitrogen. Contrary to our expectations, there was a signiW-cant, but negative, relationship between foliar nitrogen andleaf albedo, a relationship that held for both full spectrumleaf albedo as well as leaf albedo in the NIR region alone.In contrast, remote sensing data indicated an increase incanopy NIR reXectance with nitrogen fertilization. Collec-tively, these results suggest that altered nitrogen availabilitycan aVect canopy albedo, albeit by mechanisms that involvecanopy-level processes rather than changes in leaf-levelreXectance.
Keywords Albedo · Nitrogen · Leaf structure · Nitrogen fertilization · Free air CO2 enrichment
Introduction
The relationship between mass-based foliar nitrogen(N) and leaf-level photosynthetic capacity has been glob-ally documented across a wide range of plant species (Fieldand Mooney 1986; Evans 1989; Reich et al. 1997, 1999;Wright et al. 2004), and a similar trend has been observedat the canopy level for boreal and temperate forests (Ollin-ger et al. 2008). Recently, Ollinger et al. (2008) and Hollin-ger et al. (2010) demonstrated that both of these variablesare also signiWcantly and positively correlated with fullspectrum canopy albedo. Given the importance of evensmall changes in albedo on surface heat exchange, the
Communicated by Gerardo Avalos.
H. F. Wicklein (&) · S. V. Ollinger · M. E. Martin · L. C. Lepine · M. C. DayComplex Systems Research Center, Morse Hall, Institute for the Study of Earth, Oceans, and Space, University of New Hampshire, 8 College Rd, Durham, NH 03824, USAe-mail: [email protected]; [email protected]
D. Y. HollingerNorthern Research Station, US Department of Agriculture Forest Service, Durham, NH 03824, USA
M. K. BartlettDepartment of Environmental Science, Policy, and Management, University of California, Berkeley, CA 94720, USA
A. D. RichardsonDepartment of Organismic and Evolutionary Biology, Harvard University Herbarium, Harvard University, 22 Divinity Avenue, Cambridge, MA 02138, USA
R. J. NorbyEnvironmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA
123
Oecologia
Author's personal copy
occurrence of an N eVect on albedo would bear interestingand potentially important consequences for climate. How-ever, resolving underlying mechanisms of the N–albedorelationship will be important because diVerent mecha-nisms can carry diVerent implications for a response toaltered N availability.
Sources of variation in vegetation albedo occur at scalesranging from internal leaf structures to tree crowns andwhole plant canopies (Ollinger 2011). At the leaf level,reXectance is dominated by photosynthetic pigments in thevisible part of the spectrum (400–700 nm), leaf structure inthe near infrared (NIR, 700–1,350 nm), and water contentin the mid-infrared (Mid IR, 1,350–2,500 nm) (Gates et al.1965; Slaton et al. 2001; Jacquemoud et al. 2009). At thestem and canopy level, the number of scattering or absorb-ing surfaces that photons encounter is further inXuenced bystructural traits such as shoot architecture, leaf angle distri-bution, and crown geometry (e.g., Chen and Cihlar 1995;Asner 1998; Rautiainen et al. 2004). The greatest variationin reXectance at both leaf and canopy levels often occurs inthe NIR region because there are few, if any, compoundsthat absorb NIR radiation (Gates et al. 1965; Sánchez andCanton 1999; Ollinger 2011). Therefore, if structuralchanges that occur within or between leaves covary withfoliar N, they could explain the nature of the canopy-levelN–albedo relationship.
Hollinger et al. (2010) hypothesized that the associationbetween leaf N concentration (Nmass) and canopy albedomay be due to covariation between N-containing photosyn-thetic enzymes and internal leaf structures necessary to sup-port diVerent rates of photosynthesis. Particularly importantare changes in the ratio of the mesophyll surface areaexposed to intercellular air spaces to the area of the leaf(Ames/A), which has been shown to be positively correlatedwith both photosynthetic rates (Nobel et al. 1975; Longst-reth et al. 1985) and NIR reXectance (Slaton et al. 2001).Given the diVerent refractive indices of hydrated mesophyllcells and the intercellular airspace (Woolley 1971; Gaus-man et al. 1974), a higher Ames/A value should lead to moreopportunities for radiation scattering, and correspondinglyhigher reXectance.
The objectives of this study were (1) to test the hypothe-sis that the canopy level association between foliar Nmass
and canopy albedo stems from a similar relationship occur-ring at the leaf level, and (2) to determine whether twoforms of disturbance that are known to aVect N availabil-ity—N fertilization and elevated CO2—can also inXuenceleaf and canopy albedo. We measured leaf reXectance andtransmittance, as well as leaf chemical and structural traits(N concentration, leaf mass per area, equivalent waterthickness, and water content), for three deciduous speciesin the eastern US that have been subjected to either long-term N or CO2 fertilization. Data from multiple free air CO2
enrichment (FACE) sites have shown that elevated CO2
leads to an increase in leaf mass per unit area (LMA, Norbyet al. 2003) and the consequent dilution of foliar N due tothe accumulation of carbohydrates (Oren et al. 2001; Ells-worth et al. 2004; Norby and Iversen 2006). Therefore, ifthe canopy albedo–N relationship is driven by changes atthe leaf level, we would expect to see higher leaf-level leafalbedo in plots subjected to N fertilization and lower leafalbedo in plots exposed to elevated CO2, relative to thosereceiving ambient CO2 and N deposition. We also obtainedcanopy reXectance data from the airborne visible/infraredimaging spectrometer (AVIRIS) to examine whether can-opy-level spectral responses were consistent with leaf-levelresults.
Materials and methods
Study sites
We measured spectral, chemical, and structural characteris-tics of leaf samples from two sites in the eastern US: Har-vard Forest, Petersham, MA (42.5°N, 72°W) and OakRidge National Laboratory, Roane County, TN (35.9°N,84.3°W). These sites were chosen because they containdeciduous tree species relevant to our objectives andbecause they allowed us to examine the spectral response ofleaves to N fertilization and elevated CO2, both of whichrepresent important environmental change agents that areknown to alter leaf Nmass.
