Top Banner

of 15

Allometric eqn- Leucaena

Jun 04, 2018

Download

Documents

aneeshcof
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
  • 8/13/2019 Allometric eqn- Leucaena

    1/15

    Allometric models for predicting above- and belowground

    biomass ofLeucaena

    -KX2 in a shaded coffee agroecosystemin Hawaii

    Adel H. Youkhana Travis W. Idol

    Received: 15 October 2010 / Accepted: 23 April 2011 / Published online: 4 May 2011

    Springer Science+Business Media B.V. 2011

    Abstract We developed site-specific allometric

    models for Leucaena leucocephala 9 pallida var.

    KX2 trees in a shaded coffee agroecosystem in

    Hawaii to predict above- and belowground biomass

    and the regrowth potential of pollarded trees. Models

    were used to compare tree growth rates in an

    experimental agroforestry system with different pol-

    larding frequencies and additions of tree pruning

    residues as mulch. For all allometric equations, a

    simple power model (Y = aXb) provided the optimal

    prediction of biomass or regrowth after pollard-ing. For aboveground biomass components (stem,

    branches, leaves, and seed and pods), stem diameter

    alone was the best predictor variable. Stump diameter

    provided the best prediction of coarse root biomass

    and aboveground regrowth after pollarding. Predic-

    tions of biomass from generalized allometric models

    often fell outside the 95% confidence intervals of our

    site-specific models, especially as biomass increased.

    The combination of pollarding trees once per year

    plus the addition of tree mulch resulted in the greatest

    aboveground regrowth rates as well as accumulation

    of biomass and C in the stump plus coarse roots.

    Although optimal prediction required the develop-

    ment of site-specific allometric relationships, a sim-

    ple power model using stem or stump diameter alone

    can provide an accurate assessment of above- and

    belowground tree biomass, as well as regrowth

    potential under specific management scenarios.

    Keywords Allometric models Leucaenaleucocephala-KX2 Nonlinear regression

    Aboveground and belowground biomass Coarse root excavation Carbon sequestration

    Introduction

    Biomass equations form a basis for estimating

    biomass and carbon (C) pools in forestry and

    agroforestry systems (Albrecht and Kandji 2003;

    Kenzo et al. 2009; Nair et al. 2009). Allometricmodels are based on correlations between morpho-

    logical characters that can be measured in the field or

    the laboratorysuch as diameter at breast height

    (DBH) or tree height (H)and more complex size

    measurementssuch as stem volume or tree bio-

    massthat cannot be measured non-destructively. In

    forestry, these models allow for the prediction of

    ecologically or economically important pools, such as

    biomass, C, nutrients, or merchantable wood yield

    A. H. Youkhana (&) T. W. IdolDepartment of Natural Resources and Environmental

    Management, College of Tropical Agriculture and Human

    Resources, University of Hawaii at Manoa,

    1910 East-West Road, Sherman Lab # 101, Honolulu,

    HI 96822, USA

    e-mail: [email protected]

    T. W. Idol

    e-mail: [email protected]

    1 3

    Agroforest Syst (2011) 83:331345

    DOI 10.1007/s10457-011-9403-6

  • 8/13/2019 Allometric eqn- Leucaena

    2/15

    (Ares et al. 2002; Northup et al. 2005). Allometric

    models have primarily been developed for applica-

    tion in natural forests and forest plantations (Zianis

    and Mencuccini 2004). In addition, several general-

    ized allometric equations have been developed for

    estimating biomass of tropical species (Chambers

    et al. 2001; Chave et al. 2005; Overman et al. 1994;Santos-Martin et al. 2010). General models repre-

    senting multiple species, land uses, and locations can

    eliminate the investment needed to develop site- and

    species-specific models (Lambert et al. 2005). Most

    of these models use DBH, height, and sometimes

    wood density as predictor variables. Litton and

    Kauffman (2008) found that DBH alone was an

    effective predictor of all categories of aboveground

    biomass for Metrosideros polymorpha allometry in

    Hawaii. Navar (2009) developed an individual allo-

    metric equation using non-linear regression to fitparameters of the typical power equation for estimat-

    ing biomass of 19 species in Mexico. Chave et al.

    (2005) developed generalized equations for tropical

    tree species in moist, wet, and dry forests using DBH,

    height, and wood density. Sampaio and Silva (2005)

    used basal diameter as well as DBH for shrubby

    species in caatinga ecosystems of Brazil.

    The advantage of generalized equations lies in

    their broad applicability and the elimination of

    destructive harvests for each species, variety, tree

    size range, and/or location. However, using theseequations at a specific site can result in biased

    estimates of biomass and C (Cairns et al. 2003). The

    applicability of generalized models to new sites or

    species, therefore, must be tested prior to application.

    In particular, using generalized equations to estimate

    biomass in agroforestry systems may be problematic

    due to potential alterations in tree architecture and

    biomass caused by the pruning and pollarding that are

    typically used to manage shade levels and provide

    farm products (Segura et al.2006). Pruning obviously

    reduces total aboveground biomass and thus compli-cates assessments of biomass or C sequestration.

    In addition, with pollarding or pruning there is a need

    for models that predict regrowth rates of stems,

    branches, leaves, and reproductive components, since

    these separately provide important products and

    services and interact with other components of the

    system. Thus, agroforestry systems often require

    development of models specific at least to the

    management regime if not to the species or variety.

    Allometric models to predict the biomass of coarse

    roots are also needed because they contribute signif-

    icantly to total individual tree and stand biomass and

    C storage (Li et al. 2003; Santantonio et al. 1977).

    Direct assessment of belowground coarse root bio-

    mass is even more difficult than aboveground

    biomass (Huxley 1999). There is large spatial heter-ogeneity in coarse root distribution and a limited

    capacity to quantify the spatial distribution and

    area- or volume-weighted biomass of roots given

    current methods (Bengough et al. 2000). As a result,

    modeling tree root biomass allometrically has gained

    wide acceptance (Drexhage and Colin 2001). Scien-

    tists have made numerous attempts to establish a

    relationship between above- and belowground tree

    attributes at the stand and individual tree level,

    mostly using models of the relative allocation of dry

    weight between roots and aboveground parts (Bolteet al. 2004; Kenzo et al. 2009; Santantonio et al.

    1977). Similar to models of aboveground biomass,

    these studies have shown that root and main stem

    diameter are good predictors of individual root and

    total coarse root biomass, respectively (Gower et al.

    1996; Kurz et al. 1996; Li et al. 2003; Nadelhoffer

    et al.1985). For individual trees, several studies have

    developed relationships between coarse root biomass

    and stem DBH, height (H), or both (Drexhage and

    Colin 2001; Hoffmann and Usoltsev 2001; Laiho

    and Finer 1996; Santantonio et al. 1977; Thies andCunningham 1996). Haynes and Gower (1995) used

    basal stump diameter alone to predict coarse root

    biomass. This is especially appropriate for agrofor-

    estry systems in which trees are pollarded to manage

    shade levels.

    For pollarded trees in agroforestry systems,

    regrowth potential is an important process to be able

    to model and predict on an individual-tree basis and

    at the stand level. Allometric equations have not been

    developed specifically for this purpose, but it seems

    logical that measures such as main stem diameterwould be appropriate for these predictions. This

    would also allow for non-destructive assessments of

    the influence of management practices such as mulch

    addition, fertilization, tillage, etc. on this potential,

    especially over repeated pollarding or pruning cycles.

