Spacing of silver birch seedlings grown in containers of equal ...

Summary Silver birch (Betula pendula Roth) seedlingswere grown individually in containers arranged in rows radiat-ing from a central point (Nelder plot) at densities spanning therange from 207 to 891 plants m– 2. Height of one set of seed-lings was measured at weekly intervals and additional seed-lings were harvested each week for dry mass and leaf area mea-surements. Height and shoot:root dry mass ratio increased withplant density. Seedling-to-seedling variability in dry mass, butnot height, increased with increasing plant density. The red tofar-red (R:FR) photon ratio in horizontally propagated radia-tion decreased with increasing density, even when plant densi-ties and leaf area index values were low.

In a separate experiment, elongating internodes of seedlingswere irradiated locally by red and far-red light emitting diodesand stem elongation measured. Far-red light markedly in-creased stem elongation, suggesting that changes in light qual-ity sensed by growing internodes are involved in the observedresponses to growth density.

Keywords: growth analysis, growth density, neighbors, phyto-chrome, variability, variance covariate.

Introduction

Plant spacing, or density, is an important management variablein the production of forest nursery stock. In the case of con-tainer grown stock, plant density is usually directly related tocontainer volume, although the relationship can be altered bychanging the ratio of container width to depth or by addingspace between containers.

There is only limited information about how above-groundspacing affects tree seedling growth and morphology inde-pendently of belowground effects of plant density on waterand nutrient resource availability. Aphalo and Rikala (2003)reported on the effect of a limited range of container nurseryseedling densities including effects on nursery stock proper-ties and performance after transplanting to a forest site. How-ever, additional information could lead to improved methods

for producing seedlings in containers and a better understand-ing of how tree seedlings are adapted to compete with othervegetation.

Plant morphology is strongly affected by plant spacing,which affects both the availability and quality of light. Themain effect of light quality on plants is mediated by phyto-chromes, a family of photoreceptors capable of sensing thered:far-red (R:FR) photon ratio of incident light (e.g., Smith1981, 1994, Casal and Smith 1989, Smith and Whitelam1997). The responses of plants to light availability are relatedto habitat preference, shade-avoiding growth responses beingmore marked in sun plants than in shade plants (Morganand Smith 1979).

Beside avoidance mechanisms, tree seedlings adapt to shadethrough a variety of morphological and physiological re-sponses, including changes in specific leaf area (SLA). Someshade responses are regulated by the R:FR light ratio (Leeet al. 1996). However, even in the absence of shading, plantssense the presence of neighboring plants through the R:FR ra-tio of the light that they reflect and make anticipatory adaptiveresponses (Aphalo and Ballaré 1995, Ballaré 1999, Aphalo etal. 1999 and references therein).

Height growth of silver birch seedlings is stimulated by sup-plemental lateral irradiation with FR light (Aphalo and Leh-to 2001), which simulates the presence of neighboring plants.In experiments with canopies of three tree species, each grownat a range of densities, Gilbert et al. (2001) observed a strongcorrelation between extension growth and the phytochromephotoequilibrium estimated from measurements of R:FR ratioof horizontally propagated light within the canopy. The esti-mated photoequilibrium was in turn correlated with projectedleaf area (projected on the target tree and used as an estimate ofleaf area in the neighborhood of the target tree that is capableof scattering radiation in the direction of the target tree).

Data from monospecific canopies of herbaceous plants indi-cate that photomorphogenic responses to plant density (rapidelongation of shorter plants) reduces heterogeneity in plantheight relative to heterogeneity in plant dry mass (Ballaré

Tree Physiology 26, 1227–1237© 2006 Heron Publishing—Victoria, Canada

Spacing of silver birch seedlings grown in containers of equal sizeaffects their morphology and its variability

PEDRO J. APHALO1–3 and RISTO RIKALA4

1 Faculty of Forestry, University of Joensuu, P.O. Box 111, 00101 Joensuu, Finland2 Current address: Department of Biological and Environmental Sciences, P.O. Box 65, 0014 University of Helsinki, Finland3 Corresponding author ([email protected])4 Finnish Forest Research Institute, Suonenjoki Research Station, 77600 Suonenjoki, Finland

Received June 20, 2005; accepted November 11, 2005; published online June 1, 2006

at Pennsylvania State University on M

ay 16, 2016http://treephys.oxfordjournals.org/

Dow

nloaded from

http://treephys.oxfordjournals.org/

et al. 1994, Ballaré and Scopel 1997).Nelder plots (Nelder 1962), comprising rows of plants radi-

ating from a central point, have been used in previous studiesof the physiological basis of tree seedlings responses to plantdensity (e.g., Doran et al. 2001). Ritchie (1997) used Nelderplots to study the relationship between horizontally propa-gated R:FR light ratio and height growth of potted Douglas-fir(Pseudotsuga menziesii (Mirb.)) seedlings. He also measuredR:FR in Nelder plots of larger field-grown trees.

In the present study, we used Nelder plots of potted seed-lings of the broadleaf species, silver birch (Betula pendulaRoth) to study seedling responses to growth density. The ob-jective was to analyze how density affects inter-plant variabil-ity in growth, morphology and mineral nutrient content and totest the hypotheses that: (1) morphological responses to shadereduce variability in plant height relative to variability in drymass; and (2) silver birch seedlings can respond to plant den-sity by sensing the R:FR ratio of horizontally propagated light.