Harvard forest, MA
Harvard Forest (HF) is located in Central Massachusettsand has been a long-term ecological research site since1988. A chronic N fertilization experiment was establishedat HF in 1988 (Magill et al. 2004). Two stands were chosenfor N additions: a mixed hardwood stand that regeneratednaturally after a clearcut around 1945, and an even-aged redpine (Pinus resinosa Aiton) stand that was heavily dis-turbed by an ice storm in December of 2008, and not usedin this study. In each stand, four plots were established:control (no added N), low N (additions of 50 kgN ha¡1 year¡1), low N plus sulfur (not included in thisstudy), and high N (additions of 150 kg N ha¡1 year¡1).Each plot measures 30 £ 30 m. Additions of ammoniumnitrate (NH4NO3) began in 1988 and are distributed oversix equal applications during the growing season (May–September). Mean annual precipitation at HF is 1,100 mm,distributed evenly throughout the year. Ambient N deposi-tion averages 8 kg N ha¡1 year¡1 (Ollinger et al. 1993).The dominant soil types are stony- to sandy-loams formedfrom glacial till. Elevation is 385 m above sea level.
123
Oecologia
Author's personal copy
Oak Ridge National Environmental Research Park, TN
The free air CO2 enrichment (FACE) facility at the OakRidge National Laboratory (ORNL) EnvironmentalResearch Park is located in a sweetgum (LiquidambarstyraciXua L.) monoculture that was established in 1988(Norby et al. 2001). In 1996, Wve 25-m-diameter plots wereestablished: two with FACE apparatus emitting elevatedCO2, two with FACE apparatus but ambient CO2, and onewith no FACE apparatus. Exposure to elevated CO2 beganin the spring of 1998. Average daytime CO2 concentrationfor 2009 was maintained at approximately 565 ppm for theenriched plots and 401 ppm for the ambient plots (Riggset al. 2009). In 2004, an N fertilization experiment was ini-tialized in a sweetgum stand located approximately 150 mfrom the FACE site and planted at the same time. An85 m £ 50 m area was fertilized in a block pattern, witheach block containing two control plots and two fertilizedplots (both 12 £ 16 m). The fertilized plots received200 kg N ha¡1 year¡1, applied as urea each year before leafout (Iversen and Norby 2008). Mean annual temperature atOak Ridge is 14°C, and mean annual precipitation is1,371 mm, distributed evenly throughout the year. AmbientN deposition averages between 10 and 15 kg N ha¡1 year¡1
(Johnson et al. 2004). The dominant soil type is an AquicHapludult, a silty clay loam.
Data collection and analysis
Field sampling: Harvard Forest
At HF, sampling of foliage from the control and treatedhardwood plots was conducted between 20 and 23 July2009. Within each plot, Wve red maple and seven black oaktrees were randomly selected and sampled. Green leaveswere collected from the top, middle, and bottom of the can-opy using a 12-gauge shotgun. Sample collection heightswere determined using a digital hypsometer (Haglöf Ver-tex). Leaves were placed in plastic Ziploc bags and storedon ice until analysis, which was carried out within 36 h ofcollection.
Field sampling: Oak Ridge National Environmental Research Park
At ORNL, Weld sampling was conducted between 28 and30 July 2009. Using a slingshot canopy sampler (N-fertil-ized site) or tower climbing with pole pruners (FACE site),we collected green leaves from the top, middle, and bottomof the canopy. Heights were determined using either a digi-tal hypsometer (Haglöf Vertex) or measuring tapesdeployed by climbers. Within each CO2 fertilization treat-
ment (ambient and elevated CO2), we sampled ten sweet-gum trees. From each N-fertilized plot, we collected 12upper canopy and 12 lower canopy samples. The adjacentFACE ambient CO2 plots were used as the control treat-ment for the N-fertilized samples in statistical analysis. Inall cases, leaves were placed in plastic Ziploc bags andstored on ice until analysis.
Collection of Weld spectra
We measured hemispherical reXectance and transmittancespectra for healthy leaves from each individual using aportable spectrometer (ASD FieldSpec 3; analyticalSpectral Devices, Boulder, CO, USA) connected to an inte-grating sphere with an 8° near-normal incidence port tocapture both diVuse and specular reXectance (SphereOptics,Concord, NH, USA) and a halogen bulb light source. TheASD spectrophotometer was used in conjunction with anintegrating sphere, which holds leaf samples in a Wxed posi-tion and is intended to eliminate angular eVects by creatingperfectly diVuse radiation. It measures reXectance from 350to 2,500 nm, in 1-nm intervals. For each sample, measure-ments were taken for a single leaf and stacks of two, four,and eight leaves. ReXectance and transmittance spectra ofleaf stacks were taken to create optically dense layers ofleaf tissue, in order to examine how reXectance varies as afunction of the total thickness of the leaf material present.Each leaf spectrum was determined as the average of 50individual scans. The spectra were corrected for dark cur-rent, and a white reference standard was measured prior toeach set of reXectance or transmittance measurements ofone growing stack of leaves. For reXectance measurements,both the sample and the reference standard were mountedon the sphere and their positions were switched betweentwo successive scans. Spectral reXectance was then calcu-lated as the ratio of the value obtained from the sample tothe value obtained from the reference.
To calculate full spectrum (FULL, 400–2,500 nm) spec-trally weighted optical values, each reXectance or transmit-tance spectra was weighted by the solar spectrum energy(ASTM G173-03 Reference Spectra received by a 37° tiltedequator-facing surface through an air mass of 1.5, derivedfrom SMARTS v. 2.9.2; Gueymard 2004) to obtain a valuerepresenting the reXected (denoted as �FULL) or transmitted(denoted as �FULL) energy as a proportion of incident. Thesevalues approximate albedo values measured using broad-band radiometric instruments. For visible (�vis or �vis, 400–700 nm), NIR (�nir or �nir, 700–1,300 nm), and Mid IR (�midir
or �midir, 1,350–2,500 nm) portions of the spectrum, thesolar-weighted reXectance and transmittance values werecalculated by following the same process, but using onlywavelengths from the regions of interest.
