sensitivity of leaf physiognomy to climate: applications to habitat … et al._biological... ·...

INTRODUCTION

Plants have evolved to exist in conditions that are ra -re ly ideal for normal maintenance and may be at thesurvival limit. In response, plants can adapt to avoidand overcome stress by using various rapid or slow re-sponding mechanisms, such as leaf movement andphe notypic plasticity (Kato et al., 2003; Shepherd &

Griffiths, 2006; Xu et al., 2009a). As the fundamentalenergy unit of plants, leaves are considered as a nexusbetween plants and environments, which can have im-portant ecological implications for species survival,growth, and distribution. Recently, general scaling re-lationships between leaf traits and climate have beenthe subject of interest. A worldwide ‘‘economic’’ spec-trum of correlated leaf traits (that can provide a linkbet ween the various environmental factors and leaf

Journal of Biological Research-Thessaloniki 20: 312 – 325, 2013J. Biol. Res.-Thessalon. is available online at http://www.jbr.grIndexed in: WoS (Web of Science, ISI Thomson), SCOPUS, CAS (Chemical Abstracts Service) and DOAJ (Directory of Open Access Journals)

312

Sensitivity of leaf physiognomy to climate: applications

to habitat-scaled and species-based climate proxy

Fei XU1,3, Renqing WANG2,3 and Weihua GUO2,3*

1 College of Life Sciences, Shandong Normal University, 250014 Jinan, P. R. China 2 Institute of Ecology and Biodiversity, College of Life Sciences,

Shandong University, 250100 Jinan, P. R. China 3 Shandong Provincial Engineering and Technology Research Center for Vegetation Ecology,

Shandong University, 250100 Jinan, P. R. China

Received: 14 February 2013 Accepted after revision: 6 December 2013

Leaf physiognomy is climate-sensitive and used for quantitative climate reconstruction by apply-ing leaf-climate correlations. Most studies have focused on site-level means of species sets in lar -ge-scale environments. The sensitivity of leaf physiognomy to habitat microclimate within speciesis poorly known, which limits our understanding of leaf-climate relationships and applications tocli mate proxies for forest monitoring. An experiment was performed in the present study to inve -sti gate the responses of leaf size, shape, and venation pattern in the seedlings of Quercus acutissi -ma to different gradients of water and light availability. Multiple linear regressions and a contourex traction method were developed and their ability to predict microclimate was assessed by usingva riables derived from leaf physiognomy. The trends of leaf morphological variations along thegra dients share a general resource acquisition and conservation enhancement pattern. The syner -gy of leaf size, shape, and venation pattern optimized the tradeoff relationship between investmentand return of restricted resources. The water-induced plasticity of leaves was lower compared tolight-induced plasticity, which resulted in the predictive methods’ poor ability to estimate watera vailability compared to their ability to estimate light availability. The contour extraction methodwas more precise, especially for combined predictions of extreme environments because multipleli near regressions exhibited overestimation and underestimation at lower and higher gradients,re spectively. The present study demonstrated that intraspecific variations of leaf physiognomycan provide a functional link to habitat, and climate proxies based on these relationships maycon tribute useful information towards forest management.

Key words: climate proxies, contour extraction method, leaf physiognomy, multiple linear regres -sion, water and light availability.

* Corresponding author: e-mail: guo_ [email protected]

This open-access article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use,

printing, distributing, transmitting and reproduction in any medium, provided the original author and source are appropriately cited. Full

text and supplementary material (if any) is available on www.jbr.gr

fun ctions) has been identified (Niinemets, 2001; We -s to by et al., 2002; Westoby & Wright, 2003; Wright etal., 2004, 2007; Shipley et al., 2006a) and widely usedfrom functional individuals to communities and e co -sy stems (Garnier et al., 2004; Shipley et al., 2006b;We stoby & Wright, 2006). In addition, this spectrumof traits has become the proxy for reconstructing pa-leoclimates or predicting future climates (Royer et al.,2005; Whitfield, 2006).

Leaf morphological characteristics are a useful vi-sual guide for constructing relationships between dif-ferent plants as well as between plants and their envi -ron ments (Navas & Garnier, 2002; Roche et al., 2004).Leaf physiognomy can serve as an excellent tool forbotanical and ecological studies (Traiser et al., 2005).Plant species with widely varying leaf shapes, sizesand venation co-occur in vegetation. The significanceof leaf variations for species niche differentiation isstill not entirely understood, whereas the differencesin leaf physiognomy have been the subject of exten-sive research (Niinemets et al., 2007b). The morpho -lo gical traits of leaves are often used in taxonomy.However, attention needs to be given to the problemof the reliability of leaf morphological characteristics,which are dependent on environmental conditions(Viscosi et al., 2009). The development of digital ima -ge processing and analysis technology has improvedthe ability to recognize and conduct geometric mea -su rements of leaf morphology in the field (Du et al.,2007), which has led to the increased number of stud-ies on the sensitivity of leaf physiognomy to climate.

Environmental sensitivity of leaf physiognomy hasled to an upsurge in developing the technique for cli-mate proxies. The most common leaf physiognomicmethods are leaf-margin analysis and leaf-area analy-sis, both of which are based on a single variable, na me -ly, the percentage of untoothed species at a site andsite-mean leaf size, respectively (Peppe et al., 2011).The Climate-Leaf Analysis Multivariate Program,which uses additional categorical leaf states, was de ve -loped to obtain more accurate climate estimates com -pared to the results of univariate approaches (Wolfe,1995). However, errors and biases were also found inproblems related to character definitions, states, andanalysis methods in the predictive framework. An a -me liorative digital leaf physiognomy, which uses con-tinuous variables to replace the discrete ones, is con-siderably more accurate because it uses stricter chara -cter definitions (Huff et al., 2003). Multiple linear re-gression models are the preferred multivariate analy-sis methods for simple application (Royer et al., 2005;

Peppe et al., 2011). Other complex analyses based oncom puter algorithms have also been applied (Me zia -ne & Shipley, 2001; Blonder et al., 2011). These pre-vious studies were usually based on the averages ofse veral species, i.e. site-based analysis (Greenwood,2005; Royer et al., 2009). In such case, the climate waspredicted at large scales, while predictions at local sca -le have not been conducted yet. Therefore, spe cies-based analysis of leaf physiognomy may reveal newinsights concerning its relation with microclimate. Acontour method, which is usually used in geography,can display the continuous variations of multi-para-meters and account for small changes of leaf chara -cte ristics according to microclimate. In the presentstu dy, we originally developed a contour extractionme thod to construct the visual relationships betweenleaf physiognomy and microclimate.