Materials and methods

Experiments to study seedling responses to density

Silver birch seedlings were grown from seeds obtained from aseed orchard in central Finland. The seeds were germinated inpeat-filled boxes. On Week 23 of the year, when the first trueleaf was visible, the seedlings were transplanted to individualcontainers (165 ml, SC-10 super cell cone-tainers, Stuewe andsons, Corvallis, OR) filled with fertilized and limed peat (Treeseedling peat, Vapo Co., Finland; pH 5.1) and grown for oneseason. The containers were placed in purpose-built racks withrows radiating from a central point following a Nelder patternof increasing spacing between arcs. Each rack covered a66°30′ sector of a circle with 11 spokes 6°39′ apart andspanned 12 arcs at densities between 137 and 1363 seedlingsm– 2 (Figure 1). Two border rows of seedlings surrounded themeasured plants on all sides. Consequently the studied plantdensities were from 207 to 891 seedlings m– 2 (Figure 1). In a

Nelder plot, arc numbers are linearly related to the logarithmof density. There were two groups of four racks and withineach group the racks were arranged in a successive rotation of90°, forming a circle with aisles for access.

Throughout the experiment, seedlings were suppliedweekly with a fertilizer solution (Superex 6, Kekkilä Oy, Fin-land) directly applied to the peat surface. Total amounts of nu-trients in the peat's base fertilization plus the additional fertil-ization (micronutrients) were per seedling: nitrogen (N) = 17 +100 mg; phosphorus (P) = 12 + 18 mg; and potassium (K) = 24+ 120 mg. The experiment, which was conducted in a venti-lated polyethylene tunnel greenhouse, was performed in twosuccessive years. In the first year, there were four seedling har-vests, whereas in the second year there was only a single har-vest, which was made after leaf fall.

For the Year 1 harvests, seedlings from four randomly cho-sen spokes of the Nelder plot, one from each of four racks,were harvested every other week (Weeks 26 to 38 of the year).Height, leaf area, root collar diameter and dry mass of leaves,stems and roots were measured. After each harvest, the re-maining spokes were moved to close gaps in the canopy cre-ated by the removal of seedlings for analysis. Additionalharvests, were made at the time of elongation growth cessation(Week 35) and after leaf fall (after Week 40). At each of theseharvests, eight seedlings were taken per density (i.e., arc of theNelder plot), two from each of four seedling racks. In this case,each seedling main stem was cut lengthwise into 10 cm sec-tions, and the dry mass and leaf area of each section measured.

In Year 2, nondestructive height measurements were madestarting 3 weeks after transplanting. Measurements were madeon the same 12 seedlings per density at each date every weekduring the period of most rapid growth and every other weekwhen growth slowed at the end of the summer. At the singleYear-2 harvest, all seven non-border spokes of four racks ofseedlings were harvested.

In Year 1, nitrogen concentration was measured on pooled(per rack) and ground samples with a CHN analyzer (CHN-

1228 APHALO AND RIKALA

TREE PHYSIOLOGY VOLUME 26, 2006

Figure 1. Sections of a Nelder plotwith silver birch seedlings raised in165 ml containers and held in radiatingrows in purpose built racks.



Dow

nloaded from


600, Leco Corp., St. Joseph, MI). In Year 2, similarly pooledand ground samples were sent to the Waikato Stable IsotopeUnit, University of Waikato, New Zealand for separate deter-minations of carbon isotope composition and nitrogen concen-tration by GC-MS after combustion (Europa 20/20 tracemass,PDZ Europa Ltd., Northwich, U.K.).

Air temperature and humidity and photosynthetically activephoton irradiance (I) in the greenhouse were measured with aHumitemp probe (HMI33, Vaisala, Helsinki, Finland) andwith a quantum sensor (LI-190-SB, Li-Cor, Lincoln, NE), re-spectively, and the data stored by a datalogger (21-X, Camp-bell Scientific, Shepshed, U.K.). In the first year, the R:FRphoton ratio of horizontally propagated radiation was mea-sured on Weeks 27 and 28 (4 and 5 weeks after transplanting)around solar noon on days with clear skies. Measurementswere made with an R:FR sensor (SKR110, Skye Instruments,Llandrindod Wells, U.K.) with a cylindrical light collector at-tached (Ballaré et al. 1987). For the measurements, individualseedlings were replaced with the sensor and readings takenalong one spoke on each of the four racks. Additional mea-surements were taken in the space between the racks, at least2 m from the nearest seedling.

Responses to R and FR irradiation

In a controlled-environment chamber, seedlings were grownfrom seed in identical pots (i.e., 165 ml cone-tainers). Seed-lings were thinned to one per pot as the first true leaf becamevisible. The pots were filled with a mix of unfertilized peat andquartz sand (9:1, v/v), and the seedlings fertilized with a bal-anced nutrient solution (Superex 6) with concentrations in-creasing in time following a relative addition rate of 15% perday. The conditions in the growth room (GR-77, Conviron,Winnipeg, Canada) were: day/night air temperature of20/15 °C, relative humidity about 70%, an 18-h photoperiodand I of 350 µmol m– 2 s–1 with an R:FR photon ratio of 1.15.Light was supplied by both fluorescent tubes (VHO 215W,cool white, Sylvania) and 60-W incandescent lamps.

The treatment consisted in irradiating growing internodesfor 6 days with red (nominal 660 nm) and far-red (nominal735 nm) light emitting diodes (LEDs, QDDH66002 and QDD-H73502, Quantum Devices, Barneveld, WI; Figure 2). TheLEDs were located 2 cm from the target internode and pro-vided an irradiance at the stem surface of 80 µmol m– 2 s–1. Thetreatment started when the seedlings were 45 days old. Seed-lings of similar size were grouped into 14 blocks of three seed-lings and the three treatments assigned at random within eachblock.