    The objectives of the present study were to (1)

    develop allometric equations from destructive harvest

    to predict live biomass of both aboveground and

    belowground components of Leucaena-KX2 trees;

    332 Agroforest Syst (2011) 83:331345

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    3/15

    (2) evaluate the potential of generalized allometric

    models to predict the biomass of these trees; and (3)

    use the best-fit equations to compare the regrowth

    potential and biomass accumulation of trees under

    different pollarding frequencies with and without

    return of the pruning residues as mulch.

    Materials and methods

    Study site

    The study was carried out at the University of

    Hawaii, College of Tropical Agriculture and Human

    Resources, Waimanalo Research Station on the

    windward side of the island of Oahu (21 200 N,

    158 200 W). The site is 20 m above sea level and is

    classified as a humid tropical environment (Giam-belluca et al.1986). The soils are classified mainly as

    Vertic Haplustolls, dominated by the Waimanalo

    series. Further details of the study site can be found in

    Youkhana and Idol (2009).

    Leucaena variety KX2 was chosen for this study

    due to the rapid growth of resprouts after pollarding

    or coppicing, its high N fixation rate, resistance to

    psyllid insects and self-sterility (Brewbaker 2008).

    Plots were established in a seed orchard established in

    2002 with trees planted in 8 rows of 20 trees per row,

    planted on 2 9 2-m spacing. Eighteen plots wereestablished, each 4 9 6 m in size, encompassing six

    leucaena trees in a 3 9 2 tree arrangement. The inner

    6 rows and 18 trees per row were included in the

    plots, leaving a border row of trees around the entire

    orchard (Fig.1). Drip irrigation was applied to all

    plots during dry periods to maintain plant survival

    and growth.

    On 20 September 2005, all trees in the plots were

    pollarded at 1 m above ground level. This was chosen

    because the trees had previously been pollarded at

    approximately this height. The harvested biomass

    was chipped in a mechanical tree chipper and

    distributed uniformly back to the leucaena plots as

    mulch. As shoots regrew after pollarding, the two

    largest shoots initiated near the top of the stump and

    on opposite sides from each other were retained;others were removed. Again, this was based on

    the previous management regime. The boundaries

    between plots were trenched to a depth of 1 m and

    lined with plastic sheeting to prevent overlap of root

    systems between adjacent plots. On 20 August 2006,

    trees were pollarded again and assigned to one of

    two pollarding frequency treatments: once every

    6 months and once every 12 months. A second

    treatment of mulch addition was crossed with

    pollarding frequency. Coffee seedlings were planted

    within the plots on 2 9 2-m spacing to establish acoffee agroforestry system in order to investigate the

    effects of shade level and mulch additions on bean

    productivity (yield) and quality. For mulch-addition

    plots, tree pollarding residues were mechanically

    chipped and added back to the plot of origin. For the

    no-mulch plots, pollarding residues were not returned

    to the plots. The treatments were imposed for 3 years

    until September 2009.

    Measurements of aboveground biomass

    and carbon

    For the development of allometric models of above-

    ground tree biomass, 20 representative trees that

    spanned a range of stump diameters (measured at

    50 cm above the ground) were selected at the time of

    initial pollarding in 2005 (Table 1). Stump diameter

    (D) and basal diameter of the shoot emanating from

    the stump were measured using a diameter tape. The

    height (H) of each individual shoot was measured

    Pt1 2 3 4

    7 8 9 10 11 12

    5 6

    13 14 15 16 17 182 m 2 m

    4 m

    Coffea arabica var. Kona TypicaLeucaena leucocephalaKX2

    6 m

    Fig. 1 Layout of experimental plots in shade system ofLeucaena-KX2 and coffee

    Agroforest Syst (2011) 83:331345 333

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    4/15

    using a Suunto Clinometer (Philip 1994). The trees

    were pollarded and the biomass collected. Each

    tree was divided into four biomass components:

    orthotropic shoots emanating from the stump (stems),

    plagiotropic shoots attached to the stems (branches),

    leaves, and reproductive parts (seed and pods). Each

    fraction was weighed, and representative sampleswere collected for dry matter determination (60C,

    72 h until constant weight to the nearest 0.01 g).

    Wood specific gravity (oven-dry weight over green

    volume) was determined for 50 stem cross-section

    samples using the water displacement method

    (ASTM 1983); the average (0.94 g cm-3) was used

    for prediction of biomass.

    Biomass data for each individual component were

    analyzed to derive predictive equations based on stem

    D or H using a set of standard allometric models

    (Husch et al. 2003) (Table2). These included thefollowing:

    Y a bD2 1

    Y a bDcD2 2

    Y aDbHc 3

    Y aDb 4

    where Y is the dependent variable (e.g. biomass of

    stems, branches, leaves, reproductive parts or total

    shoot biomass in kg dry mass) and a, b and c areconstants. Stem D and H were log-transformed to

    create linear models analogous to Eqs.3 and 4. All

    regression equations were developed using the PROC

    REG procedure in SAS (SAS Institute Inc.1990).

    Models were selected according to goodness of fit

    measures, including the sum of squares of the

    residuals, the residual mean square (RMS), the

    coefficient of determination (R2), the Akaike Infor-

    mation Criterion (AIC), the bias-corrected AIC(AICc), and the Bayesian Information Criterion

    (BIC) (Myung et al. 2010; Spiess and Neumeyer

    2010). In addition, the residuals (observed minus

    predicted values) were plotted against the predictor

    variables to visually assess for bias in predictions.

    The coefficient of determination for each model was

    calculated as R2 = (1 - SSR)/corrected SST (Litton

    and Kauffman 2008), where SSR is the sum of

    squares of the residuals, and corrected SST is the total

    sum of squares of deviations from the overall mean.

    Models with higher R2; least bias for biomassprediction; and smallest SSR, RMSE, AIC, AICc

    and BIC were selected for each biomass component.

    Several generalized equations for predicting

    above- and belowground biomass were evaluated

    for their appropriateness for the managed Leucaena-

    KX2 trees in this study. For shoot biomass, we used

    equations from Brown (1997), Chave et al. (2005)

    and Sampaio and Silva (2005), and previous equa-

    tions developed for Leucaena-KX2 (Brewbaker

    2008) (Table3). The Brown (1997) model requires

    only stem D (cm) to predict total shoot biomass (kgdry weight); whereas, the Chave et al. (2005) models

    require species-specific information on wood specific

    gravity and either D alone or D and H specific to

    climatic zones. The two models of Sampaio and Silva

    (2005) require D alone or D and H combined to

    predict total shoot biomass. We also used two

    allometric equations developed for total shoot bio-

    mass and forage annual dry matter yields ofLeuca-

    ena-KX2 (Brewbaker2008). These models used tree

    height (cm) alone as the predictor variable (Table 3).

    The previous models used DBH to predict totalaboveground biomass. However because our trees

    were pollarded, our models predicted total biomass of

    the resprouting shoots based on basal stem D; this

    may affect the model coefficients but should not

    affect the goodness of fit, as resprouting stems

    generally maintain the same architecture and thus

    allometric relationships with biomass or volume as

    the main stem (Fownes and Harrington1992; Tewari

    et al. 2004).

