Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes

b i o m a s s a n d b i o e n e r g y x x x ( 2 0 1 3 ) 1 e9

Available online at www.sciencedirect.com

http://www.elsevier.com/locate/biombioe

Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes Shem Kuyah a,b,*, Johannes Dietz c, Catherine Muthuri a,b, Meine van Noordwijk d, Henry Neufeldt c a

Jomo Kenyatta University of Agriculture and Technology (JKUAT), P.O. Box 62000, 00200 Nairobi, Kenya World Agroforestry Centre (ICRAF), P.O. Box 30677, Nairobi 00100, Kenya c World Agroforestry Centre (ICRAF), P.O. Box 1558, Lima 12, Peru d World Agroforestry Centre (ICRAF), P.O. Box 161, Bogor 16001, Indonesia b

article info

abstract

Article history:

Farmers in developing countries are one of the world’s largest and most efficient producers of

Received 27 September 2012

sequestered carbon. However, measuring, monitoring and verifying how much carbon trees

Received in revised form

in smallholder farms are removing from the atmosphere has remained a great challenge in

6 February 2013

developing nations. Devising a reliable way for measuring carbon associated with trees in

Accepted 11 February 2013

agricultural landscapes is essential for helping smallholder farmers benefit from emerging

Available online xxx

carbon markets. This study aimed to develop biomass equations specific to dominant eucalyptus species found in agricultural landscapes in Western Kenya. Allometric relationships

Keywords:

were developed by regressing diameter at breast height (DBH) alone or DBH in combination

Biomass equations

with height, wood density or crown area against the biomass of 48 trees destructively sampled

Carbon stocks

from a 100 km2 site. DBH alone was a significant predictor variable and estimated above-

Eucalyptus

ground biomass (AGB) with over 95% accuracy. The stems, branches and leaves formed up to

Height

74, 22 and 4% of AGB, respectively, while belowground biomass (BGB) of the harvested trees

Wood density

accounted for 21% of the total tree biomass, yielding an overall root-to-shoot ratio (RS) of 0.27, which varied across tree size. Total tree biomass held in live Eucalyptus trees was estimated to be 24.4
0.01 Mg ha1, equivalent to 11.7
0.01 Mg of carbon per hectare. The equations presented provide useful tools for estimating tree carbon stocks of Eucalyptus in agricultural landscapes for bio-energy and carbon accounting. These equations can be applied to Eucalyptus in most agricultural systems with similar agro-ecological settings where tree growth parameters would fall within ranges comparable to the sampled population. ª 2013 Elsevier Ltd. All rights reserved.

1.

Introduction

Fast growing species such as Eucalyptus have been introduced into many tropical countries to mitigate the dwindling supply

of wood, especially for timber and biomass fuel [1]. In East Africa, eucalyptus species are common in farmed landscapes and the preferred species in managed plantations [2]. Eucalyptus is popular because of its fast-growing nature, multiple

* Corresponding author. Jomo Kenyatta University of Agriculture and Technology (JKUAT), P.O. Box 62000, 00200 Nairobi, Kenya. Tel.: þ254 20 722 4425; fax: þ254 20 7224001. E-mail addresses: [email protected], [email protected] (S. Kuyah). 0961-9534/$ e see front matter ª 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biombioe.2013.02.011

Please cite this article in press as: Kuyah S, et al., Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes, Biomass and Bioenergy (2013), http://dx.doi.org/ 10.1016/j.biombioe.2013.02.011