The oak species of genus Quercus L. generally ex -hi bit large plasticity in leaf morphology and have beencommonly used for the analysis of leaf-climate rela-tionships (Sisó et al., 2001; Quero et al., 2006; Royeret al., 2008; Viscosi et al., 2009; Zhu et al., 2012).Quer cus acutissima Carr. is one of the most widespreadoak species in north China and is the dominant decid-uous broadleaved species in the study area. The growthof Q. acutissima can be mainly affected by restrictionscaused by habitat conditions. Increasingly differentenvironmental conditions result in an increasingly dif-ferent phenotypic plasticity of the who le plant and leaftraits (Xu et al., 2008). In addition, many Q. acutissi-ma forests exhibit patch distribution because of an-thropogenic disturbance. Other strong competitors,such as Robinia pseudoacacia L., are gra du ally invad-ing and changing native habitats, which lead to thede cline in seedling recruitment and decrease in growthrates through aggressive capture of light and water re-sources (Xu et al., 2010). The adaptive mechanismsmay be investigated by analyzing the leaf morphologyof Q. acutissima and the function of environmentalin dication in response to diverse water and light con-ditions, which are conducive for guiding vegetationmaintenance and restoration.

We designed an experiment with controlled waterand light conditions to investigate responses of leafsize, shape, and venation pattern in the seedlings ofQ. acutissima to different gradients of water and lightavailability. In previous studies, we have found thatva riations of leaf physiognomy in the nature can be si -mu lated artificially and there are interspecific diffe ren - ces between coexisted species (Xu et al. 2008, 2009b).

Fei Xu et al. — Sensitivity of leaf physiognomy to climate 313

On this basis, the objectives of this study were to (1)investigate the role of intraspecific variation in leaf-cli mate relationships, (2) develop microclimate predi -ction models by using variables derived from leaf phy -sio gnomy, and (3) assess the potential application ofmodels as microclimate proxies.

MATERIALS AND METHODS

Study site

The study was conducted at the Fanggan ResearchStation of Shandong University, Shandong Province,China (36°26΄N, 117°27΄E). The site is characterizedby a warm temperate monsoon climate, with a meanannual temperature of 13±1°C and a mean annualprecipitation of 700 ± 25 mm, which occurs mostlyduring the summer. The soil of the area is a cinna-mon-type, and the parent material is limestone (Xu etal., 2008). Sawtooth oak (Q. acutissima) and black lo-cust (R. pseudoacacia) coexist and form mixed forestsin this area. The canopy has a dominant layer whichreaches 14 m, and the lower limits of the crown are atabout 9 m. The leaf area index can reach ~5.12 whenthe trees are flourishing in August (Xu et al., 2009b).

Plant materials

One-year-old seedlings of Q. acutissima were used asex perimental materials in this study. Acorns of Q. a -cu tissima were collected from a hill near the researchstation in early spring and planted in plastic pots (32× 29 cm in height × diameter, one acorn in each pot).The soil was a 64:22:14 (v/v/v) mixture of humic soil,sand, and loam, with a saturated water content of36% by mass, the largest volumetric water content of28%, and 68% porosity. The pH was 4.4, and the ma-jor chemical components included 88.4 g organic mat -ter, 3.7 g total nitrogen and 42.3 mg available phos -pho rus per kilogram. All the pots were regularly irri-gated and subjected to weed control before the begin-ning of the experiment.

Experimental design

The seedlings were submitted to controlled experi-ments from July to September. A factorial experimentof two factors (water and light) of four and three lev-els, respectively, was designed. Water was withheldfrom the drought groups until the soil moisture reac -hed ~50% (W2), 30% (W3), and 10% (W4) of fieldca pacity, whereas the well-watered groups receiveddaily irrigation to maintain soil water content between

70% and 80% (W1) of field capacity. All the pots weremoved into a rain-out shelter to avoid precipitationdi sturbance. The soil water content was controlled bygravimetric probe and the pots were weighed daily tomaintain the four different water contents. Top irri-gation evenly supplemented the water lost via transpi-ration and evaporation.

The light-control treatment was conducted in sha -de shelters covered by plastic films or woven black ny-lon nets. The frame of the shelter was 5.0×2.5×3.0 m(length×width×height). The microclimate was moni -tored daily by using a micro-quantum sensor and atemperature sensor of Mini-PAM (Walz GmbH, Ef -fel trich, Germany). The average Photosynthetic Ac-tive Radiation (PAR) measured from 07:00 to 16:00was 544 ± 71, 361 ± 17, and 56 ± 6.7 μmol m–2 s–1 inthe open field (L1), under plastic films (L2), and ny-lon nets (L3), respectively. The light transmission ra-tio was ~66% and 10% under the stress condition com -pared with the control. Analysis of variance (ANO-VA) found no significant difference (p=0.534) in airtemperature between the three light gradients (32.8±0.46, 32.4±0.40 and 32.1±0.39°C).