Statistical analysis

Each dataset collected from a Nelder experiment included fourracks, with individual seedlings as experimental units. How-ever, the chemical analyses were done on one pooled sampleper rack and seedling density. Univariate data were analyzedby fitting linear models generalized least squares and linearmixed effects functions from the nlme library version 3.1(Pinheiro and Bates 2000) running under R version 1.4.1 or1.6.2 (Ihaka and Gentleman 1996). The logarithm of density

(numeric variable or covariate; fixed effect) and the rack (fac-tor or categorical variable; random effect) were initially bothincluded in the models. The random effect for the racks wassmall and could not be estimated with precision and was there-fore left out of the model in the definitive analyses. The loga-rithm of density was included as a fixed effect, both as a linearterm and a quadratic term. Probabilities shown for the effect ofdensity on the mean of the response variables are for the jointtest of these two terms (see Appendix for models). The nor-mality of the distribution of the standardized residuals was as-sessed graphically with quantile–quantile normal plots, and inthe case of specific stem length (SSL) and sturdiness, it wasnecessary to use a logarithmic transformation of the data. Ho-mogeneity of variance was initially assessed visually in plotsof standardized residuals versus fitted values and of standard-ized residuals versus the logarithm of the density. Insome plots, it was evident that variability between seedlingschanged with plant density. Consequently, to test for the effectof the treatment (logarithm of density) on the variability, themodel was refit, for each response variable, adding a weightterm with a power of variance function with the logarithm ofdensity as a variance covariate. This new fit, allowing for het-erogeneous variance dependent on the density, was comparedwith the initial fit, which assumed homogeneous variance bymeans of a log likelihood test, which follows a χ2 distribution.The probabilities from this test indicate whether the variancechanged significantly with plant density. Furthermore, if theheterogeneous variance model was significantly better thanthe homogeneous variance model, the probabilities reportedfor the effect of density on the mean of the response variableare from this fit. The use of a variance covariate is analogous tothe use of an ordinary covariate in a statistical model, but it isused to describe the relationship between a covariate and thevariance of the dependent variable rather than describing itsrelationship to the mean.

Multivariate data of the height profiles for leaf area and drymass were analyzed with the MANOVA function of R. ForMANOVA, significance was tested with the Pillai-Barlett test.

In the case of data from sequential harvests, a nonlinearmixed effects model (nlme function of the nlme package) wasused to fit a growth curve with a logistic function, with rack asa grouping factor for the random effects. For the analysis ofheight growth in Year 2, a similar analysis was done, but withseedling as a grouping factor for the random effect becauseheight was repeatedly measured on the same seedlings. (InYear 1, height as well as dry mass and leaf area were measuredon different seedlings at each date). In all cases, the assump-tion of homogeneity of variance did not hold and the fittedvalue was used as a variance covariate with a power of vari-ance function (Pinheiro and Bates 2000).

Before computing the ANOVAs, values from three sup-pressed seedlings were culled from the data from the harvest atthe time of height growth cessation of Year 1. At the harvest af-ter leaf fall in Year 1, five plants out of 64 were not harvested.At the harvest in Year 2, values for four suppressed seedlingswere culled out of 224 potentially measured seedlings ofwhich three were not harvested.

TREE PHYSIOLOGY ONLINE at http://heronpublishing.com

SPACING AND THE MORPHOLOGY OF BIRCH SEEDLINGS 1229



Dow

nloaded from


Height growth data from the R and FR irradiation experi-ment were subject to repeated measures analysis, with seed-lings and blocks as grouping factors for random effects. Weused the linear mixed effects function of the nlme library,Version 3.1 (Pinheiro and Bates 2000) and R, Version 1.3.0,(Ihaka and Gentleman 1996).

Results

Red to far-red photon ratio

As reported by Aphalo et al. (1999), the R:FR ratio of horizon-tally propagated radiation decreased with increasing density,from about 0.75 outside the canopy to 0.6 at a density of207 seedlings m– 2 to 0.32 at 890 seedlings m– 2. At the lowerplant densities, the decrease in R:FR was a result of an increasein FR caused by reflection by the foliage of neighboring seed-lings, but at higher densities absorption of R also played a role.

Year 1, sequential harvests and measurements

Density had a large effect on asymptotic final height and asmall effect on the scal shape parameter of the logistic func-tion fitted. Density had a smaller, but nonetheless highly sig-nificant, effect on maximum green leaf area, but no effect onfinal total dry mass (Table 1). For these two variables, densityhad no significant effect on the time at which 50% of the as-ymptote was attained or on the shape of the fitted logisticcurve. Furthermore, this time was almost the same for heightand leaf area growth, but occurred 10 days later for dry massgrowth as would be expected for a later cessation of stemdiameter and root growth.

Year 1, at cessation of height growth

The vertical distribution of leaf area was affected by density(Figure 3; MANOVA: P < 0.0001, with P < 0.05 from uni-variate ANOVAs for the layers 0–10, 20–30, 30–40 and

> 40 cm). Total leaf area per seedling was smaller at low densi-ties compared with intermediate and high densities and itsvariability increased with increasing density (Figure 4a; Ta-ble 2). The smaller leaf areas of seedlings grown at low den-sity, despite a larger number of leaves, is explained by thesmaller size of the leaves (Figures 4b and 4c; Table 2). Thegreater number of leaves at low density had two causes: fewerleaves senesced and dropped early; and there were more leavesper branch, whereas the number of nodes on the main stem wasunaffected (overall median number of nodes = 15, first quar-tile = 14, third quartile = 16; Figures 4c and 4d; Table 2). Thenumber of leaves per seedling was more variable at low densi-ties when more leaves were located on branches; however, thesignificance of this effect depended on data from one seedlingthat had 31 leaves.

At cessation of height growth in Year 1, total dry mass perseedling was unaffected by density, but variability was higherat high densities (Figure 5a; Table 2). Dry mass allocation tostems and roots was altered, but allocation to leaves was not(Figure 5b; Table 2). Root mass ratio was highest at the lowestdensity.

Specific stem length (length per unit mass) was unaffectedby density (data not shown; Table 2) because seedlings grownat low density had shorter stems, but with less mass (Figure 5d;Table 2). Leaf area ratio (LAR) and its variability increasedwith density (Figure 5c; Table 2), which explains the increasein leaf area despite no change in total dry mass allocation toleaves.