2

b i o m a s s a n d b i o e n e r g y x x x ( 2 0 1 3 ) 1 e9

uses and importance as a ‘cash crop’ [1]. The higher proportion of Eucalyptus in agricultural landscapes is largely due to greater attention given to trees on farms, mainly because of their ability to couple economic gains with social services and environmental benefits. Despite the acknowledged importance of Eucalyptus trees, there is little knowledge about the amount of carbon that will be emitted to the atmosphere when areas dominated by these species are converted to other land uses. In Kenya, the majority of timber and non-timber wood products are obtained from farm estates, presenting an opportunity for farmers to have access to additional income from their land [3]. To meet the high demand for tree products and services, considerable efforts have been focused on conserving and also increasing trees in the landscape. Two kinds of eucalyptus trees are favored on Kenyan farms: (1) naturalized eucalyptus introduced from Australia in the colonial era, and (2) eucalyptus hybrids introduced from South Africa. Eucalyptus plantations provide raw material for industries such as sawmilling, pulp and charcoal, while eucalyptus in agricultural landscapes supply timber, poles, building material and fuelwood, both for domestic and commercial purposes. Despite considerable efforts made to increase trees on farms, quantities of major biofuel trees continue decreasing in Kenya [2]. This decline has been attributed to competing demand for timber, food crop cultivation, human settlement and harvesting trees for fuelwood [1]. The role played by woody vegetation in the global carbon cycle has led to increased interest in estimating biomass held in all land uses, in addition to forests. Estimation of tree biomass in agricultural ecosystems can be crucial for the sustainable management of woody vegetation and also an essential component of monitoring carbon sequestration [4]. Periodic measurement of biomass accumulation can be used to establish the value of a given agroforestry practice. One can further determine the production potential or suitability of a certain species for a particular purpose, e.g. charcoal production. Measurement approaches can also be designed to predict harvest yield, thus helping to assess biomass loss or accumulation over time. By establishing the rate of biomass production, one can determine carbon sequestration potential of particular species. This allows the potential of trees in agricultural landscapes to offset anthropogenic carbon emissions to be established. The Intergovernmental Panel on Climate Change, IPCC [5], provides methodologies for estimating tree carbon stocks. Without further data, IPCC default values are normally used, choosing a Tier 1 approach. This practice uses equations provided at global scale, stratified by eco-climatic zones such as those provided in Brown [6] or Chave et al. [7]. Kuyah et al. [8,9] showed that application of broadly derived forest-based equations to trees in agricultural landscapes yields biased estimates. Such bias can be minimized by applying Tier 2 or Tier 3 approaches, where country-specific models derived for local conditions are used. In this case, equations that adequately address unique project circumstances are either developed or chosen from the literature. Since tree species differ in architecture and wood gravity [10], and because the dimensional relationship of trees can be changed by silvicultural interventions, species-specific equations are necessary

to produce reliable biomass estimates [11,12]. However, insufficient species-specific equations exist, particularly in sub-Saharan Africa where less than 1% of the tree species have country-specific equations [11]. Since it is not practical to fell trees to develop equations for each species, and because destructive sampling is generally not acceptable for rare species or in areas where the eventual objective is to conserve trees, biomass equations that cover trees with comparable shapes could be developed. This study aimed to (1) build biomass equations specific to dominant eucalyptus species found in Western Kenyan agricultural landscapes, (2) establish the suitability of mixed species equations developed for agricultural landscapes, and (3) determine the biomass distribution in the above- and below-ground fractions of Eucalyptus.

2.

Methods

2.1.

Description of the study area

The study was conducted on a 100 km2 site on the Yala River basin covering part of Kakamega and Vihiga Counties in Western Kenya. The study site (latitude 0 70 51.5700 N; longitude 34 490 13.0100 E) covers an altitudinal gradient of 1430e1720 m above sea level. The regional climate around the VihigaKakamega area is influenced by two rainy seasons: long rains from March to July and inconsistent short rains from August to November, allowing up to two cropping seasons per year. The mean annual precipitation is about 1950 mm, decreasing westwards, while temperatures range between 12 and 27  C, with a mean annual temperature of 20.5  C. The landscape consists of mountainous highlands with numerous small streams and clusters of wetlands [13]. Acrisols, Ferralsols and Nitisols form the dominant soil types [14]. The soils are characterized by a good physical structure, but have low nutrient reserves due to prolonged weathering and intense agricultural use [15]. The soils are predominantly clay (46%) and silty clay (32%), although some clay loam (15%) and silt loam (5%) exist, while sand and loam together represent only 2% [13]. Agriculture is the economic mainstay in the area, with land-use systems varying from mainly subsistence smallholder farms to a few cash crop-oriented farms. Tea is the major cash crop, while maize, beans, banana and sweet potatoes are grown for subsistence. Woody vegetation forms part of the complex agricultural mosaic on smallholder farms, varying from individual free standing trees to pockets of stands that consist of indigenous and exotic forms managed in different ways [16]. Trees are grown around homesteads, in cropland and along farm boundaries, while woodlots occur mainly as small, mono-specific clusters of exotic trees, usually Eucalyptus.

2.2.

Field measurement and biomass sampling

Trees used to build allometric equations were selected from 30 m  30 m plots contained in the 100 km2 block established within the Western Kenya integrated ecosystem management project [13] and described in Kuyah et al. [8]. The block is