The Relative Water Content (RWC) and RelativeLight Intensity (RLI) were used as the standards toquantify the gradients of water and light availability.RWC was calculated by averaging the diurnal soil wa-ter content of the sampled seedlings. The average lightintensity of the sampled seedlings was obtained bymea suring PAR in four directions where the seed lingswere located. RLI was then calculated by dividing theaverage light intensity with the maximal value. Tenpots were randomly assigned to each water and lightavailability treatment. The gradients of W2 and W4were eliminated from the high (L1) and low (L3) gra-dients of light availability considering the heavy mana -gement and measurement workload. After a two-month treatment period, 15 mature leaves from threeseedlings per treatment were taken for morphologicalmeasurements.

Morphological measurements

The leaf area was measured by using a CI-203 laserarea meter (CID Inc., Washington, USA). The linearmeasurement was taken by using a digital caliper.Leaf dry mass was measured after oven drying at80°C for 48 hrs. The detailed information of the mor-phological parameter measurements and definitionsare shown in Figure 1 and Table 1.

314 Fei Xu et al. — Sensitivity of leaf physiognomy to climate

Statistical analysis

Two-way ANOVA with type III sums of squares wasused to test the interactive effects of environmentalfactors on leaf physiognomy. Pearson’s correlationsbetween leaf morphological traits and climate para -me ters were calculated, and Ordinary Least Square(OLS) regression lines were fitted to predict the habi-tat information by using variables that were consid-ered to represent the primary responses. The multi-collinearity influence could be ignored without usingelimination methods (e.g., stepwise or ridge regression)because focus was given only to the predictive powerof the functions (Mela & Kopalle, 2002; Dormann etal., 2013). The jackknife-type approach was used toevaluate the accuracy of the regression functions tobuild a group of virtual data as the test set, and theab solute residuals were used as the predictive criterion.The level of response to the variation of each factor(water and light) was estimated by using the PIRWCand PIRLI indices, respectively, which ranged from 0to 1. The index of plasticity (PI) was calculated as thedifference between the maximum and the minimummean values divided by the maximum mean value(Val ladares et al., 2000b). The changes of leaf mor -pho logical variables in the different gradients of wa-ter and light availability were illustrated in three-di-mensional surface and contour plots. The smoothingcubic spline was employed to trace the variability ofthe leaf variables (Schimek, 2000), and contour ex-traction was carried out by adjusting the step and mi-

nor number for values of the Z-axis (data of leaf mor-phology). The intersections of contours from differ-ent leaf variables, which are considered to provide in-formation about the habitat, were easily fixed by graph-merged and data-drawn application. All statisticalanalyses were performed by using the SPSS 13.0 soft-ware package (SPSS Inc., Chicago, USA). Plots weredrawn by using the Statistica 6.0 software (StatSoftInc., Oklahoma, USA).

RESULTS

LA significantly increased at increasing water andlight availability (Fig. 2A, Table 2). The range of thechanges was broader with sufficient resources thanwith seedlings under serious stress. Similar to LA, va -ri ables LDM, LL, and LW were positively correlatedwith water and light availability, respectively (Fig. 2B,D, E). SLA showed a positive correlation with wa teravailability, but was negatively correlated with light a -vai lability. The variations were nearly linear in the sur -face plot (Fig. 2C). Although LPL was significantlycorrelated with water and light availability, the trendwas not consistent with environmental gradients be-


TABLE 1. Leaf morphological parameters and their defini-tions. The two capital letters in the definition demonstratethe linear distance between the corresponding two pointspre sented in Figure 1

Variable Definition

LA (cm2) Leaf areaLDM (g) Leaf dry massSLA (cm2 g–1) Specific leaf area (ratio of leaf area

to leaf dry mass)LL (cm) Leaf length=ABLW (cm) Leaf width=FGLPL (cm) Leaf petiole length=BCLE Leaf elongation (ratio of leaf length

to leaf width)LL/LPL Leaf length to petiole length ratioLWD Leaf widest division=AD/BDLBD Leaf bulgy division=AE/BENLT (ea) Number of leaf teethMDV (cm) Mean distance between veins=

2AB/ (NLT+1)FIG. 1. Illustrated diagram of leaf morphology measure-ments in Q. acutissima. The dashed lines perpendicular tothe midrib indicate the widest (above) and bulgy (below)part of the leaf lamina. The two lines are determined by theang les between the midrib and the edge of the lamina. Thepo sitions where the angle is the smallest and the largest re -pre sent the widest and bulgy part of the leaf lamina, respecti -ve ly. The leaf is divided into three fractions (top, middle, andbot tom) by these two lines. The capital letters along the linesre present the positions located at the midrib or the edge ofthe lamina (see Table 1 for more details).


FIG. 2. Three-dimensional surface plots for LA (A), LDM (B), SLA (C), LL (D), LW (E), LPL (F), LE (G), LL/LPL (H),LWD (I), LBD (J), NLT (K) and MDV (L) of Q. acutissima in the different gradients of water and light availability. The gray -sca le maps indicate that the darker chroma corresponds to the higher value of leaf morphological variable. The spline functionis employed in the curved surface fitting and n=120 for each variable contained climate information.


TA

BL

E 2

. Pea

rson

’s c

orre

lati

on c

oeff

icie

nts

betw

een

Q. a

cuti

ssim

ale

af m

orph

olog

ical

trai

ts a

nd c

limat

e pa

ram

eter

s (N

=12

0 fo

r ea

ch le

af tr

ait,

and

n=

24 fo

r ea

ch fa

ctor

of e

nvir

on-

men

ts w

ith

five

rep

licat

ions

). S

igni

fica

nce

leve

l: ns

p>

0.05

, *p

≤0.

05, *

* p≤

0.01

, ***

p≤

0.00

1

Var

iabl

eL

A

LD

M

SLA

LL

L

WL

PL

L

E

LL

/LP

L

LW

D

LB

D

NL

T

MD

V

LD

M

0.76

***

SLA

–0

.31*

**

–0.4

3***

LL

0.