    Table 1 Diameter range of representative shoots and stumps

    of Leucaena-KX2 trees for aboveground and belowground

    study

    Aboveground study Belowground study

    Shoot D (cm) Stump D (cm) Stump D (cm)

    Range # Trees Range # Trees Range # Trees

    1.53 5 35 5 57 4

    35 5 57 5 79 3

    57 5 79 4 911 3

    79 3 911 1 1113 3

    911 2 1113 1 1315 2

    1315 2 1517 3

    1517 2 1720 2

    Total 20 20 20

    334 Agroforest Syst (2011) 83:331345

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    5/15

    Generalized models were evaluated by plotting thepredicted biomass and 95% confidence intervals from

    our site-specific model with the predicted biomass

    from each generalized model across the full range of

    D and H of Leucaena-KX2 trees measured in this

    study. Site-specific and generalized equations were

    used to predict biomass from the 2006 pollarding

    events to maintain independence from the data used

    to develop the site-specific models. If the predicted

    biomass of the generalized model fell outside of the

    best-fit confidence intervals, especially at larger D orH, then it was rejected as being appropriate for

    application to the trees in this study.

    Stump and coarse root biomass

    We used the methods of Kapeluck and Van Lear

    (1995) to predict the biomass of coarse roots ([5 mm

    diameter). Essentially, this is a two-step process in

    which a regression relationship is developed first

    Table 2 Goodness of fit comparisons of some standard allometric models explored in this study

    Model a b R2

    SSR RMS AIC AICc BIC

    For shoot D

    Stem biomass (kg)

    Y = a ? bD2

    0.19 2.03 0.91 15.22 1.04 145.53 143.73 146.52

    Y = a ? bD ? cD2

    0.03 2.10 0.97 12.65 0.72 154.89 153.09 155.89

    Y = aDb

    Hc

    0.05 2.04 0.85 22.75 2.25 152.58 149.98 154.57

    Y = aDb

    0.16 2.28 0.98 11.84 0.68 138.74 136.95 139.74

    Branches biomass (kg)

    Y = a ? bD2 0.33 1.98 0.77 32.55 2.02 151.19 149.37 152.18

    Y = a ? bD ? cD2 0.05 1.71 0.75 21.11 1.17 152.31 150.52 153.31

    Y = aDb

    Hc

    0.06 1.57 0.72 25.24 1.95 150.75 148.15 152.74

    Y = aDb

    0.12 1.96 0.78 19.55 1.14 141.38 139.58 142.37

    Leaves biomass (kg)

    Y = a ? bD2

    0.65 1.80 0.70 3.81 0.22 125.78 123.98 126.78

    Y = a ? bD ? cD2

    0.47 1.58 0.71 3.97 0.24 126.25 124.45 127.24

    Y = aDbHc 0.32 1.48 0.66 4.33 0.56 122.17 119.57 124.16

    Y = aDb

    0.03 2.10 0.73 3.66 0.19 115.39 113.59 116.38

    Reproductive parts biomass (kg)

    Y = a ? bD2

    0.44 1.42 0.49 11.58 7.05 139.98 138.17 140.96

    Y = a ? bD ? cD2

    0.60 1.55 0.51 14.65 7.33 148.75 146.95 149.74

    Y = aDb

    Hc

    0.24 1.62 0.55 17.24 8.33 141.08 139.26 142.06

    Y = aDb

    0.09 1.69 0.57 10.13 0.56 134.01 132.21 135.01

    Total biomass (kg)

    Y = a ? bD2

    0.18 2.03 0.90 32.5 1.83 159.19 157.38 160.18

    Y = a ? bD ? cD2 0.38 1.92 0.97 30.97 1.74 165.12 163.32 166.12

    Y=

    aD

    b

    H

    c

    0.34 1.94 0.83 42.24 2.37 162.06 160.26 163.06Y = aD

    b0.30 2.25 0.98 29.53 1.72 156.35 154.55 157.34

    For stump D

    Stump biomass (kg)

    Y = a ? bD2

    0.64 1.63 0.89 178.25 9.44 167.33 165.23 168.32

    Y = a ? bD ? cD2

    0.25 1.84 0.92 98.22 11.27 172.45 170.35 173.44

    Y = aDb

    0.09 1.96 0.93 22.17 1.23 160.46 158.25 161.45

    a, b and c Constants; SSR sum of squares of the residuals; RMSresiduals mean square; R2

    the coefficient of determination; AICthe

    Akaike Information Criterion; AICc the bias-corrected (AIC); BIC the Bayesian Information Criterion

    Agroforest Syst (2011) 83:331345 335

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    6/15

    between individual root diameter and root mass and

    then between stump D and predicted root mass.

    The simple power model (Eq.4) is used for both

    relationships. For this study, five Leucaena-KX2

    trees from the border rows were randomly selected to

    be excavated for measurement of total coarse root

    mass on 6 December 2007. Before excavating the

    root system, the selected trees were marked and

    measurements were made of stump D, shoot D, and

    tree H (Table1). Roots were excavated using handtools and a bulldozer with an excavator bucket to lift

    the stump. Coarse roots were excavated down to a

    diameter of 5 mm (Kapeluck and Van Lear 1995).

    After excavation, the roots and stump were washed

    with water and a stiff brush to remove soil. The

    diameter of all coarse roots was measured using a

    diameter tape or calipers where they were attached at

    the stump. After measurement, roots were cut from

    the stump and oven-dried for at least 72 h at 70C to

    a constant weight. Proximal diameters of 126 lateral

    roots from the five excavated trees were regressedagainst their respective biomass weights to develop

    the regression relationship.

    An additional 20 trees from the border rows were

    selected that spanned a range of stump D from 5 to

    20 cm. Tree dimensions were measured as before prior

    to excavation. The coarse root system was excavated

    using a bulldozer but without the requirement of

    collecting roots down to a set diameter. After washing,

    the diameter of all coarse roots at the stump was

    measured. Roots were cut from the stump, and the

    stumps were collected and dried to measure stump

    biomass.

    Stump D at 50 cm above the ground was also used

    to predict stump biomass using Eqs.1, 2, and 4,

    where Y is stump biomass in kg dry mass and D is

    stump D. For our purposes, the stump was consideredto include the aboveground shoot below the pollard-

    ing height and the belowground portion to which all

    lateral roots were attached. The best-fit model was

    chosen based on goodness of fit measures.

    Coarse root and stump samples from various

    diameter ranges were analyzed for C content using

    a combustion furnace elemental analyzer by the

    University of Hawaii Agricultural Diagnostic Service

    Center (ADSC) in order to compare C sequestration

    rates among the treatments.

    Shoot regrowth

    Stump D was used to predict annual shoot regrowth,

    i.e. the regrowth of both shoots retained on the stump

    after pollarding, using Eq.4. Separate equations were

    developed for each treatment combination and during

    each year of the study (20062009). For the 6-month

    pollarding treatment, stump D was used to predict

    annual growth by combining biomass growth over

    two successive pollarding intervals.

    Statistical analysis

    Because the influence of the pollarding and mulch

    treatments were expected to increase over time, we

    analyzed these effects on the allometric relationship

    between stump D and shoot regrowth via repeated

    measures multivariate analysis of variance (MA-

    NOVA), using the GLM procedure in SAS (Max-

    well and Delaney 2004; von Ende 1993). Best-fit

    regression equations were created within each plot

    (n = 6 trees/plot) during each year to predict shootbiomass regrowth using stump D. The coefficients

    (a and b) for each plot within a treatment were

    averaged to generate treatment means and vari-

    ances for comparison (n = 4). Where there were

    significant interactions among treatments or with

    time, specific contrasts were constructed to ana-

    lyze differences. Bonferroni adjustments were made

    to the critical P-value to account for multiple

    comparisons.