Please cite this article in press as: Kuyah S, et al., Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes, Biomass and Bioenergy (2013), http://dx.doi.org/ 10.1016/j.biombioe.2013.02.011

b i o m a s s a n d b i o e n e r g y x x x ( 2 0 1 3 ) 1 e9

divided into clusters of 2.5  2.5 km, with 10 plots in each cluster. All trees within the block were stratified into five diameter classes of <10, >10e20, >20e30, >30e50, and >50 cm to achieve an adequate spread of diameter classes in the sample. In each diameter class 10 individual trees were randomly selected for harvesting, to capture trees of different sizes in each diameter class and also to represent larger trees that should be included despite their low proportionate number, for the sake of reducing large significant errors associated with these trees. Only 8 trees were found to have DBH > 50 cm, resulting in a total of 48 trees. Trees were selected irrespective of pruning, broken tops and other defects, in order to capture a sample that is representative of the shape and structure of trees in agricultural landscapes. All trees with DBH > 2.5 cm within each 30 m  30 m plot sampled were identified and measured, yielding a further 480 trees for estimating representative biomass and consequent carbon stocks held in Eucalyptus in the area. Diameter at breast height was measured in cm at 1.3 m above the ground, while root collar diameter (RCD) was measured at 10 cm above the onset of the roots. Both DBH and RCD were measured using diameter tapes. Total tree height was measured on felled trees using a 50 m measuring tape. Crown area and wood density for harvested trees were determined as described in Kuyah et al. [8]. Measured trees were felled using a chain saw, separated into stem, branches and leaves, and the fresh weight of each component determined on-site by weighing on a balance (300 kg) to the nearest 0.1 kg. Belowground fresh weight, i.e. the stump together with coarse roots (>2 cm diameter), was estimated by excavating the whole root system within a radius of 2 m from the edge of the stump and a maximum depth of about 2 m. Soil embedded in the stump joints and on root surfaces was removed and their fresh weight determined on a 300 kg electronic balance. The biomass of roots not captured by excavation was determined by allometric equations relating the biomass of extracted root segments to their proximal and distal diameters as described in Kuyah et al. [9]. Samples were collected from each component, weighed on a 3000 g scale (
0.1 g) and transported in ziploc bags to the laboratory for dry weight determination. Two samples just above and below breast height, were collected from each stem, while three samples were collected near the stembranch bifurcation, at the middle and at the branch end of main branches and roots to account for variation in wood density. Labeled samples were oven-dried for 24 h at 105  C and their dry weight determined on a 3000 g scale. The biomass of each component was obtained by multiplying component fresh weight by their respective dry weight-tofresh weight ratio. AGB of individual trees was attained by adding up dry weights from the stems, branches and leaves. Above- and below-ground biomass was aggregated to yield total tree biomass.

2.3.

Biomass equations

Data from 48 destructively sampled trees consisting of 29 Eucalyptus camaldulensis, 11 Eucalyptus grandis and 8 Eucalyptus saligna were pooled to develop biomass equations based on the assumption that trees in the same genus exhibit

3

similarities in growth form [12], and that the allometry of woody biomass for trees growing under similar conditions does not differ significantly within the same genus [17]. The data were checked for errors prior to analysis and screened for outliers using scatter plots, the latter of which also revealed the relationship between biomass and predictor variables. Biomass values were regressed on DBH and RCD alone, and also on DBH in combination with height, wood density or crown area to obtain allometric coefficients for biomass equations. Equations were built and validated using crossvalidation with multiple sample holdouts [8]. Forty-two trees were used to form equations and the remaining 6 trees from each of the holdouts were used to validate the equations, such that each tree was used once in the training set and also once in the validation set. All equations were built using generalized linear models in Genstat 12th Edition (VSN International Ltd) and transformed to allometric power function equations, Y ¼ aXb where Y is the biomass and X is the predictor variable, e.g. DBH in ln(Y ) ¼ a þ b  ln(DBH) or RCD in ln(Y ) ¼ a þ b  ln(RCD), while a (intercept) and b (exponent) are allometric coefficients. The effects of height (H ), wood density (r) and crown area (ca) as supporting parameters to DBH were tested by adding the variables to the DBH only equation as in ln(Y ) ¼ a þ b  ln(DBH) þ c  ln(H ), ln(Y ) ¼ a þ b  ln(DBH) þ c  ln(r) and ln(Y ) ¼ a þ b  ln(DBH) þ c  ln(ca), respectively. The suitability of using local generic equations including ‘all tree species’ by Kuyah et al. [8,9] to estimate biomass of Eucalyptus was evaluated. The relative error (RE) determined as the relative difference between the predicted and the measured biomass (BM), RE% ¼ fðBMpredicted  BMmeasured Þ=BMmeasured g  100 [7], and the Akaike information criterion, AIC [18], were used to select the most suitable model. The relative error allowed the quality of the estimate to be determined by assessing the difference between the estimate and the true value [8]. The AIC is a way of selecting an equation from a set of (alternative) equations [18] by balancing changes in the goodness-of-fit versus differences in the number of parameters [19]. The best equation was the one with the lowest AIC value [18], and an RE close to zero [8]. Coefficient of determination (R2) was used to assess model fit for equations with one parameter, while equations with multiple parameters were assessed using the adjusted coefficient of determination (R2adj: ) to reflect both the number of independent variables in the equation and the sample size.