91**

* 0.

73**

*–0

.21*

LW

0.

95**

* 0.

73**

*–0

.38*

**0.

75**

*

LP

L

0.43

***

0.32

***

–0.1

3ns

0.44

***

0.37

***

LE

–0.3

6***

–0.2

6**

0.33

***

0.03

ns–0

.63*

**–0

.05n

s

LL

/LP

L0.

11ns

0.11

ns0.

01ns

0.16

ns0.

07ns

–0.8

0***

0.07

ns

LW

D0.

02ns

0.12

ns0.

00ns

-0.0

5ns

0.09

ns0.

03ns

–0.2

2*

–0.0

6ns

LB

D–0

.05n

s–0

.06n

s–0

.02n

s-0

.18*

0.

05ns

–0.0

8ns

–0.2

9**

–0.0

3ns

0.66

***

NL

T0.

17ns

0.23

* 0.

02ns

0.31

***

0.06

ns0.

02ns

0.25

**

0.18

* 0.

16ns

–0.0

5ns

MD

V0.

51**

*0.

31**

*–0

.19*

0.

46**

*0.

48**

*0.

32**

*–0

.17n

s–0

.06n

s–0

.22*

–0

.13n

s–0

.69*

**

RW

C0.

24**

0.

28**

0.

40**

*0.

24**

0.

22*

0.27

**

–0.0

4ns

–0.1

5ns

–0.0

1ns

–0.1

4ns

0.20

* –0

.02n

s

RL

I0.

42**

*0.

71**

*–0

.72*

**0.

38**

*0.

43**

*0.

23*

–0.2

0*

–0.0

3ns

–0.1

6ns

–0.2

6**

0.08

ns0.

21*

cause of a high value under combined drought andshade conditions (Fig. 2F).

LE and LL/LPL were found to increase in shadeand decrease in sunlight with increasing water availa -bi lity, whereas LWD and LBD showed inverse trends(Fig. 2G-J). Conflicting variations also occurred in thegradients of light availability. LE and LL/LPL in creas -ed in drought and decreased under well-watered con-ditions, whereas LWD and LBD decreased in droughtand increased in well-watered conditions with in creas -ing light availability. These results caused either weakor zero correlations for these four variables with wa-ter and light availability (Table 2).

NLT only showed significant correlation with wa-ter availability (Table 2). Water scarcity resulted infewer leaf teeth. NLT increased along the gradient oflight availability with abundant irrigation. On the con-trary, NLT decreased in high irradiance when thedrought stress was a burden (Fig. 2K). MDV only show -ed significant correlation with light availability andmainly increased with increasing light availability ex-cept when the water resources were sufficient. Thema ximum values of MDV appeared when high irradi -ance and serious drought were simultaneously im-posed (Fig. 2L).

The correlation coefficients of leaf traits to RWCwere smaller than those of RLI, except for LPL, LL/LPL, and NLT, whose water-induced responses werehigher and under the diagonal of the bivariate diagram(Fig. 3). A high variation existed in the degree of re-sponse to light versus water. The response to light hada mean value of 0.17 (range 0.03-0.54), while the re-sponse to water had a mean value of 0.12 (range 0.02-0.30).

The overall determination coefficient of multipleli near regression was larger for RLI than that for RWC(Table 3). The mean values of absolute residuals were9.20 for RWC and 6.43 for RLI, which indicated a 95%confidence interval on the RWC prediction wi der thanthat of RLI (Fig. 4). The predicted values were signifi - cantly overestimated and significantly underestimatedin the lower left and upper right of the scatter plots,respectively, for both RWC and RLI pre dictions by t-tests (p0.05, *p≤0.05, **p≤0.01, ***p≤0.001

Variables Coefficient Coefficientfor RWC for RLI

Intercept –154.91* –150.99*

LA (cm2) –2.56ns –1.75ns

LDM (g) 115.50*** 239.98***

SLA (cm2 g–1) 0.33*** –0.36***

LL (cm) 2.67ns –20.15*

LW (cm) 29.23ns 65.04*

LPL (cm) 76.12** 25.75*

LE – 90.42*

LBD – –23.45***

NLT (ea) 0.55* –MDV (cm) – 0.57ns

r2 0.49*** 0.81***

when both were at the lower gradients because the li -near contour was above the spline contour with thesame values in the shadowed area (Fig. 5C). The areaoutside the shadowed area indicated an underestima-tion that frequently occurs at the higher gradients.

We applied the contours of SLA and NLT, whichare widely used as important leaf traits functionallylinked to climate, in the contour extraction method togauge the accuracy of predictions (Fig. 6 and onlinesup plementary material Fig. S1). The intersections di-rectly showed the environmental information in theplots. The mean values of absolute residuals were3.45 for RWC and 2.50 for RLI. Furthermore, a pair edt-test indicated that the values of the residuals weresignificantly smaller in the contour extraction methodthan in the multiple linear regressions for RWC (p=0.014) and RLI (p=0.038), respectively.

DISCUSSION

Sensitivity of leaf size and shape to climate

In the current research, leaf size was obviously re-stricted by the shortage of water and light as shown by

LA and LDM. Although leaf size variation in Q. a cu -tis sima can be partly linked to allometric factors, theecological strategy with respect to environmentalstress has an important role (Xu et al., 2008, 2009b).Variations in leaf size along the climatic gradientsmay result from higher water demand and overheat-ing of larger leaves because of photosynthesis-transpi-ration compromise and heat dissipation. The trend ofselecting relatively small leaves may be caused by o -


100

80

60

40

20

00 20 40 60 80 100

RWC predicted (%)

RWC

act

ual (

%)

100

80

60

40

20

00 20 40 60 80 100

RLI predicted (%)

RLI

act

ual (

%)

A

B

100

80

60

40

20

0100

80

60

40

20

0100

80

60

40

20

00 20 40 60 80 100

RWC (%)

RLI

(%

)

0.2792 0.3769

0.34070.29250.27070.2378

0.1326 0.1695

0.37690.2792

0.2378 0.2707 0.2925 0.3407

0.1326 0.1695

A

B

C

FIG. 4. Actual and predicted values of RWC (A) and RLI(B) from the multiple linear regressions shown in Table 3.Dashed lines indicate 95% confidence intervals.