123
Oecologia
Author's personal copy
Leaf chemical and structural analysis
After spectral measurements were obtained, two circulardisks (2.035 cm2 area) were removed from each leaf andweighed. Leaf disks were then oven-dried at 70°C for atleast 72 h and reweighed to determine water content (% leaffresh weight), equivalent water thickness (EWT, g waterper cm2 leaf), and LMA (g leaf per m2 leaf).
The remaining leaf sample was then dried at 70°C,ground using a Wiley mill, and passed through a 1-mmmesh screen. Prior to N concentration analysis, the groundsamples were dried for 24 h at 70°C. We measured mass-based foliar N concentration (Nmass, g of N per 100 g of dryleaf matter) using a Costech Elemental Analyzer. We deter-mined N per unit leaf area (Narea) by multiplying Nmass bythe LMA of the sample (Narea = Nmass £ LMA).
Aircraft remote sensing
For the ORNL site, high spectral resolution remote sens-ing images were acquired from NASA’s AVIRIS (Air-borne Visible/InfraRed Imaging Spectrometer) instrumentwithin 10 days of our Weld sampling. AVIRIS measuresupwelling radiation in 224 contiguous optical bands from400 to 2,500 nm with a spectral resolution of 10 nm.AVIRIS was Xown on an ER-2 at an altitude of 5,500 mgiving a spatial resolution of 4.6 m. The image datasetwas orthorectiWed by NASA JPL, and atmosphericallycorrected with ImSpec LLC’s Atmospheric CorrectionNow (ACORN) v.6, transforming data from calibratedsensor radiance to apparent surface reXectance. Afterremoving strong water adsorption bands that contained nousable data, we determined canopy reXectance for eachtreatment by averaging reXectance values within eachwaveband for pixels falling within each of the plots. Dueto the spatial resolution of the image and the size of thetreatment plots, we were able to obtain 36 pixels in theambient CO2 plot, 12 pixels in the elevated CO2 plot, and24 pixels in the N-fertilized.
Statistical analysis
Summary statistics (means and standard error) were com-puted for all optical properties and leaf traits. The signiW-cances of the mean diVerences between treatments weredetermined by analysis of variance (ANOVA), with pair-wise comparisons tested using Tukey’s ‘Honest SigniWcantDiVerence’ method. Regression analysis was used to deter-mine relationships between optical properties and leaftraits. For multiple regressions the adjusted r2 was consid-ered instead of r2 because this statistic penalizes the modelfor an increased number of parameters, thereby decreasingthe likelihood of overWtting.
The Shapiro–Wilk statistic was used to test for normalityin all models, and where needed variables were power-transformed to correct for skew. Single leaf �FULL and �nir
were non-normal due to a small number of outliers (deter-mined by 1.5 £ interquartile range). Statistical tests wereperformed with and without the outliers, and, because theresulting explained variances were similar, outliers wereremoved to follow assumptions of normality in linearregression models. All statistical analysis was completedusing the software R, v.2.8.1 (R Foundation for StatisticalComputing, 2008). Reported results are for single leavesunless otherwise speciWed.
Results
Treatment diVerences
Across all species and sites, Nmass was higher in the high Ntreatment than in the low N and control treatments, whichwere not signiWcantly diVerent from each other (P < 0.001;Fig. 1a–c). For red maple and black oak at HF, there wereno diVerences between nitrogen treatments in �FULL
(Fig. 1d, e), �FULL, �nir, �nir, �midir, �midir, or FULL absorp-tion [1 ¡ (� + �)] (P > 0.1 in all cases). For the sweetgum atORNL, there were no diVerences between treatments in�FULL (except for higher mean values in the ORNL elevatedCO2 treatment than ORNL N-fertilized treatment; Fig. 1f),�FULL, �nir, or FULL absorption (P > 0.1 in all cases). Therewas higher �nir in the ORNL N-fertilized treatment than inthe ambient or elevated CO2 treatments (P < 0.01). Both�midir and �midir were higher for N-fertilized sweetgums(P < 0.01), likely due to diVerences in LMA and EWT. Forall species, �vis and �vis followed diVerences in N betweentreatments, with lower values corresponding to higher Nfertilization (P < 0.05 in all cases). There was no diVer-ence in LMA between nitrogen treatments for black oakor red maple (P > 0.1; Fig. 1g, h). However, the N-fertil-ized treatment at ORNL had lower LMA than the ambi-ent or elevated CO2 treatments (P < 0.001; Fig. 1i). EWTdid not diVer between treatments for any species(P > 0.1).
Relationships between N and leaf optical properties
The relationships between Nmass or Narea and the � and � ofall optical regions were qualitatively similar across sitesand treatments, although Narea explained less of the variancein foliar optical properties than did Nmass (Table 1). Varia-tion in Narea can be caused by changes in LMA or Nmass,although Nmass and LMA generally exhibited oppositetrends (e.g., LMA is positively correlated with �vis, whilethe relationship between Nmass and �vis is negative). Thus, it
123
Oecologia
Author's personal copy
is likely that the similarities between the Nmass and Narea
relationships are driven by changes in Nmass, but dampenedby the opposing eVects of LMA. Because the changes in
Narea can be accounted for by these two variables, subse-quent analysis and discussion of foliar N and optical prop-erties will focus on Nmass. Furthermore, Nmass was not relatedto canopy position (height from which foliage was sampled)for any of the species or treatments in our analysis (P > 0.1 inall cases), except for a weak, negative correlation in the ele-vated sweetgum treatment (r2 = 0.19, P < 0.05). As a result,we used mean values from all heights within a tree for subse-quent analyses involving Nmass.