    Table 3 Above-and-belowground generalized models used in

    this study

    Equation Source

    Aboveground biomass

    Y = 5.37 9 10-59 H

    2.714Brewbaker (2008)

    Y = 1.66 9 10-4

    9 H2.365

    Brewbaker (2008)Y = exp (-2.134 ?

    2.530 9 ln(D))

    Brown (1997)

    Y = 0.0612 9 (D 9 H)1.5811 Sampaio and Silva (2005)

    Y = 0.173 9 D2.295

    Sampaio and Silva (2005)

    Y = 0.0509 9 p D2

    H Chave-moist (2005)

    Y = 0.0776 9 (p*D2

    H)0.94

    Chave-wet (2005)

    Coarse root biomass

    Y = 0.06 9 D2.00

    Kapeluck and Van Lear

    (1995)

    For aboveground biomass, D stem diameter; for belowground

    biomass,D stump diameter

    336 Agroforest Syst (2011) 83:331345

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    7/15

    The effects of treatment and time on shoot

    regrowth and biomass of the stump and coarse roots

    at the plot level (Mg ha-1) were analyzed using

    functional data analysis procedures (Ramsay and

    Silverman 2005). Shoot regrowth and stump ?

    coarse root biomass and C content for each treatment

    combination were regressed against time to develop alinear relationship (n = 4 plots per treatment). Sig-

    nificant regression equations (non-zero slopes) indi-

    cated a significant time effect. Significantly different

    slopes among treatments indicated a significant

    treatment effect. To compare analytical procedures,

    the 2009 data on shoot regrowth and stump ? coarse

    root biomass and C content were compared among

    treatment combinations via one-way ANOVA with

    Tukeys honest significant difference as the means

    separation test.

    Results

    Biomass prediction

    The simple power model using only stem D as a

    single independent variable was found to be the

    best predictor of total aboveground biomass of

    individual shoots (Table2). Stem and total shoot

    biomass were both well-predicted (R2 = 0.98),

    while branches, leaves, and reproductive part (seedpods) were less well-predicted (Fig.2ae). Error

    variance increased for larger-diameter stems but

    showed no systematic bias (Fig.2fj). Logarithmic

    transformation of the data did not reduce this

    increase in error variance with stem D or improve

    the fit for most components.

    There was a strong relationship between shoot H

    and D (Fig.3), but shoot H alone was not as good a

    predictor of total shoot biomass (R2 = 0.79) (Fig. 4).

    Coarse root basal diameter was an effective

    predictor of individual coarse root biomass. As withother biomass components, larger diameter roots

    exhibited greater error variance, but there was no

    evidence of bias. Logarithmic transformation did

    reduce this increase in error variance and so was

    selected for the final model (Fig. 5).

    Stump D was an effective predictor of stump

    biomass (Fig.6a) and total coarse root biomass

    (Fig.6b). As with stem D, stumps with larger D

    showed greater error variance (Fig. 6c, d).

    Evaluation of generalized models

    All of the generalized models evaluated were highly

    significant (P\ 0.01), and several had R2[ 0.90

    (Table4). None of them, however, provided a better

    fit than the site-specific models. In addition, biomass

    predicted from all of the generalized models tendedto fall outside of the confidence intervals from the

    best-fit models, especially at larger D or H (Fig. 7).

    The two generalized tropical tree models by Chave

    et al. (2005) underestimated aboveground biomass

    relative to the site-specific model at all tree diameters

    (Fig.7c) but did not show an increasing deviation as

    stem D increased.

    The previously published models for Leucaena-

    KX2 (Brewbaker 2008) also showed divergence

    from our site-specific model using H as the

    predictor variable (Fig.8). Model (1) provided abetter fit but underestimated biomass at moderate

    heights (47 m) when applied to the data in our

    study.

    Shoot regrowth

    There were no significant time or treatment effects

    of pollarding or mulching on the prediction of

    individual shoot biomass (kg stem-1) based on stem

    D. On the other hand, there was a significant declinefrom 2006 to 2009 in the exponential parameter for

    stump D as a predictor variable for shoot regrowth

    in the no-mulch treatments (Table5), indicating a

    decline in shoot regrowth potential. There were no

    differences in the relationship due to pollarding

    frequency.

    Total shoot regrowth at the plot level (Mg ha-1

    year-1) increased significantly over time for the

    treatment combination pollarding once-mulch addi-

    tion. For the other treatment combinations, shoot

    regrowth declined significantly over time in the orderpollarding once-no mulch, pollarding twice-mulch,

    pollarding twice-no mulch (Fig.9; Table6). The

    one-way ANOVA comparing treatments during 2009

    showed the same significant differences (data not

    shown).

    Stump plus coarse root biomass and C sequestra-

    tion on a plot-level basis increased significantly over

    time for all treatments (Table6). The pollard once-

    mulch addition treatment was significantly greater

    Agroforest Syst (2011) 83:331345 337

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    8/15

    than the pollard twice-no mulch treatment, adding

    approx. 2.4 Mg ha-1 of biomass (1.1 Mg ha-1 of C)

    from 2006 to 2009 (Fig. 10). The one-way ANOVA

    comparing treatments during 2009 showed the same

    significant difference (data not shown).

    Discussion

    In general, stem and stump diameter were good

    predictors of biomass and regrowth after pollarding

    forLeucaena-KX2 using a simple power relationship.

    A

    y = 0.16 D2.28

    R2

    = 0.98

    S

    temb

    iomass(kg)

    0

    5

    10

    15

    20

    25F

    Stemb

    iomassresiduals

    -4

    -2

    0

    2

    4

    B

    y = 0.12 D1.96

    R2

    = 0.78

    Branchesbiomass(kg)

    0

    2

    4

    6

    8

    G

    Branchesbiomassresiduals

    -4

    -2

    0

    2

    4

    C

    y = 0.03 D 2.10

    R2= 0.73

    Leafbiomass(kg)

    0

    1

    2

    3

    4

    H

    Leafbiomassresidu

    als

    -4

    -2

    0

    2

    4

    D

    y = 0.09 D1.69

    R2= 0.57

    Reproduc

    tivebiomass(kg)

    0

    2

    4

    6

    I

    Reproductiveresiduals

    -4

    -2

    0

    2

    4

    E

    y = 0.30 D2.25

    R2

    = 0.98

    Shoot D (cm)

    Totalshootbiomass(kg)

    0

    10

    20

    30

    J

    0 2 4 6 8 0 2 4 6 8

    Totalshootbiomassresiduals

    -4

    -2

    0

    2

    4

    Fig. 2 Allometric models

    for predicting: a stem,

    b branches, c leaves,

    d reproductive component

    (pods and seeds), ande total

    shoot biomass (kg) from

    shoot diameter (cm) in

    individuals ofLeucaena

    -KX2, and fj biomass

    residuals (observed minus

    predicted values). P \ 0.01

    for all models

    338 Agroforest Syst (2011) 83:331345

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    9/15

    For individual stems, basal stem D was quite accuratein predicting biomass of the main stem, but less so for

    branches, leaves, or seed pods. The main stem

    represents the major proportion of aboveground

    biomass (Kaonga and Bayliss-Smith 2009; Santos-

    Martin et al.2010; Senelwa and Sims1998; Van et al.