FIG. 5. Contour plots for LDM of Q. acutissima in the dif-ferent gradients of water and light availability. The linear (A)and spline (B) functions are employed to test the effect offitted curve. The scatters are the locations of LDM in everytreatment, and the means are listed as the labels. Whenoverlapped, these two contour plot lines with the same val-ues intersected at two points (C). The diagram can be divid-ed into two parts by sketching the intersections together. Thearrow in the shadowed area indicates the average values ofRWC and RLI that should be overestimated by using the lin-ear function instead of the spline function. The arrows in thewhite area contrast with the shadowed area.


FIG. 6. Validation experiment for the RWC and RLI of Q. acutissima in the treatment of W1L1 (A), W3L1 (B), W1L2 (C),W2L2 (D), W3L2 (E), W4L2 (F), W1L3 (G) and W3L3 (H) by using the contour extraction method. Solid and dashed linesrefer to the contours of SLA and NLT, respectively. The intersections are established by the tool of drawing data. When twoin tersections appear in a plot, the mean value is used as a substitute.

ve rall resource limitation in stressful environments,which makes the construction of large leaves with ex-tensive vascular and cell-wall fractions overly expensi -ve and will reduce the investment in LDM (Niine -mets et al., 2007a). Furthermore, smaller leaves havean advantage in minimizing self-shading (Falster &Westoby, 2003).

An integrated variation of LA and LDM can beex pressed by SLA. This functional trait of leaves isim portant in connection with other traits and the cli-mate (Wright et al., 2004). To achieve different ameli -o rative aims under resource stress, the variations alongthe gradients of water and light availability exhibiteddivergence. Lower values in water-limited environ-ments tend to correspond to relatively high invest-ments in leaf defenses and long leaf lifespan. Shade-tolerant leaves have remarkably larger SLA, which isan accommodation to decrease self-shading of chlo -ro plasts in the abaxial surface of leaves (Quero et al.,2006). Another advantage is that the construction andmaintenance costs of the production of leaves withmo re symplastic component in shade are reduced (Lusket al., 2008).

LE, calculated as LL divided by LW, characterizesthe overall slenderness of the leaves (Niinemets et al.,2007b), and often changes with leaf size because of theeffect of allometry. However, plasticity to environ -men tal conditions also occurs (Tsialtas & Maslaris,2007; Xu et al., 2009b). Leaves became longer insteadof wider in the open field with the decrease of wateravailability. This correlation most likely allows theleaves to reduce transpiration by reducing the size ofthe boundary layer and to shed heat better in warmha bitats. Leaves were also narrower in the shade andmoist conditions, which is thought to be an adapta-tion to decrease self-shading and curliness of leaveswith high SLA.

Leaf shape was not only represented by the inte-grated changes of leaf major and minor axes, but thedistribution of leaf area fractions (partitioned by LWDand LBD) in the present research also includes addi-tional information. When LE increased, lamina areaslocated close to the leaf apex decreased, which repre-sents the narrow part of the leaf increasing in propor-tion to the whole leaf lamina. The cooperation be-tween LE and leaf area fractions was also displayedin the case of the formation of wider leaves, which arecapable to maximize their areas for light capture.

Synergy between leaf venation pattern and leaf dimen-

sion

In the present study, the changes of LPL were consis-tent with leaf size, which provides an available proxyfor leaf size from petiole dimensions when the leafblade was disfeatured (Jordan, 2011). The elongationof petioles will achieve optimal leaf display to dealwith the shade stress. The increases in the relative di -stance of LA from the stem by longer petioles can re -du ce the between-row shading. Furthermore, a leafwith a narrow blade can reduce both between-rowand within-row shadings (Takenaka, 1994). Leaf peti-ole and shape alternatively contributed to larger lightcapture in this research, shown by the similar changesin LE and LL/LPL, as plants had a tradeoff betweenthe need for increasing interception areas and sup-port structures. Increasing the investment in petiolesre quires the synthesis of more xylogens and limits thebiomass invested into functional leaf activity. Longerpetioles also have a disadvantage which leads to thebending of the shaded leaf in a moist habitat (Pickupet al., 2005).

Large leaves require disproportionately more massallocated to petioles and veins for mechanical support(Niklas et al., 2007), especially for thin leaves. A sha -de-tolerant leaf shape, whose centroid is far from theleaf base, requires higher vein density to withstandthe increased bending moment. Also, closer spacingof veins (decreased MDV) results in both higher wa-ter fluxes and carbon assimilation rates because ofshorter path lengths between veins and stomata (Bro-dribb et al., 2007; Brodribb & Field, 2010). However,in vestments in support will transform the investedbiomass into functional leaf activity. This inverse re-lationship is likely because of the displacement of thelamina tissue by non-photosynthetic venation tissue(Poorter et al., 2006). Xeromorphic leaves eliminatedthe dependence of veins because of the self-supportof lamina, which can rely on lamina cells (non-specificsupport) in addition to vasculature (Niinemets, 2001).