Across all species, Nmass was the best predictor of FULL,NIR, and visible � and � for all species combined (Fig. 2),with the exception of �FULL, which was not correlated withNmass (P > 0.05). However, contrary to our expectations, therelationships between Nmass and � were negative (Fig. 2a–c;P < 0.001 in all cases). There was a positive relationshipbetween both �FULL and �nir and Nmass, whereas the relation-ship between Nmass and �vis was negative (Fig. 2d–f;P < 0.001 in all cases). The relationships between Nmass and� in the FULL and NIR regions are primarily due to diVer-ences between species [within species, these relationshipswere not signiWcant (P > 0.1)], with the exception of sweet-gum �FULL which showed a slight negative relationshipwith Nmass (P < 0.05, r2 = 0.06), whereas the relationships
Fig. 1 ANOVA results (means § standard error) for Nmass (a–c), solarweighted full spectrum reXectance (�FULL; d–f), and LMA (g–i) treatmentdiVerences. Results for black oak (site: HF) are depicted with black
bars, red maple (site: HF) in gray bars, and sweetgum (site: ORNL) inwhite bars. Means with diVerent letters were signiWcantly diVerent inpair-wise comparisons (Tukey’s multiple comparison test)
Cont. LowN HighN
Cont. LowN HighN Cont. LowN HighN Cont. LowN HighN
Cont. LowN HighN Cont. LowN HighN
Leaf
Nm
ass
(%)
0
1
2
3
4 aa a
b
Ref
lect
ance
FU
LL
0.0
0.1
0.2
0.3
0.4 d
LMA
(g
m−2
)
0
20
40
60
80
100 g
Leaf
Nm
ass
(%)
0
1
2
3
4 b
a ab
Ref
lect
ance
FU
LL0.0
0.1
0.2
0.3
0.4 e
LMA
(g
m−2
)
0
20
40
60
80
100 h
Elev Amb N Fert
Treatment
Leaf
Nm
ass
(%)
0
1
2
3
4 c
a a
b
Elev Amb N Fert
Treatment
Ref
lect
ance
FU
LL
0.0
0.1
0.2
0.3
0.4 fa ab b
Elev Amb N Fert
TreatmentLM
A (
g m
−2)
0
20
40
60
80
100 i aa
b
Table 1 Comparison of regression statistics for spectrally weightedreXectance and transmittance foliar N on a mass (Nmass, %) and area basis(Narea, g m¡2), reporting the coeYcient of determination (r2), P value, andthe sign of the slope of the regression line (‘Sign’ column) for each model
ReXectance and transmittance values are all weighted by the solarspectrum
ns insigniWcant trends
Response variable Source of variation
r2 Nmass r2 Narea
P value Sign P value Sign
FULL reXectance 0.26 <0.001 ¡ 0.15 <0.001 ¡FULL transmittance ns ns
NIR reXectance 0.17 <0.001 ¡ 0.06 <0.01 ¡NIR transmittance 0.10 <0.001 + 0.05 <0.01 +
Mid IR reXectance ns 0.09 <0.001 ¡Mid IR transmittance 0.21 <0.001 + ns
VIS reXectance 0.51 <0.001 ¡ 0.17 <0.001 ¡VIS transmittance 0.45 <0.001 ¡ 0.31 <0.001 ¡
123
Oecologia
Author's personal copy
between Nmass and the visible region are signiWcant evenwithin species (P < 0.05 in all cases).
Although Nmass was not a good predictor of �midir or �midir,the relationship was signiWcant and positive for �midir
(r2 = 0.21, P < 0.001). There was no relationship betweenNmass and �midir (P > 0.1). Absorption was positively corre-lated with Nmass in the visible (r2 = 0.52, P < 0.001) and theMid IR (r2 = 0.17, P < 0.001), but not in the NIR (P > 0.1),leading to a weak, positive relationship between totalFULL absorption and Nmass (r2 = 0.07, P < 0.01). Therewas some colinearity between spectral response variables,which is not surprising, as they are all part of the sameoverall spectra. However, the diVerent regions respondeddiVerently, because absorbers are located primarily in thevisible (pigment absorption) and Mid IR (water absorp-tion), with the NIR being primarily inXuenced by scattering(Ollinger 2011).
The above results describe how Nmass varied with � and �for a single leaf. In leaf stacks, the correlation coeYcientsfor the relationships between Nmass and foliar optical prop-erties decreased as the number of leaves increased. Despitethis, the negative slopes of the relationship between Nmass
and �FULL remained intact for all leaf stacks (Fig. 3;P < 0.001 in all cases).
Relationships between leaf traits and leaf optical properties
Across all species and treatments, there was a weak, positivecorrelation between LMA and �FULL (P < 0.05), and weak,negative correlation between LMA and both �FULL and �nir
(Fig. 4; P < 0.001 in both cases). There were stronger nega-tive correlations between LMA and �midir (r2 = 0.38,P < 0.001) and �midir (r
2 = 0.40, P < 0.001), although this islikely due to the inXuence of EWT. LMA was positively cor-
Fig. 2 Regression between foli-ar Nmass and a solar-weighted full spectrum reXectance (�FULL; 400–2,500 nm), b solar-weight-ed NIR reXectance (�nir; 700–1,300 nm), and c solar-weighted visible reXectance (�vis; 400–700 nm; data log-transformed), d solar-weighted full spectrum transmittance (�FULL), e solar-weighted NIR transmittance (�nir), and f solar-weighted visi-ble transmittance (�vis; data log-transformed). The nonlinear relationship between Nmass and both �vis and �vis appears to be due to diVerences in slope between species that vary in their N concentrations. All correlations are signiWcant at P < 0.001. Black oak (BO), red maple (RM) and sweetgum (SG) were all included in the regression analysis
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
0.25
0.30
0.35
0.40
0.45
Ref
lect
ance
FU
LL
SGRMBO
r2 = 0.26a
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
0.25
0.30
0.35
0.40
0.45
Tra
nsm
ittan
ceF
ULL
d ns
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.50.35
0.40
0.45
0.50
0.55
0.60
0.65R
efle
ctan
ceni
r
r2 = 0.17
b
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.50.35
0.40
0.45
0.50
0.55
0.60
0.65
Tra
nsm
ittan
ceni
r
r2 = 0.10e
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
0.05
0.10
0.15
0.20
Leaf Nmass (%)
Ref
lect
ance
vis
r2 = 0.60c
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
0.05
0.10
0.15
0.20
Leaf Nmass (%)
Tra
nsm
ittan
cevi
s
r2 = 0.50f
123
Oecologia
Author's personal copy
related with �vis (r2 = 0.20, P < 0.001), whereas there was no
relationship between �vis and LMA (P > 0.1).Equivalent water thickness was negatively correlated
with �midir and �midir and was the best overall predictor ofMid IR foliar optical properties (Fig. 5; P < 0.001 in allcases). EWT was weakly, but positively, correlated with�FULL (r2 = 0.06, P < 0.01), �nir (r2 = 0.04, P < 0.01), and�vis (r2 = 0.28, P < 0.001), and negatively correlated with�FULL (r2 = 0.10, P < 0.001) and �nir (r
2 = 0.09, P < 0.001).Relative water content (g H2O g¡1 fresh leaf tissue) was notcorrelated with any foliar optical parameter we considered(P > 0.1 in all cases).