3.

Results

3.1.

Dendrometric relationships

Fig. 1 shows the regression of biomass as a function of a) DBH, and b) RCD for estimating the biomass of different components. Diameter at breast height was strongly and significantly correlated with above- and below-ground biomass, accounting for 98 and 96% of the variation in above- and below-ground biomass fractions, respectively (Fig. 1a). Likewise, the regression of stem and branch biomass as a function of DBH was significant (P < 0.001) with high R2 values of 0.97 and 0.95, respectively. Foliage, however, had a moderate

Please cite this article in press as: Kuyah S, et al., Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes, Biomass and Bioenergy (2013), http://dx.doi.org/ 10.1016/j.biombioe.2013.02.011

4

b i o m a s s a n d b i o e n e r g y x x x ( 2 0 1 3 ) 1 e9

Fig. 1 e The regression of biomass as a function of a) diameter at breast height, and b) root collar diameter, for aboveground biomass (AGB), belowground biomass (BGB), stem biomass, branch biomass and leaf biomass in Eucalyptus.

correlation, indicating that DBH accounted only for about 85% of the variation in leaf biomass. RCD was more closely related to BGB (R2 ¼ 0.91) than to AGB (R2 ¼ 0.88) and explained 87, 86 and 82% of the variation observed in stem, branch and leaf biomass, respectively (Fig. 1b). Tree height showed a strong positive correlation with DBH, a near normal distribution and greater variance for harvested trees of larger diameter, DBH > 30 cm (Fig. 2a). The height distribution for non-harvested trees was negatively skewed, with 58 and 35% of the trees measured having values below 10 m and between 10 and 20 m, respectively (Fig. 2b). A moderate correlation was observed between DBH and the height of harvested (R2 ¼ 0.86) and non-harvested trees (R2 ¼ 0.69), respectively.

3.2.

Biomass equations

Empirical allometric coefficients for estimating biomass of different components based on DBH and RCD are presented in Table 1. DBH was a significant predictor variable for all components (P < 0.001) and estimated stem and AGB with a small RE (2.5%). However, DBH overestimated belowground and

branch biomass by 30% and foliage biomass by 16%. RCD was a significant predictor variable for all biomass components (P < 0.001), but overestimated the biomass of most components by 30%, except foliage biomass which was estimated with a small RE of 3.5%. Local generic equations by Kuyah et al. [8,9] for estimating above-, AGB ¼ 0.091  (DBH)2.472, and belowground biomass, BGB ¼ 0.048  (DBH)2.303, in agricultural landscapes overestimated biomass by 10 and 47.8%, respectively. Whereas both equations by Kuyah et al. [8,9] and the equations developed for Eucalyptus generally overestimated biomass (Fig. 3a and b), the equations for Eucalyptus still yielded estimates about twice as accurate as the generic equations. The diameter-based equation for stem and AGB showed low RE, <5% across the diameter classes above 10 cm DBH (Fig. 4a). However, DBH showed variation in error disaggregation across the size classes for belowground biomass, overestimating the biomass of trees with DBH < 20 cm by over 30%. Diameter-based equations for estimating branch and leaf biomass showed high and variable RE across tree size. Allometric equations based on RCD overestimated all components by over 20% for most diameter classes, except for trees with DBH > 50 cm, where biomass was estimated with RE of about 10%.

Fig. 2 e The relationship between height and diameter at breast height of eucalyptus trees for a) harvested trees, and b) nonharvested trees sampled in Kakamega-Vihiga area in Western Kenya. The vertical bars represent height distribution across diameter classes based on stratification of DBH into <10, 20, 30, 40 and >50 cm. Please cite this article in press as: Kuyah S, et al., Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes, Biomass and Bioenergy (2013), http://dx.doi.org/ 10.1016/j.biombioe.2013.02.011

b i o m a s s a n d b i o e n e r g y x x x ( 2 0 1 3 ) 1 e9

Table 1 e Allometric coefficients and the standard error of the estimate (SEE) for estimating biomass of different compartments using a) diameter at breast height, b) root collar diameter and c) the equations by Kuyah et al. [8,9]. R2 is the coefficient of determination, RE is the relative error. Compartment Allometric coefficients P-value R2 RE% a(SEE) a) Aboveground Belowground Total tree biomass Stem Branches Leaves b) Aboveground Belowground Total tree biomass Stem Branches Leaves c) Aboveground Belowground

b(SEE)

0.085(0.090) 2.471(0.029) 0.029(0.205) 2.432(0.065) 0.114(0.093) 2.463(0.029)