Nitrogen content per seedling was unaffected by density, al-though its variability was greater at high plant densities (Fig-ure 6a; Table 2). However, leaf nitrogen content ration per unitdry mass basis, was higher in seedlings grown at high density(Figure 6b; Table 2). Nitrogen concentration in stems was sig-nificantly affected by density, but the pattern was not as clearas in leaves. Nitrogen concentration in roots was unaffectedby density. Nitrogen concentration per unit leaf area slightly



Figure 2. Light emitting diodes adja-cent to a seedling internode in the irra-diation experiment.



Dow

nloaded from


decreased with increasing density, whereas SLA increased(Figure 7; Table 2). Variability in SLA also increased withdensity.

Year 1, after leaf fall

After leaf fall, total dry mass of seedlings at a low density wasslightly more than that of seedlings grown at high density (datanot shown; Table 2). Similarly root collar diameters weregreater in seedlings at the low densities (data not shown; Ta-ble 2). More dry mass was allocated to roots at the lower den-sities. At low densities, plants were shorter (data not shown;Table 2) with slightly lower specific stem length (data notshown; Table 2).

Nitrogen content per seedling was higher at low density thanat high density (data not shown; Table 2). Nitrogen concentra-tions in stems and roots were unaffected by density (Table 2).

Year 2, height growth throughout the growing season

Height increased with increasing plant density (Figure 8).Based on an analysis with a nonlinear mixed effects model tofit a logistic curve, we found a significant effect of density onthe asymptotic final height (P < 0.0001) with a small, but sig-

nificant effect on the timing of 50% final height (P < 0.0001)and steepness of the curve (P = 0.13). The effect of density ontiming was such that height growth stopped slightly earlier atlow densities than at high densities: the largest difference was1.2 days for attaining half final height (see Table 1 for parame-ter estimates).

Year 2, after leaf fall

Seedlings grown at low densities were the shortest and the re-sponse was not linear with the logarithm of density, but tendedto level off at the highest densities (Figure 9d; Table 3). Spe-cific stem length was smaller in seedlings grown at low density(data not shown; Table 3).

Total dry mass of the seedlings was unaffected by density;however, dry mass variability increased with density (Fig-ure 9a; Table 3). Density had no effect on root collar diameter,but it increased variability in root collar diameter (data notshown; Table 3). Dry mass allocation to roots and stems



Table 1. Fit of the logistic function to growth data from sequential harvests (Year 1) or repeated measurements (Year 2). Abbreviations: Asym isthe asymptote (or the size at the end of the growth season); xmid is the week of the year when 50% of Asym was reached; and scal is a scale param-eter giving the steepness of the curve. For each parameter, b0 is the value at a position between Arcs 6 and 7 (or a density of 435 seedlings per m2);b1 is the slope as a function of arc number (where arc number is linearly related to the logarithm of density); b2 is the corresponding quadratic term;and df1 and df2 are the degrees of freedom of the F value. The F value is for a joint test of significance for the effect of density (b1 and b2) for thethree parameters. For Year 2, xmid b2 = –0.0136 was significantly different from zero.

Asym b0 Asym b1 Asym b2 xmid b0 scal b0 df1, df2 F P

Year 1Height (mm) 460 –23.2 –3.59 30.3 1.41 6, 212 25.1 < 0.0001Area (cm2) 444 –16.7 –2.11 30.4 1.13 6, 180 8.78 < 0.0001Dry mass (g) 4.00 0.0 0.0 31.7 1.38 6, 212 0.95 0.46

Year 2Height (mm) 593 –32.7 –6.57 29.6 1.31 6, 847 91,805 < 0.0001

Figure 3. Vertical profile of leaf area of seedlings grown at differentdensities (means of two adjacent arcs of the Nelder experiments). Ab-breviations and symbols: LAI = leaf area index; d = average growingdensity (two arcs); and numbers in reverse print = total leaf area perseedling. Measurements were made at the time of cessation of heightgrowth (Year 1 experiment). Values are means for 16 seedlings (twoseedlings from each of four racks and two adjacent densities).

Figure 4. Leaf areas of seedlings grown at different densities. (a) Totalleaf area per seedling; (b) mean size of leaves; (c) number of leavesper seedling; and (d) number of nodes on the main stem with missingleaves. Measurements were made at the time of cessation of heightgrowth (Year 1 experiment). Values are means for eight seedlings(two seedlings from each of four racks). Error bars indicate ± onestandard error.



Dow

nloaded from


changed drastically in response to density, with more roots atlow densities (Figure 9b; Table 3). The ratio of stem diameterto seedling height, was higher and more variable in seedlingsgrown at low density (Figure 9c; Table 3).

Nitrogen content per seedling and nitrogen concentrationsin stems and roots were not significantly affected by density(Figures 10a, 10c and 10d; Table 3). The proportion of nitro-gen allocated to roots decreased with increasing density (Fig-ure 10b; Table 3), in parallel with the changes in root massratio.

Carbon isotope ratio was strongly affected by plant density(Figure 11; Table 3), indicating that intercellular carbon diox-ide (CO2) concentration (C i) was on average lower in plantsgrown at low density. Low C i can be the result of water stresscausing stomatal closure or high photosynthetically active ra-diation (PAR) increasing the internal demand for CO2.