Multiple regression models using all combinations ofleaf traits did not improve predictions for �FULL, �FULL, �nir,or �nir above that which was obtained by Nmass alone.
Relationships between leaf traits
The relationship between LMA and EWT was signiWcantand positive both within (black oak: r2 = 0.73; red maple:r2 = 0.36; sweetgum: r2 = 0.46; P < 0.001 in all cases) andacross (r2 = 0.54, P < 0.001) all species. LMA and Nmass
were negatively correlated within sweetgums (r2 = 0.22,
Fig. 3 Foliar Nmass versus solar weighted full spectrum reXectance(�FULL) for stacks of one (r2 = 0.25, P < 0.001), two (r2 = 0.18,P < 0.001), four (r2 = 0.15, P < 0.001), and eight (r2 = 0.10, P < 0.001)leaves. Black oak (BO), red maple (RM) and sweetgum (SG) were allincluded in the regression analysis. Although the y-intercept diVers, allstacks show similar negative relationships between foliar N and �FULL
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
Leaf Nmass (%)
Ref
lect
ance
FU
LLsingle leaftwo leavesfour leaveseight leaves
Fig. 4 Regression between LMA and a solar-weighted full spectrum reXectance (�FULL; P < 0.05), b solar-weighted NIR reXectance (�nir; P > 0.1), c solar-weighted full spectrum transmittance (�FULL; P < 0.001), and d solar-weight-ed NIR transmittance (�nir; P < 0.001). Black oak (BO), red maple (RM) and sweetgum (SG) were all included in the regres-sion analysis
20 40 60 80 120
0.25
0.30
0.35
0.40
0.45
0.50
Ref
lect
ance
FU
LL
SGRMBO
r2 = 0.02a
20 40 60 80 1200.2
0.3
0.4
0.5
0.6
Tra
nsm
ittan
ceF
ULL
c r2 = 0.08
20 40 60 80 1200.25
0.30
0.35
0.40
0.45
0.50
LMA (g m−2)
Ref
lect
ance
nir
b ns
20 40 60 80 1200.2
0.3
0.4
0.5
0.6
LMA (g m−2)
Tra
nsm
ittan
ceni
r
d r2 = 0.06
123
Oecologia
Author's personal copy
P < 0.001), but the relationship was not signiWcant in blackoaks, red maples, or across all species. EWT and Nmass werenot signiWcantly correlated within any species (P > 0.1 inall cases), but with all species combined they were slightlynegatively related (r2 = 0.05, P < 0.01).
Canopy-level reXectance
Despite the absence of diVerences in leaf-level reXectancebetween treatments, reXectance data obtained from theAVIRIS sensor showed that whole-canopy NIR reXectancewas higher for the N-fertilized sweetgum treatment than theambient CO2 sweetgum treatment (Fig. 6; P < 0.001). Can-opy NIR reXectance from the elevated CO2 sweetgum treat-ment was lower than the ambient CO2 sweetgum treatment,although the trend was not signiWcant (Fig. 6; P > 0.1)and the diVerence was small compared to that between theN-fertilized and ambient CO2 treatments. There was little to
no diVerence in canopy reXectance in the visible or MIRportions of the spectrum.
Discussion
A primary goal of this study was to examine relationshipsbetween leaf traits and leaf optical properties that mighthelp explain the positive correlation between canopy Nmass
and full spectrum canopy albedo seen in temperate andboreal forests (Ollinger et al. 2008; Hollinger et al. 2010).Our results do not support the hypothesis that the canopy-level trend was caused by a similar trend occurring at theleaf level. There were no diVerences in leaf �FULL or �FULL
between comparable N or CO2 treatments, despite changesin leaf Nmass caused by N fertilization and increases inwhole canopy NIR reXectance, as measured using AVIRIS,in the N-fertilized plots at ORNL. Earlier work showed areduction in leaf Nmass in CO2-treated plots relative to con-trols at ORNL (Norby and Iversen 2006). Although we alsosaw lower Nmass in CO2-treated plots, the diVerence wassmaller and not signiWcant. As a result, our prediction thatfoliar � would decrease with increased CO2 fertilizationwas neither supported nor refuted.
Whereas the slope of the previously observed relation-ship between Nmass and canopy full spectrum albedo is pos-itive, we observed a negative relationship at the leaf level,when all species and treatments were combined. This pat-tern was driven in part by the visible region, where high Nfoliage is expected to absorb more light due to increasedpigment concentration. Nevertheless, the relationship wasalso negative, albeit weaker, even when restricted to theNIR region.