<0.001 0.994 2.5 <0.001 0.962 29.9 <0.001 0.993 6.0

0.058(0.109) 0.009(0.272) 0.042(0.217) 0.040(0.274) 0.011(0.233) 0.051(0.247)

2.496(0.035) 2.652(0.087) 1.847(0.069) 2.458(0.079) 2.469(0.067) 2.461(0.071)

<0.001 <0.001 <0.001 <0.001 <0.001 <0.001

0.027(0.276) 0.004(0.435) 0.024(0.298) 0.091(0.131) 0.048(0.199)

2.483(0.079) 2.631(0.125) 1.830(0.085) 2.472(0.039) 2.303(0.059)

<0.001 0.943 28.5 <0.001 0.872 94.4 <0.001 0.899 3.5 10.0 47.8

0.991 0.947 0.938 0.941 0.971 0.962

3.6 30.0 16.1 29.2 30.7 24.2

Height was a significant predictor variable only for aboveground, stem and branch biomass (Table 2; P < 0.05). Crown area was a significant predictor variable for only branch biomass, while wood density was not a significant predictor variable for any of the components. Height, crown area and wood density did not improve R2, which was already high (R2 ¼ 0.99). AIC showed that tree height was a more suitable supportive proxy for estimation of AGB of Eucalyptus (AIC ¼ 46) than wood density (AIC ¼ 61) and crown area (AIC ¼ 62).

3.3.

Biomass estimates

The contribution of different components to the total tree biomass varied considerably. AGB accounted for most of the total tree biomass, with the stems, branches and leaves

5

contributing 73.7, 22.3 and 4.0% to AGB, respectively. Much of the tree biomass is held in the stem, which constitutes more than half of the total tree biomass, while branches and leaves make up 17.6 and 3.2% of the total tree biomass, respectively (Fig. 5). While the proportion of stem biomass was almost constant for all sizes classes, the percentage of branch biomass showed, on average, an increase with tree size, although the trend was not continuous. Changes in biomass allocated to leaves complemented those observed in the stem and branches inversely. The proportion of foliage declined from 14.3% in small trees (DBH < 10 cm) to 3.3% in high biomass trees (DBH > 50 cm). At any diameter, the variation in the proportion of AGB in the stem was large and constant. The BGB of the harvested trees accounted for 21.1% of the total tree biomass, yielding an overall RS of 0.27, which varied across the diameter classes. The biomass of coarse roots not captured by excavation was estimated to be 0.85 Mg, which is 9.2% of the total BGB for harvested trees. RS ranged between 0.14 and 1.07, with a mean and standard error (SE) of 0.31
0.02 and a median of 0.28. RS for individual trees showed a slight decrease with increasing tree size (Fig. 6a and b). High RS values, >0.5 were mainly associated with coppiced trees as indicated in Fig. 6a, where RS values are plotted against DBH; this is less sensitive to effects of silvicultural intervention than RCD. The equations with DBH alone, AGB ¼ 0.085  (DBH)2.471 and BGB ¼ 0.029  (DBH)2.432estimated biomass held in Eucalyptus to be 18.7
0.2, and 5.6
0.1 Mg ha1 for above- and below-ground fractions, respectively. The biomass of stems, branches and leaves was estimated to be 13.8
0.02, 2.7
0.004 and 1.3
0.001 Mg ha1, respectively. This trend is consistent with the respective proportions outlined for aboveand below-ground segments of harvested trees. Biomass estimates obtained from DBH-only equations for different components were converted to carbon stocks using the 47.5% carbon fraction determined for Eucalyptus by element analysis. Eucalyptus in agricultural landscapes of Western Kenya was then estimated to stock about 9.0
0.01 Mg of carbon per hectare in AGB, most of which is held in the stem, 6.6
0.01 Mg ha1, while belowground systems contribute

Fig. 3 e Comparisons of biomass estimated by the equations developed for Eucalyptus and the local generic (mixed species) equations by Kuyah et al. [8,9] for estimating a) aboveground biomass (AGB), and b) belowground biomass (BGB). Please cite this article in press as: Kuyah S, et al., Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes, Biomass and Bioenergy (2013), http://dx.doi.org/ 10.1016/j.biombioe.2013.02.011

6

b i o m a s s a n d b i o e n e r g y x x x ( 2 0 1 3 ) 1 e9

Fig. 4 e Variability of the relative error associated with equations developed for estimating: a) aboveground biomass, b) belowground biomass, c) total tree biomass, d) stem biomass, e) branch biomass, and f) leaf biomass in Eucalyptus using diameter at breast height (DBH) and root collar diameter (RCD). The error bars indicate the standard errors of the average relative error per diameter class.

about 2.7
0.003 Mg of carbon per hectare. In total, Eucalyptus dominated agricultural landscapes in Western Kenya were estimated to stock 11.7
0.01 Mg of carbon per hectare in live tree biomass, on average.