Characteristics of leaf teeth are functional traits toreflect a tradeoff between carbon uptake and waterloss. The relationship between leaf teeth and enhanc -ed sap flow may help explain why leaf teeth are moreabsent in drought environments where the water costassociated with teeth may be more important (Peppeet al., 2011). This pulse in gas-exchange activity maybe also adaptive in a shade condition because it mayextend the season of potential growth (Royer & Wilf,2006). In contrast, tooth-driven pulse in sap flow is


more muted for thin leaves in wet conditions, becausethe potential benefit is outweighed by the diminishingproportion of functional leaf tissues to support struc-tures. The present study highlights water availabilityas an important factor that affects NLT. Variations ofthe leaf-margin state in light availability may also be alikely effect of water availability because shade stresscauses less water consumption.

Application to predictions

The study results are consistent with the previous find -ings, which showed that seedlings of deciduous spe -cies are more tolerant to shade but not necessarily todrought (Quero et al., 2006; Sanz-Pérez & Castro-Diez, 2010; Zhu et al., 2012). The adaptive mecha-nisms are mainly to lessen shade stress, even underse rious drought. Also, the correlation coefficients andplasticity indices are larger under light availabilitycom pared to water availability. The fact that neithermultiple linear regression nor contour extraction isbet ter in estimating RWC than RLI indicates thatleaf traits respond less to water than to light availabi -li ty. This trend was also found in models predictinglarge-scale precipitation (Jacobs, 2002; Peppe et al.,2011).

The parameters taken into account ensure the ac-curacy of multiple linear regressions. The inclusion ofcertain leaf morphological variables (e.g. variables ofleaf-margin state), which are significantly correlatedwith the predicted climate, will improve the predic-tive power (Royer et al., 2012). Although some leaftraits exhibit multi-collinearity with each other, thismulti-collinearity may not be a problem for the pre-diction. The models can achieve accurate predictionswhen the overall determination coefficients r2 are lar -ge enough, because the combined linear relationshipsof explanatory variables will continue to apply in allca ses (Mela & Kopalle, 2002). The sample and amountof independent variables will also affect the coeffi-cients of each variable and the constants of the wholeregression functions to cause the instability of the pre -dictive system. Even if the slopes of the regressionsthat include the diverse variables show no statisticaldif ference, the probability of misestimating may in-crease toward the ends, which reflects an overestima-tion to the lower end and underestimation to the upperend in the present study. This bias is caused by the ri -gid characteristics of the linear model, which can notbe adjusted with the values of non-linear variables.

The contour extraction method in the present re-search had some advantages in climate prediction overmultiple linear regressions. First, the characteristicsof our prediction system are flexible for using splinecur ves. Therefore, a good foundation for predictiveaccuracy is established. Second, more alternatives areavailable in choosing leaf traits for prediction. Anytwo or more leaf traits that are not perfectly collinearcan locate the climate information. The contour ex-traction method is suitable for application to dead orfossil leaves, because some leaf traits (e.g. SLA) can-not be reliably reconstructed (Royer et al., 2007).Third, the bi-variables of environmental gradients canbe synchronously estimated, which means that thevariability of leaf traits should consider the effects ofdouble environmental factors at the same time. How-ever, the effects are discrete in multiple linear regres-sions because of one function for each environmentalfactor prediction. The ignored factor will introducean error in estimation of another factor, because at agi ven value of one factor, the leaf traits will also varysignificantly as a result of the effect of the other fac-tors (Peppe et al., 2011). This may be another reasonwhy the accuracy of multiple linear regressions is low-er than that of contour extraction method. However,the contour extraction method has been recognizedto have its own shortcomings in prediction. The valueintervals for the test set are limited because the con-tours are drawn based on the sample set. Any leaf who -se values of traits are beyond the range of the modelswill have no contour to extract. Therefore, this me -thod does not fit for leaves derived from different re-gions or growth phases as well as those abnormallyderived. Furthermore, the method cannot performba tch processing as one value of leaf trait correspondsto one contour to extract.

Implications and future research

Sensitivity of leaf physiognomy to climate establisheda foundation for constructing and improving leaf-cli-mate analysis methods for microclimate prediction.The oak species investigated in the present study sha -re a general trend of leaf traits that can be suited to atradeoff demand for resources as well as part of acon servative resource-use strategy (Valladares et al.,2000a; Xu et al., 2010). The response of the leaves caneven occur due to the effect of climate in the current-year growing season, which serves as a significant ex-tension for dynamic prediction by investigating theleaves in different temporal and spatial scales.


The contour extraction method in the present stu -dy is more precise compared with the multiple linearregressions, especially for the combined predictionsof extreme environments. However, the diversity ofthe sampled leaves may result in the invalidity of thismethod. Therefore, the allometric relationships ofleaf traits must be established to improve the applica-bility of the contour extraction method (Niklas et al.,2009). In addition, other environmental factors (e.g.air temperature, soil fertility, wind) may affect theleaf morphological plasticity in the field (Royer et al.,2008). Regional differences and genetic variabilitywill also introduce errors to leaf-climate relationships(Hameed et al., 2012). Further extensions, coupledwith more detailed experimental measurements ofthe traits and modified relationships between leaf andclimate, will enable leaf physiognomy to become a ro-bust proxy to macro- and microclimate.

ACKNOWLEDGMENTS

The present work was financially supported by theNational Natural Science Foundation of China (Nos.31200143 and 31270374). We are grateful to YinghuaWei and Yuanzu Xu for building the experimental e -quip ment and to Weihong Xu and Yue Yu for the mea -surement assistance.

REFERENCES

Blonder B, Violle C, Bentley LP, Enquist BJ, 2011. Vena-tion networks and the origin of the leaf economics spec-trum. Ecology Letters, 14: 91-100.

Brodribb TJ, Field TS, 2010. Leaf hydraulic evolution led asurge in leaf photosynthetic capacity during early an-giosperm diversification. Ecology Letters, 13: 175-183.

Brodribb TJ, Feild TS, Jordan GJ, 2007. Leaf maximumpho tosynthetic rate and venation are linked by hydrau -lics. Plant Physiology, 144: 1890-1898.