To what can we attribute these results? Previous studieshave linked scattering and NIR reXectance to leaf structuralparameters such as leaf thickness, percent intercellular airspace (%IAS), or the Ames/A ratio (Knapp and Carter 1998;Gausman et al. 1970; Slaton et al. 2001). In our own data,
Fig. 5 Both a solar-weighted reXectance (�) and b solar-weightedtransmittance (�) declined in the Mid IR region (700–1,350 nm) as leafEWT increased. This was due to the strong positive correlationbetween Mid IR absorption [1 ¡ (� + �)] and EWT (c), which demon-
strates the importance of water abosrption in the Mid IR in inXuencingscattering from this region. All correlations are signiWcant atP < 0.001. Black oak (BO), red maple (RM) and sweetgum (SG) wereall included in the regression analysis
6 8 12 16
0.18
0.20
0.22
0.24
0.26
0.28
EWT (mg cm−2)
Ref
lect
ance
mid
ir
SGRMBO
r2 = 0.35a
6 8 12 16
0.20
0.25
0.30
0.35
0.40
0.45
EWT (mg cm−2)
Tra
nsm
ittan
cem
idir r2 = 0.28
b
6 8 12 160.30
0.35
0.40
0.45
0.50
0.55
0.60
EWT (mg cm−2)
Abs
orpt
ion m
idir
r2 = 0.65
c
Fig. 6 AVIRIS canopy reXectance across the full spectrum for N-fer-tilized, ambient CO2, and elevated CO2 sweetgum treatments at ORNL
500 1000 1500 2000 2500
0.0
0.1
0.2
0.3
0.4
0.5
Wavelength
Can
opy
refle
ctan
ce
N fertilizedAmbient CO2
Elevated CO2
123
Oecologia
Author's personal copy
there was no correlation between �nir and LMA, althoughthis could be due to the confounding eVects on reXectanceof leaf thickness and leaf density, two variables that deter-mine LMA, but have opposing eVects on other leaf traitssuch as the fraction of leaf volume composed of intercellu-lar air space (increased leaf thickness increases %IAS,whereas increased leaf density decreases %IAS; Niinemets1999, 2001). A number of studies have observed positiverelationships between NIR reXectance and %IAS (Castro-Esau et al. 2006; Gausman et al. 1970; Slaton et al. 2001),while others suggest varying relationships between %IASand foliar Nmass. In a study involving fertilization of tall fes-cue (Festuca arundinacea Schreb.), Rademacher and Nel-son (2001) found that the ratio of IAS to total mesophyllspace was lower in high N foliage because N fertilizationenhanced mesophyll area to a greater degree than leaf vol-ume. Conversely, Niinemets (1999), across a broad rangeof woody vegetation, observed a negative relationshipbetween leaf density and both Nmass and the fraction of leafmesophyll as intercellular airspace. This implies that leaveswith high foliar N should be less dense and have a greater%IAS.
In contrast to our leaf level results, remote sensing datacollected by the AVIRIS sensor showed an increase inwhole-canopy NIR reXectance with N fertilization in theORNL sweetgum plantation, a response that is consistentwith observations along natural gradients in plant N status(Ollinger et al. 2008; Hollinger et al. 2010). Collectively,these results suggest that the association between Nmass andcanopy NIR reXectance is driven by stem or canopy levelproperties that are either inXuenced by, or covary with, leafNmass.
In our study, the only canopy structural variable wewere able to evaluate was leaf area index (LAI, leaf areaper unit area of ground), which was higher in the ORNLN-fertilized plots than in the control plots (LAI » 5.5 and4.5, respectively; Richard Norby and Colleen Iversen, per-sonal communication). To evaluate the potential eVect ofthis diVerence on the canopy reXectance values observedby AVIRIS, we used SAIL-2 (Scattering from ArbitrarilyInclined Leaves; Verhoef 1984; Braswell et al. 1996), aradiative transfer model that has been used extensively inthe literature (e.g., Huemmrich and Goward 1997; Andrieuet al. 1997; Daughtry et al. 2000; Zhang et al. 2006).Although SAIL is known to be sensitive to changes inLAI, especially in low-LAI systems (e.g., Asner 1998),model runs for the control and N-fertilized plots, con-ducted using measured LAI and measured leaf and back-ground reXectance and transmittance spectra, produced anegligible diVerence in canopy reXectance (Fig. 7). Thissuggests that canopy properties other than LAI are drivingthe observed changes in canopy reXectance between thestands.
Other canopy properties that may provide a link betweenplant N status and canopy reXectance include shoot archi-tecture (Smolander and Stenberg 2003; Malenovský et al.2008), crown geometry (e.g., Rautiainen et al. 2004)- andleaf angle distribution (Asner 1998; Close and Beadle2006). Each of these are inXuenced by a variety of resourceoptimization mechanisms that are known to vary with plantnutrient status, perhaps resulting in associations betweenleaf traits and canopy structural properties that have yet tobe explored (e.g., Ollinger 2011). As an example, N fertil-ization experiments involving eucalyptus (Close andBeadle 2006), wheat (Brooks et al. 2000)- and rice (Tariet al. 2009) have all demonstrated reductions in mean leafangle (i.e.- leaves become more horizontal) with improvedN nutrition. Because low leaf angle is associated withincreased reXectance, particularly in the NIR region (Asner1998; Ollinger 2011), such a response could contribute to apositive relationship between Nmass and canopy albedo if itwere to occur over broad gradients within native ecosys-tems. Canopy leaf angle is a challenging parameter to mea-sure, and at present, we lack the data needed to evaluate theextent to which variation in leaf angle, or other relatedparameters, may have contributed to our results.
Conclusions
This study investigated the importance of leaf-level reXec-tance properties as a potential cause of the relationshipbetween foliar Nmass and canopy full spectrum albedo.Although we cannot completely rule out the importance ofmechanisms involving leaf-level radiation scattering, our
Fig. 7 Results of SAIL-2 model runs for the ambient CO2 and N-fer-tilized plots, conducted using measured LAI and measured leaf andbackground reXectance and transmittance spectra. Resulting spectrashow a negligible diVerence in canopy reXectance
500 1000 1500 2000
0.0
0.1
0.2
0.3
0.4
0.5
Wavelength
SA
IL c
anop
y re
flect
ance
Ambient CO2 LAIN Fertilized LAI
123
Oecologia
Author's personal copy
results suggest that they are not the dominant inXuence onthis relationship. Instead, we suggest that the principalmechanism lies in associations between leaf-level Nmass andoptically important canopy structural properties that haveyet to be thoroughly explored. Although measuring canopystructural properties is challenging, future work should con-sider variables such as leaf arrangement and foliage clump-ing, leaf angle distribution, and crown geometry, inconjunction with leaf traits and canopy spectral properties.If met with success, such investigations would improve ouroverall understanding of how biogeochemical processesinXuence biophysical properties that are relevant to surfaceenergy exchange and interactions between ecosystems andclimate.