4.

Discussion

Diameter at breast height had a strong positive relationship with all biomass components evaluated (R2 ¼ >0.80). Similarly good fits of high R2 have been reported by Zewdie et al. [20] for branches (0.79), stem (0.94) and AGB (0.86), regressing against

DBH in Eucalyptus globulus plantations of central Ethiopia. The relationship between leaf biomass and DBH was less pronounced than between stem biomass and DBH, which makes sense considering that leaves are transient tissues and also affected more by changes that alter allometric relationships [21]. For example, removing large branches to provide wood and trimming the crown to reduce competition with other trees and crops are common practices in agricultural landscapes. Such interventions, together with interplant competition and other factors, modify the canopy size of individual trees and are likely to affect allometric relationships between diameter and component biomass.

Table 2 e Allometric coefficients and the standard error of the estimate (SEE) for estimating a) aboveground biomass, b) stem biomass and c) branch biomass using diameter at breast height (DBH) as the primary predictor variable, and height, wood density or crown area as supporting variables. R2adj: is the adjusted coefficient of determination, AIC is the Akaike information criterion (AIC), and RE is the relative error. Predictor variable

a)

b)

c)

DBH, DBH, DBH, DBH, DBH, DBH, DBH, DBH, DBH,

height wood density crown area height wood density crown area height wood density crown area

Allometric coefficients a(SEE)

b(SEE)

c(SEE)

0.067(0.143) 0.095(0.161) 0.079(0.114) 0.032(0.141) 0.072(0.201) 0.044(0.131) 0.018(0.428) 0.008(0.505) 0.015(0.310)

2.307(0.083) 2.464(0.032) 2.476(0.062) 2.089(0.082) 2.477(0.040) 2.663(0.071) 3.164(0.249) 2.662(0.102) 2.138(0.167)

0.260(0.123) 0.120(0.106) 0.016(0.038) 0.646(0.122) 0.184(0.133) 0.085(0.044) 0.803(0.370) 0.028(0.334) 0.347(0.103)

P-value

R2adj:

AIC

RE%

0.041 0.268 0.680 <0.001 0.175 0.059 0.036 0.934 0.002

0.994 0.991 0.991 0.994 0.989 0.988 0.948 0.933 0.933

46 61 62 43 74 80 43 53 53

2.3 9.0 13.1 79.4 8.0 3.3 27.6 13.6 0.5

Please cite this article in press as: Kuyah S, et al., Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes, Biomass and Bioenergy (2013), http://dx.doi.org/ 10.1016/j.biombioe.2013.02.011

b i o m a s s a n d b i o e n e r g y x x x ( 2 0 1 3 ) 1 e9

Fig. 5 e The proportion of biomass for the different components of the trees as a percentage of the total tree biomass. Roots represent the biomass of stumps and coarse roots.

Diameter at breast height alone was the best independent variable for describing the different biomass components, estimating stem, aboveground and total tree biomass with about 95% accuracy. BGB was overestimated by both RCD and DBH based equations, confirming previous reports that BGB is a major component of uncertainty in measuring total tree biomass [22]. While RCD is a useful parameter to estimate belowground biomass, especially when trees are already cut down and only stump dimensions are available [23], the bias associated with RCD as proxy for biomass in this study was high and showed greater variability across tree size than DBH. This high and inconsistent RE could be attributed partly to uncertainties in measuring RCD where stems tend to exhibit a much more fluted cross section. This is even more pronounced with increasing tree size. The biomass of branches and leaves could not be accurately estimated using DBH, possibly due to the ephemeral nature of these components [21].

7

The biomass of small trees was generally overestimated, while the tendency to overestimate biomass dropped with increasing tree size. A similar observation was made by Pe´rez-Cruzado and Rodrı´guez-Soalleiro [24]on Eucalyptus nitens plantations in northwestern Spain. This indicates that error in biomass estimation depends on the average tree size. Other authors have reported the importance of tree size in both formulation and use of allometric equations [10,23]. Chave et al. [25] reported that biomass values of the smallest trees strongly affect values of allometric coefficients, while Kuyah et al. [8] showed that it is difficult to accurately estimate the biomass of small trees. Both below- and aboveground biomass were overestimated by between 10 and 48% using the equations by Kuyah et al. [8,9], which had been established under the dominance of Eucalyptus trees. This underscores the need for equations that adequately describe the architecture of the target population and indicates that genetic differences in tree form seem to have a significant influence on metric scaling relationships. Height, wood density and crown area data did not improve the relationship between DBH and any of the tree biomass components. Whereas including height as an additional predictor variable to DBH has been reported to increase R2 for E. globulus coppice plantations [20] and other mixed species equations [25,26], findings from this study showed that addition of height data neither improved R2, nor significantly reduced the RE in diameter-based equations. This observation is in line with several conclusions that height, as an additional predictor, only adds marginal value to the predictive ability of diameter-based equations [27e29]. Height as a supporting predictor variable was a significant predictor of stem biomass, and has been found to improve stem biomass estimates elsewhere [24]. However, height did not improve accuracy of estimating AGB due to the compensatory relationship between stem and canopy mass [30]. This compensatory relationship resulted in similar AGB for trees of the same diameter, but different partitioning in leaves, branches and stems. Compensatory relationships have been observed in E. nitens plantations in northwestern Spain Pe´rez-Cruzado and Rodrı´guez-Soalleiro