Light quality experiment

Far-red light from LEDs markedly and quickly increased stemelongation (Figure 12; P < 0.0001 for the FR treatment by time



Table 2. Summary of ANOVA from linear mixed-effects model fits to data from the experiment of Year 1. Abbreviations: LMR = leaf mass ratio;SMR = stem mass ratio; RMR = root mass ratio; LAR = leaf area ratio; SLA = specific leaf area; SSL = specific stem length; leaf [Nmass] = nitrogenconcentration in leaves expressed per unit dry mass; leaf [Narea] = nitrogen concentration in leaves expressed per unit area; stem (root) [Nmass] = ni-trogen concentration in stems (roots) expressed per unit dry mass; seedling N = nitrogen content per seedling; and LNR = leaf nitrogen ratio (pro-portion of nitrogen content of whole plant that was in leaves).

Effect of density on mean Effect on variability

df number df density F P L ratio P

Harvest 1Height 2 55 13.73 < 0.0001 0.384 0.536Leaf area 2 55 9.15 < 0.0001 4.473 0.034No. leaves 2 55 8.65 0.0005 12.016 0.0005No. dropped 2 55 43.49 < 0.0001 1.526 0.2167No. nodes main 2 55 0.981 0.382 0.431 0.511Collar diam. 2 55 1.14 0.327 4.107 0.0427Dry mass 2 55 0.42 0.660 7.364 0.0067LMR 2 55 0.60 0.552 0.601 0.438SMR 2 55 50.79 < 0.0001 0.666 0.414RMR 2 55 31.69 < 0.0001 4.909 0.027LAR 2 55 9.33 0.0003 5.355 0.0207SLA 2 55 25.02 < 0.0001 3.936 0.0472SSL 2 55 1.70 0.192 2.660 0.103Leaf [Nmass] 2 26 8.76 0.0012 2.724 0.099Leaf [Narea] 2 26 6.67 0.0046 0.0001 0.99Stem [Nmass] 2 19 9.63 0.0013 0.368 0.544Root [Nmass] 2 26 0.17 0.845 2.945 0.086Seedling N 2 26 2.46 0.105 5.124 0.0236LNR 2 26 8.27 0.0005 0.144 0.704

Harvest 2Height 2 52 9.47 0.0003 0.355 0.551Dry mass 2 52 3.23 0.0476 2.404 0.121SWR 2 52 44.10 < 0.0001 0.036 0.849RWR 2 52 44.10 < 0.0001 0.036 0.849SSL 2 52 6.12 0.0041 10.699 0.0011Stem [Nmass] 2 26 1.28 0.295 0.177 0.674Root [Nmass] 2 26 1.98 0.159 1.144 0.2849Seedling N 2 19 1.61 0.226 5.938 0.015

Figure 5. Dry mass of seedlings grown at different densities. (a) Totaldry mass per seedling; (b) mass ratios; (c) leaf area ratio; and (d)height. Measurements were made at the time of cessation of heightgrowth (Year 1 experiment). Values are means for eight seedlings(two seedlings from each of four racks). Error bars indicate ± onestandard error.



Dow

nloaded from


interaction), whereas red light from LEDs had little effect onstem elongation (P = 0.34 for the R treatment by time interac-tion).

Discussion

Interactions between neighboring plants can be either positive(facilitative) or negative (competitive). Such effects must havea role in determining the outcome of interspecific competition(e.g., Callaway and Walker 1997), but they may be equally rel-evant to an understanding of intraspecific interactions. In ourexperiments, belowground interactions were prevented byraising seedlings in containers and supplying them with equalamounts of mineral nutrients. The only possible effect ofneighboring plants on the environment of roots was throughsoil temperature, with lower soil temperatures measured (datanot shown) at high densities because of shading.

Seedling dry matter accumulation decreased slightly withplant density according to measurements at the end of theYear 1 experiment. However, plant dry mass was unaffected byplant density in Year 2 when 28 instead of eight seedlings were

measured at each density. This result probably reflects the ef-fects of plant density on both environmental conditions andcompensatory physiological and morphological responses ofthe plants. When height growth ceased, leaf area index (LAI)ranged from about 44 at the highest density to about 7.5 at thelowest density. Thus, the main alterations in the environmentwith increasing density were likely: (1) decreased availableenergy for photosynthesis per unit leaf area; and (2) decreasedtranspiration demand (less incident energy per unit leaf area,higher air humidity, less wind). The main adjustments in plantmorphology with increasing plant density were: (1) a reduc-tion in root mass ratio; (2) an increase in stem height; and (3)the formation of thinner but larger leaves (changes in LAR andSLA).

In the experiment of Aphalo and Rikala (2003), dry mass ofbirch seedlings measured after leaf fall decreased with plant



Figure 9. (a) Seedling dry mass per seedling; (b) root mass ratio; (c)diameter:height ratio; and (d) height versus plant density. Measure-ments were made after leaf fall (Year 2 experiment). Values are meansfor 28 seedlings, seven from each of four Nelder racks. Error bars in-dicate ± one standard error.

Figure 7. (a) Leaf nitrogen (N) concentration per unit projected area;and (b) specific leaf area (SLA). Measurements were made at the timeof cessation of height growth (Year 1 experiment). Values are meansfor eight seedlings (four pooled samples for N determination). Errorbars indicate ± one standard error.

Figure 6. (a) Nitrogen (N) content per seedling; (b) N concentrationper unit dry mass in leaves; (c) stems; and (d) roots. Measurementswere made at the time of cessation of height growth (Year 1 experi-ment). Values are means of four samples of two seedlings, each from adifferent Nelder rack. Error bars indicate ± one standard error.

Figure 8. Height versus plant density. Values are means for 12 seed-lings, three from each of four racks. The same seedlings were mea-sured on different dates, indicated as week of the year (Year 2experiment).



Dow

nloaded from


density, contrary to what we observed here. In the earlierstudy, height increased with plant density while dry matter al-location to roots decreased, as reported here. However, in theearlier study, acclimation responses to density were evidentlyinsufficient to compensate for the reduced light availability.As in our experiment, Aphalo and Rikala (2003) found thatconcentrations of N (and of K and P) in stems + roots were un-affected by density.