Acknowledgments We thank G. James Collatz for helpful com-ments on a draft of this manuscript, Rob Braswell for providing theSAIL-2 model code, and Richard Norby, Colleen Iversen, and JeVeryWarren for support at ORNL. We are indebted to Michael Eastwood,ER-2 pilots Denis Steel, Tim Williams, and the rest of the AVIRISteam for aircraft data acquisition. This work was funded by a grantfrom the North American Carbon Program (NACP) NASA’s Terres-trial Ecology and Carbon Cycle Science Programs and a graduate fel-lowship provided by the Research and Discover program. The ORNLFACE experiment was supported by the US Department of Energy,OYce of Science, Biological and Environmental Research Program.A.D.R. and M.K.B. acknowledge support, through the Harvard ForestREU program, from the National Science Foundation (Grant DBI-04-52254).
References
Andrieu B, Baret F, Jacquemoud S, Malthus T, Steven M (1997) Eval-uation of an improved version of SAIL model for simulating bidi-rectional reXectance of sugar beet canopies. Remote Sens Environ60:247–257
Asner GP (1998) Biophysical and biochemical sources of variability incanopy reXectance. Remote Sens Environ 64:234–253
Braswell BH, Schimel DS, Privette JL, Moore B III, Emery WJ, Sulz-man EW, Hudak AT (1996) Extracting ecological and biophysi-cal information from AVHRR optical data: an integratedalgorithm based on inverse modeling. J Geophys Res 101:23335–23348
Brooks TJ, Wall GW, Pinter PJ Jr, Kimball BA, LaMorte RL, LeavittSW, Matthias AD, Adamsen FJ, Hunsaker DJ, Webber AN(2000) Acclimation response of spring wheat in a free-air CO2enrichment (FACE) atmosphere with variable soil nitrogenregimes. 3. Canopy architecture and gas exchange. PhotosynthRes 66:97–108
Castro-Esau KL, Sanchez-Azofeifa GA, Rivard B, Wright AJ, Ques-ada M (2006) Variability in leaf optical properties of Mesoamer-ican trees and the potential for species classiWcation. Am J Bot94:517–530
Chen JM, Cihlar J (1995) Quantifying the eVect of canopy architectureon optical measurements of leaf area index using two gap size anal-ysis methods. IEEE Transact Geosci Remote Sens 33:777–787
Close DC, Beadle CL (2006) Leaf angle responds to nitrogen supply ineucalypt seedlings. Is it a photoprotective mechanism? Tree Phys-iol 26:743–748
Daughtry CST, Walthall CL, Kim MS, Brown de Colstoun E, McMurt-rey JE III (2000) Estimating corn leaf chlorophyll concentrationfrom leaf and canopy reXectance. Remote Sens Environ 74:229–239
Ellsworth DS, Reich PB, Naumburg ES, Koch GW, Kubiske ME,Smith SD (2004) Photosythesis, carboxilation and leaf nitrogenresponses of 16 species to elevated pCO2 across four free-air CO2enrichment experiments in forest, grassland and desert. GlobChange Biol 10:2121–2138
Evans JR (1989) Photosynthesis and nitrogen relationships in leaves ofC3 plants. Oecologia 78:9–19
Field C, Mooney HA (1986) The photosynthesis-nitrogen relationshipin wild plants. In: Givinsh TI (ed) On the economy of form andfunction. Cambridge University Press, Cambridge, pp 25–55
Gates DM, Keegan HJ, Schleter JC, Weidner VR (1965) Spectral prop-erties of plants. Appl Opt 4:11–20
Gausman HW, Allen WA, Cardenas R, Richardson AJ (1970) Relationof light reXectance to historical and physical evaluations of cottonleaf maturity. Appl Opt 9:545–552
Gausman HW, Allen WA, Escobar DE (1974) Refractive index ofplant cell walls. Appl Opt 13:109–111
Gueymard C (2004) The sun’s total and spectral irradiance for solar en-ergy applications and solar radiation models. Sol Energy 76:423–453
Hollinger DY, Ollinger SV, Richardson AD, Meyers TP, Dail DB,Martin ME, Scott NA, Arkebauer TJ, Baldocchi DD, Clark KL,Curtis PS, Davis KJ, Desai AR, Dragoni D, Goulden ML, Gu L,Katul GG, Pallardy SG, Paw UKT, Schmid HP, Stoy PC, SuykerAE, Verma SB (2010) Albedo estimates for land surface modelsand support for a new paradigm based on foliage nitrogen concen-tration. Glob Change Biol 16:696–710
Huemmrich KF, Goward SN (1997) Vegetation canopy PAR absorp-tance and NDVI: an assessment for ten tree species with the SAILmodel. Remote Sens Environ 61:254–269
Iversen CM, Norby RJ (2008) Nitrogen limitation in a sweetgum plan-tation: implications for carbon allocation and storage. Can J ForRes 38:1021–1032
Jacquemoud S, Verhoef W, Baret F, Bacour C, Zarco-Tejada PJ, AsnerGP, François C, Ustin SL (2009) PROSPECT + SAIL models: areview of use for vegetation characterization. Remote Sens Envi-ron 113:S56–S66
Johnson DW, Cheng W, Joslin JD, Norby RJ, Edwards NT, Todd DEJr (2004) EVects of elevated CO2 on nutrient cycling in a sweet-gum plantation. Biogeochemistry 69:379–403
Knapp AK, Carter GA (1998) Variability in leaf optical propertiesamong 26 species from a broad range of habitats. Am J Bot85:940–946