    2000), so stem D was sufficient for predicting total

    shoot biomass. Repeated pollarding and pollarding

    frequency did not alter the allometric relationship

    between stem D and stem biomass. Other studies

    have found similar results (Fownes and Harrington

    1992; Tewari et al. 2004), which suggests that

    pollarding does not alter the morphology, and thus

    allometry, of resprouting stems.

    Many authors have confirmed that using stem D

    alone (DBH or basal stem D) is a reliable predictorof total aboveground biomass for a variety of

    species and ecosystems (Bartelink 1998; Fownes

    and Harrington 1992; Litton and Kauffman 2008;

    Sampaio and Silva 2005; Segura and Kanninen

    2005). However, Harrington (1979) and Claesson

    et al. (2001) argue that allometric equations for

    biomass estimation based on only one response

    variable are of questionable accuracy. In our study,

    we also used tree height (H) as a second predictor

    variable, since it is often measured in forest inven-

    tories and has been used in allometric models topredict biomass either singly or in combination with

    D. Our results showed that H alone was not as good a

    predictor of total stem biomass, and the combination

    of D and H did not improve the prediction compared

    with D alone. This has also been shown in other

    studies for a range of species (Onyekwelu2004; Salis

    et al.2006; Tumwebaze2008). It is not the number of

    predictor variables built into the model but rather the

    H = 15.18 (1-exp(-0.08D))

    R2

    = 0.80

    Shoot D (cm)

    2 4 6 8

    Sho

    otH

    (m)

    0

    2

    4

    6

    8

    10

    Fig. 3 Diameter versus height relationship for Leucaena-KX2

    shoots that were harvested to develop the allometric model for

    predicting total biomass of stem

    A y = 0.64x1.80

    R2= 0.79

    H (m)

    Totalbiomass(kg)

    0

    10

    20

    30

    B

    0 2 4 6 8 0 2 4 6 8

    Totalbiomass(kg)residuals

    -10

    -5

    0

    5

    10Fig. 4 a Allometric model

    for predicting total biomass(kg) from stem height

    (m) for Leucaena-KX2 and

    b biomass residual of stem

    height

    B

    lnCoarserootbiomassresiduals

    -4

    -2

    0

    2

    4A R

    2= 0.88

    ln Coarse root D (cm)

    -2 -1 0 1 2 3-2 -1 0 1 2 3

    lnCoarserootbiomass(kg)

    -8

    -6

    -4

    -2

    0

    2Fig. 5 a Predicting

    biomass (kg) from coarse

    root diameter (cm) of

    Leucaena-KX2 in Hawaii

    using log transformed linear

    model and b biomass

    residuals of log transformed

    coarse root diameter

    Agroforest Syst (2011) 83:331345 339

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    10/15

    significance of the predictive strength of the individ-

    ual variables that counts. For most studies, predic-

    tions based on stem D result in less error than H alone

    (Otieno et al.1991), and the combination of D and H

    does not necessarily reduce this error (e.g. Litton and

    Kauffman2008).

    Similar to stem D, coarse root basal diameter was an

    effective predictor of individual coarse root biomass.

    In addition, stump D was an accurate predictor of

    total coarse root biomass. These relations provide an

    effective means to estimate biomass, and thus

    belowground carbon pools, using common inventory

    variables.

    The simple exponential (power) equation proved

    superior to other more complex equations for predicting

    biomass and regrowth rates (Table4). This has been the

    most frequently used equation to estimate biomass

    based on stem D, representing plants from different

    A

    y = 0.09 D 1.96

    R2 = 0.93

    Stu

    mpbiomass(kg)

    0

    5

    10

    15

    C

    Stum

    pbiomassresiduals

    -2

    0

    2

    B

    y = 0.06 D2.00

    R2

    = 0.91

    Stump D (cm)

    Totalcoarserootbiomass(kg)

    0

    5

    10

    15

    D

    0 5 10 15 0 5 10 15Totalcoarserootbiomassresiduals

    -2

    0

    2

    Fig. 6 Allometric models

    for predicting biomass (kg)

    for a stump, b total coarse

    root, and c and d biomass

    residuals (observed minus

    predicted values). P \ 0.01

    for all models

    Table 4 Goodness of fit

    comparisons ofLeucaena-

    KX2 model and other

    generalized models for

    predict aboveground and

    belowground biomass (kg)

    D Basal stem diameter

    Models R2 SSR RMS P value

    Individual shoot biomass

    Leucaena KX2-Hawaii

    Y = 0.30 9 D2.25 0.97 7.01 0.37 \0.01

    Y = 0.64 9 H1.80 0.80 30.69 1.90 \0.01

    Brewbaker (2008)

    Y = 5.37 9 10-5

    9 H2.714

    0.82 43.90 2.17 \0.01Y = 1.66 9 10-

    49 H

    2.3650.75 54.22 4.24 \0.01

    Brown (1997)

    Y = exp (-2.134 ? 2.530 9 ln(D)) 0.86 53.85 2.99 \0.01

    Chave et al. (2005):

    Y = 0.0509 9 p D2

    H (moist) 0.92 16.18 0.80 \0.01

    Y = 0.0776 9 (p*D2

    H)0.94

    (wet) 0.85 42.07 2.45 \0.01

    Sampaio and Silva (2005):

    Y = 0.0612 9 (D 9 H)1.5811

    0.86 251.23 17.49 \0.01

    Y = 0.173 9 D2.295

    0.91 184.42 6.36 \0.01

    340 Agroforest Syst (2011) 83:331345

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    11/15

    environments and sizes (Zianis and Mencuccini2004).Although much attention has been placed on developing

    generalized allometric models for tropical trees (Brown

    1997; Zianis and Mencuccini2004; Chave et al.2005;

    Pilli et al. 2006), using generalized equations can lead to

    a bias in estimating biomass for a particular species

    (Clark et al.2001; Cairns et al.2003; Chave et al. 2005;

    Litton and Kauffman 2008; Pilli et al. 2006), Ulti-

    mately, this is more problematic than simply poorer

    accuracy, as bias results in unreliable estimates within

    certain tree size ranges and not just wider confidence

    intervals across the entire range. Major sources of

    uncertainties include (i) differences in site indices,

    including soil and climatic conditions; (ii) unquantified

    diameter and height ranges (Specht and West 2003);(iii) narrow ranges of diameter and height (Burrows

    et al.2000); and (iv) differences in age classes and tree

    density (Chen et al. 1998; Lin et al. 2001). Thus,

    generalized models are subject to the same limitations

    that constrain the application of site-and species-

    specific models.