Dormann CF, Elith J, Bacher S, Buchmann C, Carl G, Car-ré G, Marquéz JRG, Gruber B, Lafourcade B, LeitãoPJ, et al., 2013. Collinearity: a review of methods to dealwith it and a simulation study evaluating their perform-ance. Ecography, 36: 27-46.

Du JX, Wang XF, Zhang GJ, 2007. Leaf shape based plantspecies recognition. Applied Mathematics and Compu-tation, 185: 883-893.

Falster DS, Westoby M, 2003. Leaf size and angle vary wi -de ly across species: what consequences for light inter-ception? New Phytologist, 158: 509-525.

Garnier E, Cortez J, Billès G, Navas ML, Roumet C, De-bussche M, Laurent G, Blanchard A, Aubry D, Bell-mann A, et al., 2004. Plant functional markers captureecosystem properties during secondary succession. E co -

lo gy, 85: 2630-2637.Greenwood DR, 2005. Leaf form and the reconstruction of

past climates. New Phytologist, 166: 355-357.Hameed M, Batool S, Naz N, Nawaz T, Ashraf M, 2012.

Leaf structural modifications for drought tolerance insome differentially adapted ecotypes of blue panic (Pa -ni cum antidotale Retz.). Acta Physiologiae Plantarum,34: 1479-1491.

Huff PM, Wilf P, Azumah EJ, 2003. Digital future for pale -o climate estimation from fossil leaves? Preliminary re-sults. Palaios, 18: 266-274.

Jacobs BF, 2002. Estimation of low-latitude paleoclimatesusing fossil angiosperm leaves: examples from the Mio -cene Tugen Hills, Kenya. Paleobiology, 28: 399-421.

Jordan GJ, 2011. A critical framework for the assessmentof biological palaeoproxies: predicting past climate andlevels of atmospheric CO2 from fossil leaves. New Phy-tologist, 192: 29-44.

Kato E, Nagano H, Yamamura S, Ueda M, 2003. Syntheticinhibitor of leaf-closure that reveals the biological im-portance of leaf-movement for the survival of legumi-nous plants. Tetrahedron, 59: 5909-5917.

Lusk CH, Reich PB, Montgomery RA, Ackerly DD, Ca -ven der-Bares J, 2008. Why are evergreen leaves so con-trary about shade? Trends in Ecology and Evolution, 23:299-303.

Mela CF, Kopalle PK, 2002. The impact of collinearity onre gression analysis: the asymmetric effect of negativeand positive correlations. Applied Economics, 34: 667-677.

Meziane D, Shipley B, 2001. Direct and indirect relation-ships between specific leaf area, leaf nitrogen and leafgas exchange. Effects of irradiance and nutrient supply.Annals of Botany, 88: 915-927.

Navas ML, Garnier E, 2002. Plasticity of whole plant andleaf traits in Rubia peregrina in response to light, nutri-ent and water availability. Acta Oecologica, 23: 375-383.

Niinemets Ü, 2001. Global-scale climatic controls of leafdry mass per area, density, and thickness in trees andshrubs. Ecology, 82: 453-469.

Niinemets Ü, Portsmuth A, Tena D, Tobias M, MatesanzS, Valladares F, 2007a. Do we underestimate the im-portance of leaf size in plant economics? Disproportio -nal scaling of support costs within the spectrum of leafphysiognomy. Annals of Botany, 100: 283-303.

Niinemets Ü, Portsmuth A, Tobias M, 2007b. Leaf shapeand venation pattern alter the support investments wi thinleaf lamina in temperate species: a neglected source ofleaf physiological differentiation? Functional Ecology,21: 28-40.

Niklas KJ, Cobb ED, Niinemets Ü, Reich PB, Sellin A, Shi -pley B, Wright IJ, 2007. “Diminishing returns” in thescaling of functional leaf traits across and within speciesgroups. Proceedings of the National Academy of Sciencesof the United States of America, 104: 8891-8896.


Niklas KJ, Cobb ED, Spatz HC, 2009. Predicting the allo -me try of leaf surface area and dry mass. American Jour-nal of Botany, 96: 531-536.

Peppe DJ, Royer DL, Cariglino B, Oliver SY, Newman S,Leight E, Enikolopov G, Fernandez-Burgos M, Her-rera F, Adams JM, et al., 2011. Sensitivity of leaf sizeand shape to climate: global patterns and paleoclimaticapplications. New Phytologist, 190: 724-739.

Pickup M, Westoby M, Basden A, 2005. Dry mass costs ofdeploying leaf area in relation to leaf size. FunctionalEcology, 19: 88-97.

Poorter H, Pepin S, Rijkers T, de Jong Y, Evans JR, KörnerC, 2006. Construction costs, chemical composition andpayback time of high- and low-irradiance leaves. Jour-nal of Experimental Botany, 57: 355-371.

Quero JL, Villar R, Marañón T, Zamora R, 2006. Interac-tions of drought and shade effects on seedlings of fourQuer cus species: physiological and structural leaf re-sponses. New Phytologist, 170: 819-834.

Roche P, Díaz-Burlinson N, Gachet S, 2004. Congruencyanalysis of species ranking based on leaf traits: whichtraits are the more reliable? Plant Ecology, 174: 37-48.

Royer DL, Wilf P, 2006. Why do toothed leaves correlatewith cold climates? Gas exchange at leaf margins provi -des new insights into a classic paleotemperature proxy.International Journal of Plant Sciences, 167: 11-18.

Royer DL, Wilf P, Janesko DA, Kowalski EA, Dilcher DL,2005. Correlations of climate and plant ecology to leafsize and shape: potential proxies for the fossil record.American Journal of Botany, 92: 1141-1151.

Royer DL, Sack L, Wilf P, Lusk CH, Jordan GJ, NiinemetsÜ, Wright IJ, Westoby M, Cariglino B, Coley PD, et al.,2007. Fossil leaf economics quantified: calibration,Eocene case study, and implications. Paleobiology, 33:574-589.