Both the means and the variances of seedling morphologicalcharacteristics are important determinants of planting stockquality, the goal being to produce plants that are uniformlywell adapted for transplanting to the forest. The means and

variances of seedling morphological traits must also be impor-tant in determining the outcome of interactions among plantsin natural stands and the performance of the stand as a whole.

In each of the Nelder experiments reported here, the vari-ance in height did not increase with density, whereas the vari-ance in plant dry mass did. Thus, it appears that throughphenotypic plasticity in height growth and other traits, seed-lings of low dry mass were able to avoid growth suppressionby larger neighboring plants (see Aphalo et al. 1999). Amongother morphological responses to high plant density was alate-season increase in SSL. Both the mean and the variance ofSSL increased, accompanying the increased variance in wholeplant dry mass. In contrast, the number of leaves was morevariable at low than at high density because no branches wereformed at high plant densities and the number of leaves was,



Table 3. Summary of ANOVA from linear mixed-effects model fit to data from the experiment of Year 2. Abbreviations: SMR = stem mass ratio;RMR = root mass ratio; SSL = specific stem length; stem (root) [Nmass] = nitrogen concentration in stems (roots) expressed per unit dry mass; seed-ling N = nitrogen content per seedling; and stem (root) δ13C = stable carbon isotope composition in stems (roots).

Effect of density on mean Effect on variability

df number df density F P L ratio P

Harvest 2Height 2 211 101.25 < 0.0001 0.042 0.837Diameter 2 211 1.47 0.2322 7.931 0.0049Dry mass 2 211 1.19 0.306 7.507 0.0061SMR 2 211 200.25 < 0.0001 1.197 0.2738RMR 2 211 200.25 < 0.0001 1.197 0.2738SSL 2 211 26.78 < 0.0001 15.256 0.0001Sturdiness index 2 211 106.80 < 0.0001 38.018 < 0.0001Stem [Nmass] 2 26 0.803 0.459 1.537 0.215Root [Nmass] 2 26 0.062 0.940 0.2131 0.644Seedling N 2 26 2.647 0.0899 0.4273 0.513Stem δ13C 2 26 17.76 < 0.0001 0.1450 0.7033Root δ13C 2 26 70.73 < 0.0001 0.9947 0.3186

Figure 10. (a) Seedling nitrogen (N) content; (b) root N:mass ratio; (c)N concentration in stems; and (d) N concentration in roots versusplant density. Measurements were made after leaf fall (Year 2 experi-ment). Nitrogen concentrations are expressed per unit dry mass. Ni-trogen concentration determinations were made on four pooledsamples per density, one sample from each of four racks. Each pooledsample comprised seven seedlings from the same rack (total of28 seedlings per value shown). Error bars indicate ± one standard er-ror.

Figure 11. Carbon isotope ratio (δ 13C) of seedling stems versus plantdensity. Measurements were made after leaf fall (Year 2 experiment).Values are means of four pooled samples, one for each density andNelder rack, from a total of 28 seedlings per density. Error bars indi-cate ± one standard error.



Dow

nloaded from


thus, fixed by the number of nodes on the main stem. However,some lower leaves dropped prematurely at the highest densi-ties.

The observed morphological responses to plant density areassumed to have been primarily in response to light, althoughother density dependent environmental factors may also haveplayed a role. The effect of the response to plant density was toreduce size stratification and inequality in the sharing of re-sources (e.g., light) among individuals, leading to a more uni-form population than would otherwise be expected (Ballaréet al. 1994, Ballaré and Scopel 1997, Aphalo et al. 1999). Be-cause small individuals become relatively more elongated,they capture proportionally more light than if they had had thesame morphology as the dominant individuals. In our experi-ment, a few suppressed individuals were found, but only at thehighest densities.

It is generally accepted that accelerated elongation in re-sponse to increasing plant density is a shade avoidance mecha-nism of competition for light (e.g., Schmitt et al. 1995, Dornet al. 2000), which is limited in scope mainly by the impact ofincreased height on the mechanical stability of the plant. Mod-eling studies indicate that the optimal internode length forlight capture increases with plant density and is also related toleaf lamina length (Sekimura et al. 2000). In our experiment,leaf size increased with density, so based on these optimizationarguments, one could expect that natural selection would haveselected for elongation of internodes.

In the LED irradiation experiment, a rapid and highly signif-icant acceleration of stem height growth in response to FR wasobserved. Responses of silver birch to light quality have beenreported previously (Aphalo and Lehto 1997, 2001); however,in earlier experiments the whole seedling was irradiated,whereas in our study only the growing internode was irradi-ated directly, although some stray light probably reached theadjacent leaves. From measurements during the early stages ofthe Nelder experiment, we know that the R:FR ratio at the irra-diated stem position decreased markedly with density. These

results led us to infer that, at least at early stages of canopy de-velopment, photomorphogenesis played an important role inthe responses of seedlings to plant density. Furthermore, thephotomorphogenic response was likely mediated by changesin the horizontal R:FR photon flux ratio, as has been demon-strated in some herbaceous plants (Ballaré et al. 1987, 1990).

However, non-photomorphogenetic responses to densitycannot be excluded. For example, ethlene accumulation withinthe canopy has been identified as an additional signal ofcrowding sensed by plants (Pierik et al. 2004). Other responsescould include effects of decreased PAR on lower leaves, de-creased wind speed, increased relative humidity and decreasedsoil temperature at higher densities. However, in our experi-ment, these changes in the environment of the seedlings musthave been most marked late in the development of the canopy.Such changes should affect leaf gas exchange, with lowerintercellular CO2 concentration under the higher irradianceand higher evaporative demand conditions of the low densitycanopy. The decrease in δ13C with density is probably a reflec-tion of this effect. However, the decrease in δ13C in plant tis-sues could also have been caused partly by changes in δ13C inthe air within the canopy as a result of the effect of plant den-sity on the contribution of soil and plant respiration to carbondioxide in the canopy air space. A similar effect of irradiancehas been reported for measurements along an irradiance cline(Ehleringer et al. 1986).