    Simply adding additional variables, such as tree

    height and wood density, to account for more sources

    of variation, may not overcome these potential

    A

    0

    10

    20

    30

    40

    Leucaena

    Site specific modelUper 95% Conf. Int. of site modelLower 95% Conf. Int. of site modelBrown (moist) 1997

    B

    Totalsh

    ootbiomass(kg)

    0

    10

    20

    30

    40

    C

    Shoot D (cm)

    0 2 4 6 80

    10

    20

    30

    40

    Sampaio & Silva (1) 2005Sampaio & Silva (2) 2005

    Chave (wet) 2005Chave (moist) 2005

    -KX2 Waimanalo

    Fig. 7 Comparison of predicted biomass in generalized

    models using: a D alone, b D alone and both D and H, and

    c D, H, and P (wood specific gravity) as predictor variables

    with predicted biomass and values of 95% confidence intervals

    ofLeucaena-KX2 site-specific model

    H (m)

    0 2 4 6 8

    Totalshootbiomass(kg)

    0

    10

    20

    30

    40

    Leucaena-KX2 Waimanalo

    Site specific model

    Brewbaker (1) 2008

    Brewbaker (2) 2008

    Lower 95% Conf. Int.of site model

    Upper 95% Conf. Int. of site model

    Fig. 8 Comparison of predicted values of individual shoot

    biomass (kg) in Brewbaker (2008) models (1) and (2) with

    predicted biomass and values of 95% confidence intervals of

    Leucaena-KX2 site-specific models, uses shoot height (H) as

    predictor variable

    Table 5 Comparison of coefficients from best-fit allometric

    models (Y = aDb

    ) developed during 2006 and 2009 for pre-

    diction of total shoot biomass regrowth (kg) ofLeucaena-KX2

    (Y) based on stump diameter (D)

    Year a (SE) b (SE) R2

    No mulch

    2006 0.25 (0.17) 2.00 (0.20) 0.93

    2009 0.68* (0.34) 1.49* (1.55) 0.91

    Mulch

    2006 0.26 (0.08) 2.02 (0.14) 0.912009 0.31

    (0.23) 2.06

    (2.55) 0.90

    * Indicates a significant difference between 2006 and 2009 at

    P\ 0.025

    Indicates a significant difference between mulch and

    no-mulch treatments at P \0.025

    Agroforest Syst (2011) 83:331345 341

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    12/15

    limitations. In our study, the models of Chave et al.

    (2005) showed a consistent underestimate of shoot

    biomass, despite the use of stem D, H, and wood

    density. Unlike many of the other generalized

    models, however, there was no increasing divergence

    with increasing stem D, suggesting at least a

    reasonable level of precision in its prediction of

    biomass. Across a wide range of species or site

    conditions, therefore, such models may be beneficial.

    For forest plantations or agroforestry systems, how-ever, the effort required to develop site-specific

    models is recouped in the ability to quickly and

    reliably predict biomass using simple inventory

    measures.

    The 2-step procedure for predicting coarse root

    biomass by Kapeluck and Van Lear (1995) also proved

    useful in this study, despite the potential problems

    introduced by pollarding. One potential concern with

    this method is the use of an allometric equation to

    predict individual root biomass in order to develop the

    allometric relationship between stump D and totalcoarse root biomass. This additional step is introduced

    to save time and effort in developing a site-specific

    allometric. If individual root biomass were poorly

    predicted from root diameter or if there were evidence

    of bias in root biomass prediction, such concerns would

    be valid. In our case, there was no evidence of bias in

    the prediction of individual root biomass, and stump D

    was able to predict total coarse root biomass with

    reasonable accuracy and precision.

    Year

    2006 2007 2008 2009

    Totalshootregrowth(Mgha-1)

    0

    4

    8

    12

    Pollard once + Mulch

    Pollard twice + Mulch

    Pollard once + No mulch

    Pollard twice + No mulch

    Fig. 9 Effect of pollarding frequency (once and twice) and

    mulching (mulch, no mulch) on total shoot regrowth

    (Mg ha-1

    year-1

    ) from 2005 to 2009

    Table 6 Comparison of time and treatment effects on shoot

    growth and stump ? root biomass and C sequestration

    Treatment Slope SE R2 P-value

    Shoot growth

    Pollard once ? no mulch -0.46ab 0.07 0.80 \0.01

    Pollard twice ? no mulch -1.21c 0.06 0.97 \0.01

    Pollard once ? mulch 0.79a 0.15 0.74 \0.01

    Pollard twice ? mulch -0.48b 0.11 0.64 \0.01

    (Stump ? root) biomass

    Pollard once ? no mulch 0.41ab 0.15 0.43 0.02

    Pollard twice ? no mulch 0.24b 0.07 0.55 \0.01

    Pollard once ? mulch 0.84a 0.13 0.82 \0.01

    Pollard twice ? mulch 0.46ab 0.11 0.64 \0.01

    (Stump ? root) C sequestration

    Pollard once ? no mulch 0.19ab 0.07 0.43 0.02

    Pollard twice ? no mulch 0.11b 0.31 0.55 \0.01

    Pollard once ? mulch 0.38a 0.06 0.82 \0.01

    Pollard twice ? mulch 0.21ab 0.05 0.64 \0.01

    Slope coefficient for the time regression, standard error (SE),

    coefficient of determination (R2

    ), and significance of the slope

    term (P-value)

    Means followed by the same letter do not differ significantly

    according to Tukeys honest significant different test

    A

    0

    2

    4

    6

    8

    B

    Year

    2006 2007 2008 2009

    Mgha-1

    0

    1

    2

    3

    4

    Pollard once + Mulch

    Pollard twice + Mulch

    Pollard once + No mulchPollard twice + No mulch

    Fig. 10 Effect of pollarding frequency (once and twice)

    and mulching (mulch, no mulch) on stump ? coarse root:

    a biomass and b carbon sequestration from 2006 to 2009

    342 Agroforest Syst (2011) 83:331345

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    13/15

    Both pollarding frequency and mulch additions

    had significant effects on stem regrowth rates.

    Pollarding twice per year led to a significant decline

    in shoot regrowth over time, regardless of mulch

    addition. The negative effect of more frequent

    pollarding has been reported previously (Duguma

    et al. 1988). After pollarding, plants must use storedreserves to initiate new growth and then replace those

    lost reserves with new assimilates to maintain

    regrowth potential (Lovett and Tobiessen 1993).

    Lower stem biomass production with shorter harvest-

    ing cycles may also be influenced by the season of

    cutting. Seasonal variations in solar radiation and

    average temperature are known to be highly corre-

    lated with leaf yield, leaf N and wood yield of

    Leucaena (Evensen1985). During the winter season

    in Hawaii, days are measurably shorter (10.5 h day

    length minimum vs. 13.5 h maximum), averagetemperatures are *4C cooler, and rainfall and thus

    cloud cover are generally higher. This may result in

    slower regrowth of shoots pollarded just before or

    during winter and therefore a delay in the replenish-

    ment of stored reserves, reducing annual biomass

    production.

    Mulch addition had a positive effect on shoot

    regrowth. Mulch from these pollarded trees provides

    significant quantities of N and organic matter to the

    soil on an annual basis (Youkhana and Idol2009) and

    thus can be used as an organic amendment for soilimprovement. Repeated removal of this nutrient-rich

    organic matter would likely result in nutrient limita-

    tion to the pollarded trees. As with any harvested

    crop, ifLeucaena were to be used in a cut-and-carry

    system for animal fodder, green manure or mulch

    production, restoration of nutrient losses in harvested

    biomass is a key for sustainability of shoot regrowth

    and harvestable yields.

    Finally, the increase in plot-level stump ? coarse

    root biomass and C sequestration from 2006 to 2009

    under even the best conditions (pollard once ?

    mulch) were modest (*1.1 Mg ha-1 of C) compared

    to most tropical forest plantations or unmanaged

    stands. Predicted coarse root biomass of individual

    trees was *15% of shoot regrowth potential on an

    annual basis, slightly lower than the 20% average

    area-based (Mg ha-1) root:shoot ratio predicted for a

    large dataset of upland forests from Cairns et al.