Fig. 6 e Variability of root-to-shoot ratio (RS) for individual Eucalyptus trees across a) diameter at breast height, and b) root collar diameter. The solid circles indicate the mean RS with standard error for each of the diameter classes. Please cite this article in press as: Kuyah S, et al., Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes, Biomass and Bioenergy (2013), http://dx.doi.org/ 10.1016/j.biombioe.2013.02.011

8

b i o m a s s a n d b i o e n e r g y x x x ( 2 0 1 3 ) 1 e9

[24] and Eucalyptus pilularis from contrasting sites in Australia [30]. Wood density marginally improved model fit for AGB and stem biomass, but did not improve the predictive power of the diameter-based equations, contrary to previous reports [7,8,27]. This could be due to much lower variation in wood density relative to tree size between trees sampled; hence stem wood density did not appear to affect the allometric relationship between DBH and biomass of aboveground parts [30]. Height and wood density are susceptible to the influence of site factors and differences between tree species. The influence of these parameters was not obvious in the trees evaluated, possibly because they belong to the same genus with one species, E. camaldulensis dominating the dataset and were sampled from a site with near similar climatic and edaphic conditions. Eucalyptus species have the potential to contribute a significant amount of carbon sequestered in Western Kenya’s agricultural landscapes since they constitute 59% of all trees encountered in agricultural landscapes [8]. This dominance has been observed for decades [17]. The stem biomass accounted for the largest proportion of individual tree AGB across all size classes, followed by branches and leaves. This division agrees with results from the studies on growth of E. globulus in central Ethiopian plantations [20] and E. nitens plantations on the coast of Arauco, Chile [31]; both studies reported the stem biomass proportion greater than 75% of AGB. The proportion of leaf biomass (4%) determined in this study compares well with 3% given by Mun˜oz et al. [31]. The proportion of stem biomass increased slightly with tree size, consistent with the biomechanical requirements of the stem; that as tree biomass increases, the stem increases in diameter in order to provide greater mechanical strength for supporting the growing weight. Whereas the biomass of stem, branches and BGB generally increased proportionally with tree size, the biomass of leaves tended to decrease. Zewdie et al. [20] also showed that trees allocate proportionally less biomass to leaves and more to stems as they age, resulting in a decrease in leaf biomass with tree size. The BGB is an important carbon pool for many vegetation types and land-use systems identified by the IPCC. The proportion of BGB relative to total tree biomass determined in this study (21%) is slightly lower than 26% given by Cairns et al. [32] for tropical hardwood species. The RS median value determined in this study (0.28) is identical to the RS median determined for mixed species by Kuyah et al. [9], but higher than the IPCC default value of 0.24
0.14 for tropical hardwood species [32]. The median RS was preferred to the RS mean (0.31
0.02) in order to reduce the influence of large outliers in the dataset arising from coppices or pruning. However, both the RS mean and RS median are lower than 0.38 obtained by Eamus et al. [33] for an open eucalyptus forest in a savanna of northern Australia. Trees in the surveyed landscape are more likely to have invested less in BGB as water and nutrients are not considered limiting in the area. The low RS compared to other studies may indicate that high biomass proportion was allocated to stem, branches and leaves as a possible mechanism to confer a competitive advantage, allowing trees to out-compete neighbors by growing in height and expanding crown area to shade out competitors.

5.

Conclusion

Reliable methods for estimating carbon in the trees of agricultural landscapes are required if farmers are to benefit from any carbon sequestered by their trees. Diameter-based equations predicted biomass of most compartments with 95% accuracy, and with about the same RE across trees of different size. Given that DBH is easy to measure with high accuracy, the equations provide a useful tool for estimating biomass and carbon stocks of Eucalyptus for purposes such as bio-energy and carbon sequestration. These equations can be best applied to Eucalyptus trees in agricultural landscapes in similar agro-ecological zones, provided that tree growth parameters fall within similar ranges to those of the sampled population.