Longstreth DJ, Bolaños JA, Goddard RH (1985) Photosynthetic rateand mesophyll surface area in expanding leaves of Alternantheraphiloxeroides grown at two light levels. Am J Bot 72:14–19
Magill AH, Aber JD, Currie WS, NadelhoVer KJ, Martin ME, McDo-well WH, Melillo JM, Steudler P (2004) Ecosystem response to15 years of chronic nitrogen additions at the Havard ForestLTER, Massachusetts, USA. For Ecol Manag 196:7–28
Malenovský Z, Martin E, Homolová L, Gastellu-Etchegorry J-P, Zuri-ta-Milla R, Schaepman ME, Pokorný R, Clevers JGPW, Cudlín P(2008) InXuence of woody elements of a Norway spruce canopyon nadir reXectance simulated by the DART model at very highspatial resolution. Remote Sens Environ 112:1–18
Niinemets U (1999) Research review: components of leaf dry mass perarea-thickness and density-alter leaf photosynthetic capacity inreverse directions in woody plants. New Phytol 144:35–47
Niinemets U (2001) Global-scale climatic controls of leaf dry mass perarea, density, and thickness in trees and shrubs. Ecology 82:453–469
123
Oecologia
Author's personal copy
Nobel PS, Zaragoza LJ, Smith WK (1975) Relation between meso-phyll surface area, photosynthetic rate, and illumination level dur-ing development for leaves of Plectranthus parviXorus Henckel.Plant Physiol 55:1067–1070
Norby RJ, Iversen CM (2006) Nitrogen uptake, ditribution, turnover,and eYciency of use in a CO2-enriched sweetgum forest. Ecology87:5–14
Norby RJ, Todd DE, Fults J, Johnson DW (2001) Allometric determi-nation of tree growth in a CO2-enriched sweetgum stand. NewPhytol 150:477–487
Norby RJ, Sholtis JD, Gunderson CA, Jawdy SS (2003) Leaf dynamicsof a deciduous forest canopy: no response to elevated CO2. Oec-ologia 136:574–584
Ollinger SV (2011) Sources of variability in canopy reXectance and theconvergent properties of plants. New Phytol 189:375–394
Ollinger SV, Aber JD, Lovett GM, Millham SE, Lathrop RG, Ellis JE(1993) A spatial model of atmospheric deposition for the North-eastern US. Ecol Appl 3:459–472
Ollinger SV, Richardson AD, Martin ME, Hollinger DY, Frolking S,Reich PB, Plourde LC, Katul GG, Munger JW, Oren R, SmithML, Paw UKT, Bolstad PV, Cook BD, Day MC, Martin TA,Monson RK, Schmid HP (2008) Canopy nitrogen, carbon assim-ilation and albedo in temperate and boreal forests: functional rela-tions and potential climate feedbacks. Proc Nat Acad Sci105:19335–19340
Oren R, Ellsworth DS, Johnsen KH, Phillips N, Ewers BE, Maier C,Schäfer KVR, McCarthy H, Hendrey G, McNulty SG, Katul GG(2001) Soil fertility limits carbon sequestration by forest ecosys-tems in a CO2-enriched atmosphere. Nature 411:469–472
Rademacher IF, Nelson CJ (2001) Nitrogen eVects on leaf anatomywithin the intercalary meristems of tall fescue leaf blades. AnnBot 88:893–903
Rautiainen M, Stenberg P, Nilson T, Kuusk A (2004) The eVect ofcrown shape on the reXectance of coniferous stands. Remote SensEnviron 89:41–52
Reich PB, Walters MB, Ellsworth DS (1997) From tropics to tundra:global convergence in plant functioning. Proc Nat Acad Sci94:13730–13734
Reich PB, Ellsworth DS, Walters MB, Vose JM, Gresham C, Volin JC,Bowman WD (1999) Generality of leaf train relationships: a testacross six biomes. Ecology 80:1955–1969
Riggs JS, Tharp ML, Norby RJ (2009) ORNL FACE CO2 data. carbondioxide information analysis center, Oak Ridge, TN, USA. http://cdiac.ornl.gov. Accessed 15-July-2010
Sánchez J, Canton MP (1999) Space imaging processing. CRC PressLLC, Boca Raton
Slaton MR, Hunt ER, Smith WK (2001) Estimating near-infrared leafreXectance from leaf structural characteristics. Am J Bot 88:278–284
Smolander S, Stenberg P (2003) A method to account for shoot scaleclumping in coniferous canopy reXectance models. Remote SensEnviron 88:363–373
Tari DB, Gazanchian A, Pirdashti HA, Nasiri M (2009) Flag leaf mor-phophysiological response to diVerent agronomical treatments ina promising line of rice (Oryza sativa L.). Am-Euras J Agric Envi-ron Sci 5:403–408
Verhoef W (1984) Light scattering by leaf layers with application tocanopy reXectance modeling: the SAIL model. Remote SensEnviron 16:125–141
Woolley JT (1971) ReXectance and transmittance of light by leaves.Plant Physiol 47:656–662
Wright IJ, Reich PB, Westoby M, Ackerly DD, Baruch Z, Bongers F,Cavender-Bares J, Chapin T, Cornelissen JHC, Diemer M, FlexasJ, Garnier E, Groom PK, Gulias J, Hikosaka K, Lamont BB, LeeT, Lee W, Lusk C, Midgley JJ, Navas M-L, Niinemets U, OleksynJ, Osada N, Poorter H, Poot P, Prior L, Pyankov VI, Roumet C,Thomas SC, Tjoelker MG, Veneklaas EJ, Villar R (2004) Theworldwide leaf economics spectrum. Nature 428:821–827
Zhang Q, Xiao X, Braswell BH, Linder E, Ollinger S, Smith ML, Jen-kins JP, Baret F, Richardson AD, Moore B III, Minocha R (2006)Characterization of seasonal variation of forest canopy in atemperate deciduous broadleaf forest, using daily MODIS data.Remote Sens Environ 105:189–203
123