    (1997). The low C sequestration rate is thus an

    inevitable trade-off of managing shade levels through

    pollarding. However, a previous study showed that

    returning mulch to the soil surface increased C

    sequestration as soil organic matter by 10.8 Mg ha-1

    over the same time period (Youkhana and Idol2009).

    Unfortunately, current programs to provide compen-

    sation or recognition of C sequestration in forestry

    and agroforestry systems focus exclusively on theaboveground biomass components (Idol et al. 2011).

    Conclusion

    In a coffea-Leucaena agroecosystem, aboveground

    biomass and shoot regrowth after pollarding can be

    reliably predicted using stem and stump diameter,

    respectively, using a simple exponential relationship.

    Similarly, basal root diameter and stump diameter

    can reliably predict individual and total coarse rootdiameter using the same power function. For this

    system, biomass predictions from generalized models

    are inadequate, as they result in over- or under-

    estimates of biomass, especially as tree size

    increases. Repeated pollarding did not alter the

    allometric relationship between stem D and shoot

    biomass, but shoot regrowth potential and stump ?

    coarse root biomass growth over time were both

    affected by pollarding frequency and mulch addition.

    The allometric equations developed in this study

    should be sufficient for modeling of overall produc-tion and C sequestration of this multipurpose tree

    under standard management practices.

    References

    Albrecht A, Kandji ST (2003) Carbon sequestration in tropical

    agroforestry systems. Agric Ecosyst Environ 99:1527

    Ares A, Quesada JP, Boniche J, Yost RS, Molina E, Smith J

    (2002) Allometric relationships in Bactris gasipaes for

    heart-of-palm production agroecosystems in Costa Rica.

    J Agric Sci 138:285292

    ASTM (1983) Standard test methods for specific gravity of

    wood and wood-base materials. Edition D 2395-83.

    ASTM, Philadelphia, pp 353360

    Bartelink HH (1998) A model of dry matter partitioning in

    trees. Tree Physiol 18:91101

    Bengough AG, Castrignano A, Pages L, Van Noordwijk M

    (2000) Sampling strategies, scaling, and statistics. In:

    Smit AL, Bengough AG, Engels C, Van Noordwijk M,

    Pellerin S, Van de Geijn SC (eds) Root methods: a

    handbook. Springer, Berlin, pp 147173

    Agroforest Syst (2011) 83:331345 343

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    14/15

    Bolte A, Rahmann T, Kuhr M, Pogoda P, Murach D, Gadow K

    (2004) Relationships between tree dimension and coarse

    root biomass in mixed stands of European beech (Fagus

    sylvaticaL.) and Norway spruce (Picea abies[L.] Karst.).

    Plant Soil 264(12):111

    Brewbaker JL (2008) Registration of KX2-Hawaii, interspe-

    cific-hybrid leucaena. J Plant Regist 2:190193

    Brown S (1997) Estimating biomass and biomass change of

    tropical forests: a primer. Forestry paper 134. United

    Nations Food and Agriculture Organization, Rome

    Burrows WH, Hoffmann MB, Crompton JF, Back PV, Tait LJ

    (2000) Allometric relationships and community estimates

    for some dominant Eucalypts in central Queensland

    woodlands. Austr J Bot 48(6):707714

    Cairns MA, Brown S, Helmer EH, Baumgardner GA (1997)

    Root biomass allocation in the worlds upland forests.

    Oecologia 111:111

    Cairns MA, Olmsted I, Granados J, Argaez J (2003) Compo-

    sition and aboveground tree biomass of a dry semi-ever-

    green forest on Mexicos Yucatan Peninsula. For Ecol

    Manag 186:125132

    Chambers JQ, dos Santos J, Ribeiro RJ, Higuchi N (2001) Treedamage, allometric relationships, and above-ground net

    primary production in central Amazon forest. For Ecol

    Manag 152:7384

    Chave J, Andalo C, Brown S, Cairns MA, Chambers JQ, Eamus

    D, Folster H, Fromard F, Higuchi N, Kira T, Lescure JP,

    NelsonBW, OgawaH, Puig H, RieraB, Yamakura T (2005)

    Tree allometry and improved estimation of carbon stocks

    and balance in tropical forests. Oecologia 145:8799

    Chen TH, Sheu BH, Chang CT (1998) Accumulation of stand

    biomass and nutrient contents of Casuarina plantations in

    Suhu coastal area. Taiwan J For Sci 13(4):335349

    Claesson S, Sahlen K, Lundmark T (2001) Functions for bio-

    mass estimation of young Pinus sylvestris, Picea abies

    andBetulaspp. from stands in Northern Sweden with highstand densities. Scand J For Res 16:138146

    Clark DA, Brown S, Kicklighter DW, Chambers JQ, Tomlison

    JR, Ni J (2001) Measuring net primary production in for-

    ests: concepts and field methods. Ecol Appl 11:356370

    Drexhage M, Colin F (2001) Estimation root system biomass

    from breast-height diameters. Forestry 74:491497

    Duguma B, Kang BT, Okali DU (1988) Effect of pruning

    intensities of three woody leguminous species grown in

    alley cropping with maize and cowpea on an alfisol.

    Agrofor Syst 6:1935

    Evensen CLI (1985) Effect of season and forage yield variation

    of Leucaena. Leucaena Res Rep 6:8890

    Fownes JH, Harrington RA (1992) Allometry of woody bio-

    mass and leaf area in five tropical multipurpose trees.

    J Trop For Sci 4(4):317330

    Giambelluca TW, Nullet MA, Schroeder TA (1986) Rainfall

    atlas of Hawaii. Report R76. State of Hawaii. Department

    of Land and Natural Resources, Division of Water and

    Land Development, Honolulu

    Gower ST, Pongracic S, Landsberg JJ (1996) A global trend in

    belowground carbon allocation: can we use relationship at

    smaller scales? Ecology 77:17501755

    Harrington G (1979) Estimating aboveground biomass of trees

    and shrubs in a Eucalyptus populnea F. Mull. Woodland

    by regression of mass on trunk diameter and plant height.

    Aust J Bot 27:135143

    Haynes BE, Gower ST (1995) Belowground carbon allocation

    in unfertilized and fertilized red pine plantations in

    northern Wisconsin. Tree Physiol 15:317325

    Hoffmann CW, Usoltsev VA (2001) Modeling root biomass

    distribution in Pinus sylvestris forests of the Turgai

    Depression of Kazakhstan. For Ecol Manag 149:103114

    Husch B, Beers TW, Kershaw JA Jr (2003) Forest mensuration,

    4th edn. Wiley, Hoboken

    Huxley P (1999) Tropical agroforestry. Blackwell Science,

    Oxford

    Idol T, Haggar J, Cox L (2011) Ecosystem services from

    smallholder forestry and agroforestry in the tropics. In:

    Campbell WP, Lopez SL (eds) Integrating agriculture,

    conservation, and ecotourism: examples from the field.