To the best of our knowledge, the statistical method used totest for the effect of treatment on variability has not previouslybeen used by plant physiologists. The method has value for an-alyzing dose-response experiments or other continuous exper-iments with covariates. Application of these methods is nowpossible with high-quality free software.

In conclusion, as hypothesized, light quality is a determi-nant of birch seedling responses to plant density during theearly stages of canopy development, but the effects of other en-vironmental variables likely play a role also, especially duringthe later stages of canopy development. Studying density ef-fects on the variance of adaptive plant characteristics in addi-tion to effects on the mean adds a new dimension to ourunderstanding of seedling responses to the environment. Theinter-plant variation in the magnitude of the shade-avoidanceresponse seemed to partly compensate for the inter-plant vari-ation in growth. This plasticity must have important conse-quences for height distribution within a population.

Of practical significance, our study shows that the density ofnursery-grown tree seedlings influences not only mean seed-ling traits, but also seedling uniformity, both of which are im-portant factors in the production of high quality forest plantingstock. From a theoretical perspective, between plant variabil-ity is a key factor relating physiological processes to popula-tion and ecological processes.

Acknowledgments

We thank Raija Kuismin, Ritva Pitkänen and Pekka Voipio for skillfultechnical assistance. We also thank the developers of the R system andthe nlme package for their efforts to make modern statistical methods



Figure 12. Height growth of seedlings irradiated with far-red (�) andred (�) light emitting diodes (LEDs) and controls (�). Values aremeans of 14 seedlings. Error bars indicate ± one standard error. Ab-breviation: d = days.



Dow

nloaded from


widely available at no cost to the user and for their commitment tosupporting users. This research was jointly funded by the Finnish For-est Research Institute, Jenny ja Antti Wihurin rahasto (Grant 00013)and the Finnish Academy (Project 48606).

References

Aphalo, P.J. and C.L. Ballaré. 1995. On the importance of informa-tion-acquiring systems in plant-plant interactions. Funct. Ecol.9:5–14.

Aphalo, P.J. and T. Lehto. 1997. Effects of light quality on growth andN accumulation in birch seedlings. Tree Physiol. 17:125–132.

Aphalo, P.J. and T. Lehto. 2001. Effect of lateral far-red light supple-mentation on the growth and morphology of birch seedlings and itsinteraction with mineral nutrition. Trees 15:297–303.

Aphalo, P.J. and R. Rikala. 2003. Apparent quality and actual fieldperformance of silver-birch seedlings grown in containers of differ-ent volume and at different spacings. New For. 25:93–108.

Aphalo, P.J., C.L. Ballaré and A.L. Scopel. 1999. Plant-plant signal-ing, the shade avoidance response and competition. J. Exp. Bot.50:1629–1634.

Ballaré, C.L. 1999. Keeping up with the neighbours: phytochromesensing and other signaling mechanisms. Trends Plant Sci. 4:97–102.

Ballaré, C.L. and A.L. Scopel. 1997. Phytochrome signalling in plantcanopies: testing its population-level implications with photo-receptor mutants of Arabidopsis. Funct. Ecol. 11:441–450.

Ballaré, C.L., A.L. Scopel and R.A. Sánchez. 1990. Far-red radiationreflected from adjacent leaves: an early signal of competition inplant canopies. Science 247:329–332.

Ballaré, C.L., A.L. Scopel, E.T. Jordan and R.D. Vierstra. 1994. Sig-nalling among neighboring plants and the development of size in-equalities in plant populations. Proc. Natl. Acad. Sci. USA 91:10094–10098.

Ballaré, C.L., R.A. Sánchez, A.L. Scopel, J.J. Casal and C.M. Ghersa.1987. Early detection of neighbour plants by phytochrome percep-tion of spectral changes in reflected sunlight. Plant Cell Environ.10:551–557.

Callaway, R.M. and L.R. Walker. 1997. Competition and facilitation:A synthetic approach to interactions in plant communities. Ecology78:1958–1965.

Casal, J.J. and H. Smith. 1989. The function, action and adaptive sig-nificance of phytochrome in light-grown plants. Plant Cell Envi-ron. 12:855–862.

Doran, K., R.W. Ruess, F.G. Plumley and T.L. Wurtz. 2001. Photo-synthetic responses of white spruce saplings (Picea glauca) to con-trolled density gradients of spruce and green alder (Alnus crispa).Ecoscience 8:76–88.

Dorn, L.A., E.H. Pyle and J. Schmitt. 2000. Plasticity to light cues andresources in Arabidopsis thaliana: testing for adaptive value andcosts. Evolution 54:1982–1994.

Ehleringer, J.R., C.B. Field, Z.F. Lin and C.Y. Kuo. 1986. Leaf carbonisotope and mineral composition in subtropical plants along anirradiance cline. Oecologia 70:520–526.

Gilbert, I.R., P.G. Jarvis and H. Smith. 2001. Proximity signal andshade avoidance differences between early and late successionaltrees. Nature 411:792–795.

Ihaka, R. and R. Gentleman. 1996. R: A language for data analysisand graphics. J. Comput. Graph. Stat. 5:299–314.

Lee, D.W., K. Baskaran, M. Mansor, H. Mohamad and S.K. Yap.1996. Irradiance and spectral quality affect Asian tropical rain for-est tree seedling development. Ecology 77:568–580.

Morgan, D.C. and H. Smith. 1979. A systematic relationship betweenphytochrome-controlled development and species habitat, forplants grown in simulated natural radiation. Planta 145:253–258.