Royer DL, McElwain JC, Adams JM, Wilf P, 2008. Sensiti -vi ty of leaf size and shape to climate within Acer rubrumand Quercus kelloggii. New Phytologist, 179: 808-817.

Royer DL, Kooyman RM, Little SA, Wilf P, 2009. Ecologyof leaf teeth: a multi-site analysis from an Australiansubtropical rainforest. American Journal of Botany, 96:738-750.

Royer DL, Peppe DJ, Wheeler EA, Niinemets Ü, 2012. Ro -les of climate and functional traits in controlling tooth edvs. untoothed leaf margins. American Journal of Botany,99: 915-922.

Sanz-Pérez V, Castro-Díez P, 2010. Summer water stressand shade alter bud size and budburst date in three me -di terranean Quercus species. Trees, 24: 89-97.

Schimek MG, 2000. Estimation and inference in partiallylinear models with smoothing splines. Journal of Statis-tical Planning and Inference, 91: 525-540.

Shepherd T, Griffiths DW, 2006. The effects of stress onplant cuticular waxes. New Phytologist, 171: 469-499.

Shipley B, Lechowicz MJ, Wright IJ, Reich PB, 2006a. Fun-

damental trade-offs generating the worldwide leaf eco-nomics spectrum. Ecology, 87: 535-541.

Shipley B, Vile D, Garnier É, 2006b. From plant traits toplant communities: a statistical mechanistic approachto biodiversity. Science, 314: 812-814.

Sisó S, Camarero JJ, Gil-Pelegrín E, 2001. Relationship be-tween hydraulic resistance and leaf morphology in broad -leaf Quercus species: a new interpretation of leaf loba-tion. Trees, 15: 341-345.

Takenaka A, 1994. Effects of leaf blade narrowness andpetiole length on the light capture efficiency of a shoot.Ecological Research, 9: 109-114.

Traiser C, Klotz S, Uhl D, Mosbrugger V, 2005. Environ-mental signals from leaves – a physiognomic analysis ofEuropean vegetation. New Phytologist, 166: 465-484.

Tsialtas JT, Maslaris N, 2007. Leaf shape and its relation-ship with leaf area index in a sugar beet (Beta vulgatisL.) cultivar. Photosynthetica, 45: 527-532.

Valladares F, Martínez-Ferri E, Balaguer L, Pérez-CoronaE, Manrique E, 2000a. Low leaf-level response to lightand nutrients in Mediterranean evergreen oaks: a con-servative resource-use strategy? New Phytologist, 148:79-91.

Valladares F, Wright SJ, Lasso E, Kitajima K, Pearcy RW,2000b. Plastic phenotypic response to light of 16 con-generic shrubs from a Panamanian rainforest. Ecology,81: 1925-1936.

Viscosi V, Lepais O, Gerber S, Fortini P, 2009. Leaf mor-phological analyses in four European oak species (Quer -cus) and their hybrids: a comparison of traditional andgeometric morphometric methods. Plant Biosystems,143: 564-574.

Westoby M, Wright IJ, 2003. The leaf size – twig size spec-trum and its relationship to other important spectra ofvariation among species. Oecologia, 135: 621-628.

Westoby M, Wright IJ, 2006. Land-plant ecology on the ba-sis of functional traits. Trends in Ecology and Evolution,21: 261-268.

Westoby M, Falster DS, Moles AT, Vesk PA, Wright IJ,2002. Plant ecological strategies: some leading dimen-sions of variation between species. Annual Review ofEcology and Systematics, 33: 125-159.

Whitfield J, 2006. The cost of leafing. Nature, 444: 539-541.Wolfe JA, 1995. Paleoclimatic estimates from Tertiary leaf

assemblages. Annual Review of Earth and Planetary Sci-ences, 23: 119-142.

Wright IJ, Reich PB, Westoby M, Ackerly DD, Baruch Z,Bongers F, Cavender-Bares J, Chapin T, CornelissenJHC, Diemer M, et al. 2004. The worldwide leaf econo -mics spectrum. Nature, 428: 821-827.

Wright IJ, Ackerly DD, Bongers F, Harms KE, Ibarra-Man riquez G, Martinez-Ramos M, Mazer SJ, Muller-Landau HC, Paz H, Pitman NCA, et al., 2007. Relation-ships among ecologically important dimensions of planttrait variation in seven neotropical forests. Annals of



Bo tany, 99: 1003-1015.Xu F, Guo W, Xu W, Wang R, 2008. Habitat effects on leaf

morphological plasticity in Quercus acutissima. Acta Bi-ologica Cracoviensia Series Botanica, 50: 19-26.

Xu F, Guo W, Wang R, Xu W, Du N, Wang Y, 2009a. Leafmovement and photosynthetic plasticity of black locust(Robinia pseudoacacia) alleviate stress under differentlight and water conditions. Acta Physiologiae Plantarum,31: 553-563.

Xu F, Guo W, Xu W, Wei Y, Wang R, 2009b. Leaf mor -pho logy correlates with water and light availability: what

consequences for simple and compound leaves? Pro -gress in Natural Science, 19: 1789-1798.

Xu F, Guo W, Xu W, Wang R, 2010. Effects of light inten-sity on growth and photosynthesis of seedlings of Quer-cus acutissima and Robinia pseudoacacia. Acta Ecologi-ca Sinica, 30: 3098-3107.

Zhu Y, Kang H, Xie Q, Wang Z, Yin S, Liu C, 2012. Patternof leaf vein density and climate relationship of Quer cusvariabilis populations remains unchanged with environ-mental changes. Trees, 26: 597-607.

sensitivity of leaf physiognomy to climate: applications to habitat … et al._biological... ·...

Documents