    Springer, New York (in press)

    Kaonga ML, Bayliss-Smith TB (2009) Allometric models for

    estimation of aboveground carbon stocks in improved

    fallows in eastern Zambia. Agrofor Syst 78(3):217232

    Kapeluck PR, Van Lear DH (1995) A technique for estimating

    below-stump biomass of mature loblolly pine plantations.Can J For Res 25:355360

    Kenzo T, Ichie T, Hattori D, Itioka T, Handa Ch, Ohkubo T,

    Kendawang JJ, Nakamura M, Sakaguchi M, Takahashi N,

    Okamoto M, Tanaka-Oda A, Sakurai K, Ninomiya I

    (2009) Development of allometric relationships for

    accurate estimation of above-and below-ground biomass

    in tropical secondary forests in Sarawak, Malaysia. J Trop

    Ecol 25:371386

    Kurz WA, Beukema SJ, Apps MJ (1996) Estimation root

    biomass and dynamics for the carbon budget of the

    Canadian forest sector. Can J For Res 26:19731979

    Laiho R, Finer L (1996) Changes in root biomass after water-

    level draw down on pine mires in Southern Finland. Scand

    J For Res 11:251260Lambert MC, Ung CH, Raulier F (2005) Canadian national tree

    aboveground biomass equations. Can J For Res 35:

    19962018

    Li Z, Kurz WA, Apps MJ, Beukema SJ (2003) Belowground

    biomass dynamics in the Carbon Budget Model of

    Canadian Forest Sector: recent improvements and impli-

    cations for the estimation of NPP and NEP. Can J For Res

    33:126136

    Lin K, Ma F, Tang S (2001) Allometric equations for pre-

    dicting aboveground biomass of tree species in Fushani

    forest. Taiwan J For Sci 16(3):143151

    Litton CM, Kauffman JB (2008) Allometric models for pre-

    dicting aboveground biomass in two widespread woody

    plants in Hawaii. Biotropica 40:313320

    Lovett GM, Tobiessen P (1993) Carbon and nitrogen assimi-

    lation in red oaks (Quercus rubra L.) subject to defolia-

    tion and nitrogen stress. Tree Physiol 12:259269

    Maxwell SE, Delaney HD (2004) Designing experiments and

    analyzing, 2nd edn. Lawrence Erlbaum Associates, Hills-

    dale. Chapters 13 and 14

    Myung JI, Tang Y, Pitt MA (2010) Evaluation and comparison

    of computational models. In: Johnson ML (Ed) Essential

    numerical computer methods. Elsevier Science & Tech-

    nology Books, Amsterdam. Chapter 21, pp 511527

    344 Agroforest Syst (2011) 83:331345

    1 3

  • 8/13/2019 Allometric eqn- Leucaena

    15/15

    Nadelhoffer KJ, Aber JD, Melillo JM (1985) Fine root, net

    primary production, and soil nitrogen availabilitya new

    hypothesis. Ecology 66:13771390

    Nair PKR, Kumar BM, Nair VD (2009) Agroforestry as a

    strategy for carbon sequestration. J Plant Nutr Soil Sci

    172:1023

    Navar J (2009) Allometric equation for tree species and carbon

    stocks for forests of northwestern Mexico. For Ecol

    Manag 257:427434

    Northup BK, Zitzer SF, Archer S, McMurtry CR, Boutton TW

    (2005) Aboveground biomass and carbon and nitrogen

    content of woody species in a subtropical thornscrub

    parkland. J Arid Environ 62:2343

    Onyekwelu JC (2004) Above-ground biomass production and

    biomass equations for even-aged Gmelina arborea

    (ROXB) plantations in South-western Nigeria. Biomass

    Bioenergy 26:3946

    Otieno K, Onim JFM, Bryant MJ, Dzowela BH (1991) The

    relation between biomass yield and linear measures of

    growth in Sesbania sesban in western Kenya. Agrofor

    Syst 13:131141

    Overman JPM, Witte HJL, Saldarriaga JG (1994) Evaluation ofregression models for above-ground biomass determina-

    tion in Amazon rainforest. J Trop Ecol 10:207218

    Philip MS (1994) Measuring trees and forests, 2nd edn. CAB

    International, Wallingford

    Pilli R, Anfodillo T, Carrer M (2006) Towards a functional and

    simplified allometry for estimating forest biomass. For

    Ecol Manag 237:583593

    Ramsay JO, Silverman BW (2005) Functional data analysis,

    2nd edn. Springer, New York

    Salis SM, Assis MA, Mattos PP, Pia ACS (2006) Estimating

    the aboveground biomass and wood volume of savanna

    woodlands in Brazils Pantanal wetlands based on allo-

    metric correlations. For Ecol Manag 228:6168

    Sampaio EV, Silva GC (2005) Biomass equations for Braziliansemiarid caatinga plants. Acta Bot Bras 19(4):935943

    Santantonio DF, Herman RK, Overton WS (1977) Root bio-

    mass studies in forest ecosystems. Pedobiologia 17:131

    Santos-Martin F, Navaro-Cerrillo RM, Mulia R, van Noo-

    rdwijk M (2010) Allometric equations based on a fractal

    branching model for estimating aboveground biomass of

    four native tree species in the Philippines. Agrofor Syst

    78:193202

    SAS Institute Inc (1990) SAS/STAT users guide. Version 9,

    4th edn. SAS Institute Inc, Cary

    Segura M, Kanninen M (2005) Allometric models for tree

    volume and total aboveground biomass in a tropical

    humid forest in Costa Rica. Biotropica 37:28

    Segura M, Kanninen M, Suarez D (2006) Allometric models

    for estimating aboveground biomass of shade trees and

    coffee bushes grown together. Agrofor Syst 68:143150

    Senelwa K, Sims REH (1998) Tree biomass equations for short

    rotation Eucalyptus grown in New Zealand. Biomass

    Bioenergy 13:133140

    Specht A, West PW (2003) Estimation of sequestered carbon

    on farm forest plantation in northern New South Wales,

    Australia. Biomass Bioenergy 25(4):363379

    Spiess A, Neumeyer N (2010) An evaluation ofR2

    as an inade-

    quatemeasurefor nonlinearmodels in pharmacological and

    biochemical research: a Monte Carlo approach. BMC

    Pharmacol 10:6. http://www.biomedcentral.com/1471-

    2210/10/6

    Tewari SK, Katiyar RS, Ram B, Misra PN (2004) Effect of age

    and season of harvesting on the growth, coppicing char-

    acteristics and biomass productivity of Leucaena leuco-

    cephala and Vitex negundo. Biomass Bioenergy 26(3):

    229234Thies WG, Cunningham PG (1996) Estimating large-root

    biomass from stump and breast-height diameters for

    Douglas-fir in western Oregon. Can J For Res 26:237243

    Tumwebaze SB (2008) Quantifying the amount and spatial

    distribution of soil organic carbon under the linear

    simultaneous agroforestry system. Doctoral thesis, State

    University of New York, College of Environmental Sci-

    ence and Forestry, Syracuse

    Van TK, Rayachhetry MB, Centre D (2000) Estimating

    aboveground biomass of Melaleuca quinquenenervia in

    Florida, USA. J Aquat Plant Manag 38:6267

    von Ende CH (1993) Repeated-measures analysis: growth and

    other time-dependent measures. In: Scheiner SM, Gu-

    revitch J (eds) Design and analysis of ecological experi-ments. Chapman & Hall, New York, pp 113137

    Youkhana A, Idol T (2009) Tree pruning mulch increases soil

    C and N in a shaded coffee agroecosystem in Hawaii. Soil

    Boil Biochem 41(12):25272534

    Zianis D, Mencuccini M (2004) On simplifying allometric

    analysis of forest biomass. For Ecol Manag 187:311332

    Agroforest Syst (2011) 83:331345 345

    1 3

    http://www.biomedcentral.com/1471-2210/10/6http://www.biomedcentral.com/1471-2210/10/6http://www.biomedcentral.com/1471-2210/10/6http://www.biomedcentral.com/1471-2210/10/6