Nelder, J.A. 1962. New kinds of systematic designs for spacing exper-iments. Biometrics 18:283–307.

Pierik, R., G.C. Whitelam, L.A. Voesenck, H. de Kroon and E.J.W.Visser. 2004. Canopy studies on ethylene-insensitive tobacco iden-tify ethylene as a novel element in blue light and plant–plant sig-naling. Plant J. 38:310–319.

Pinheiro, J.C. and D.M. Bates. 2000. Mixed-effects models in S andS-Plus. Springer, NY, 528 p.

Ritchie, G.A. 1997. Evidence for red:far-red signaling and photo-morphogenic growth response in Douglas-fir (Pseudotsuga men-ziesii) seedlings. Tree Physiol. 17:161–168.

Schmitt, J., A.C. McCormac and H. Smith. 1995. A test of the adap-tive plasticity hypothesis using transgenic and mutant plants dis-abled in phytochrome-mediated elongation responses to neighbors.Am. Nat. 146:937–953.

Sekimura, T., T. Roose, B. Li, P.K. Maini, J.i. Suzuki and T. Hara.2000. The effect of population density on shoot morphology ofherbs in relation to light capture by leaves. Ecol. Model. 128:51–62.

Smith, H. 1981. Light quality as an ecological factor. In Plants andtheir Atmospheric Environment, 21st Symposium of the BritishEcological Society. Eds. J. Grace, E.D. Ford and P.G. Jarvis.Blackwell Scientific Publications, Oxford, pp 93–110.

Smith, H. 1994. Sensing the light environment: the functions of thephytochrome family. In Photomorphogenesis in Plants. Eds.R.E. Kendrick and G.H.M. Kronenberg. 2nd Edn. Kluwer Aca-demic, Dordrecht, pp 377–416.

Smith, H. and G.C. Whitelam. 1997. The shade avoidance syndrome -multiple responses mediated by multiple phytochromes. Plant CellEnviron. 20:840–844.

Appendix

Statistical models

This appendix provides a summary of the statistical modelswe used based on the extensive discussion in Pinheiroand Bates (2000). For the analysis of the growth curves in Ta-ble 1, we used a nonlinear mixed-effects model with a simplelogistic function with three parameters. The model was fit bythe method of maximum likelihood. The formulation used forthe logistic is:

y x= +φ1 2 31{ ( ) )}exp(– – φ φ (1)

where y is height, leaf area, or dry mass, x is time and the threeparameters are, φ1 is the horizontal asymptote, φ2 (xmid) is thetime at y = φ1/2 and φ3 is a scale parameter on the x axis givingthe steepness of the curve.

The model is then:

y f vij ij ij ij= +( ),φ ε (2)

with i = 1,...,M the groups and j = 1,..., n i the observations ineach of the i groups. The logistic function is f , a covariate vec-tor, in our case time, is v ij and ε ij is a normally distributedwithin-group error term. However, we do not assume homo-





Dow

nloaded from


scedastic within-group errors: εi ~ N(0,σ 2Λ i).The parameter vector φ ij is modeled as:

φ βij ij ij iA B= + b (3)

with b i ~ N(0,Ψ); where β is a p-dimensional vector of fixedeffects and b i is a q-dimensional random effects vector associ-ated with the ith group with variance-covariance matrix Ψ.

In our case the fixed effects were initially given as a sec-ond-degree polynomial on the covariate row number in theNelder pattern (proportional to logarithm of growing density).In addition the origin of these row numbers was shifted to thecenter of the table. In nlme:

fixed Asym xmid scal I(row 6.5) I((row 6.5) 2)~ ^= + + − + −

During model building some of the (insignificant) second-degree terms were removed.

The random effects were given as an intercept, with seedlingor table as the grouping factor depending on whether the datawas sequentially measured on the same seedlings or not.

In nlme, for the first case:

random = ~1|seedling

The heterogeneity of variance was modeled, in this case, us-ing the expected values as a variance covariate in a power ofvariance function. In nlme:

weight = varPower()

The variance model being:

Var ( ij ijε σδ

) = 2 2v (4)

where v ij is the variance covariate and δ a parameter.

For the analyses in Tables 2 and 3, from observations for asingle occasion rather than growth curves, we fitted by themethod of restricted maximum likelihood second degree poly-nomial regressions against the logarithm of growing density,initially including a random effect for the intercept, taking intoconsideration the grouping by rack. These initial linear mixedeffects models were of the form (in matrix notation):

y bi i i i iX Z= + +β ε (5)

where subscript i indicate the group, b i ~ N(0,Ψ), ε i ~N(0,σ2I) and b i is a random-effects vector describing the shiftin intercept with group, in our case rack. The matrix Ψ in thiscase is equal to σ b

2 ,,the random effects variance.Because the random effects were very small, and in practice

not estimable with precision, in the analyses presented weused generalized least squares models (sensu Pinheiroand Bates 2000), lacking random effects, but including a flexi-ble model for the variance. The usual regression model being:

y x xij ij ij ij= + + +β β β ε0 1 22 (6)

where y ij is the observed response and x ij is the logarithm ofgrowing density.

To test for the change in variability of the response variablesas a function of density, we fitted in each case a model withfixed variance (the default in lme and gls), where ε ij ~ N(0,σ 2),as shown above, and a model with variance modeled as a func-tion of a covariate logarithm of growing density. Each pair offits was compared to test the significance of the effect of den-sity on the variability. The only difference with the variancemodel used for the growth curves is that a different covariatewas used (see Equation 4). In gls, for example, for diameter:

model = diam ~ log(dens) + I(log(dens)^2)

weight = varPower(form = ~ log(dens)).





Dow

nloaded from


Spacing of silver birch seedlings grown in containers of equal ...

Documents

Transcript of Spacing of silver birch seedlings grown in containers of equal ...