Acknowledgment This work was carried out within the Carbon Benefits Project and was supported by a grant from the Global Environment Facility. The authors recognize Tom Ochinga, Joash Mango and Walter Adongo for helping with data collection, the farmers who allowed their trees to be felled and Kristi Foster for reading the advanced draft of this article.

references

[1] Dessie G, Erkossa T. Eucalyptus in East Africa. In: Socioeconomic and environmental issues. Rome, Italy: FAO; 2011 May. p. 42, Working paper FP46/E. [2] Kituyi E, Marufu L, Wandiga SO, Jumba IO, Andreae MO, Helas G. Biofuel availability and domestic use patterns in Kenya. Biomass Bioenerg 2001;20:71e82. [3] Kiplagat JK, Wang RZ, Li TX. Renewable energy in Kenya: resource potential and status of exploitation. Renew Sust Energ Rev 2011;15:2960e73. [4] Eamus D, McGuinness K, Burrows W. Review of allometric woody biomass for Queensland, the Northern Territory and Western Australia. Canberra, Australia: Australian Greenhouse Office; 2000 August. p. 64, Report no. 5A. [5] IPCC. Good practice guidance for land use, land-use change and forestry. In: Penman J, editor. IPCC national greenhouse gas inventories programme. Hayama, Japan: Institute for Global Environmental Strategies (IGES) for the Intergovernmental Panel on Climate Change (IPCC); 2003. [6] Brown S. Estimating biomass and biomass change of tropical forests: a primer. Rome, Italy: FAO; 1997. p. 55, FAO forestry paper e 134. [7] Chave J, Andalo C, Brown S, Cairns M, Chambers J, Eamus D, et al. Tree allometry and improved estimation of carbon stocks and balance in tropical forests. Oecologia 2005;145:87e99. [8] Kuyah S, Dietz J, Muthuri C, Jamnadass R, Mwangi P, Coe R, et al. Allometric equations for estimating biomass in agricultural landscapes: I. aboveground biomass. Agric Ecosyst Environ 2012;158:216e24. [9] Kuyah S, Dietz J, Muthuri C, Jamnadass R, Mwangi P, Coe R, et al. Allometric equations for estimating biomass in agricultural landscapes: II. belowground biomass. Agric Ecosyst Environ 2012;158:225e34. [10] Chave J, Condit R, Aguilar S, Hernandez A, Lao S, Perez R. Error propagation and scaling for tropical forest biomass estimates. Philos Trans R Soc Lond 2004;359:409e20.

Please cite this article in press as: Kuyah S, et al., Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes, Biomass and Bioenergy (2013), http://dx.doi.org/ 10.1016/j.biombioe.2013.02.011

b i o m a s s a n d b i o e n e r g y x x x ( 2 0 1 3 ) 1 e9

[11] Henry M, Picard N, Trotta C, Manlay RJ, Valentini R, Bernoux M, et al. Estimating tree biomass of sub-Saharan African forests: a review of available allometric equations. Silva Fenn 2011;45:477e569. [12] Fatemi FR, Yanai RD, Hamburg SP, Vadeboncoeur MA, Arthur MA, Briggs RD, et al. Allometric equations for young northern hardwoods: the importance of age-specific equations for estimating aboveground biomass. Can J For Res 2011;891:881e91. [13] Boye A, Verchot L, Zomer RJ. Baseline report, Yala and Nzoia river basins. Western Kenya intergrated ecosystem management project. Findings from the baseline surveys. Nairobi, Kenya: World Agroforestry Centre; 2008 March. p. 71. [14] Jaetzold R, Schmidt H. Farm management handbook of Kenya. In: Natural conditions and farm management information, vol. II. Nairobi, Kenya: Ministry of Agriculture; 1982. [15] Shepherd KD, Soule MJ. Soil fertility management in west Kenya: dynamic simulation of productivity, profitability and sustainability at different resource endowment levels. Agric Ecosyst Environ 1998;71:131e45. [16] Henry M, Tittonell P, Manlay RJ, Bernoux M, Albrecht A, Vanlauwe B. Biodiversity, carbon stocks and sequestration potential in aboveground biomass in smallholder farming systems of Western Kenya. Agric Ecosyst Environ 2009;129: 238e52. [17] Senelwa K, Sims REH. Tree biomass equations for short rotation eucalyptus grown in New Zealand. Biomass Bioenerg 1998;13:133e40. [18] Akaike H. J Econom 1981;16:3e14. [19] Motulsky H, Cristopoulus A. Fitting models to biological data using linear and nonlinear regression. A practical guide to fitting. SanDiego, CA: GraphPad Software Inc, www. graphpad.com; 2003. [20] Zewdie M, Olsson M, Verwijst T. Above-ground biomass production and allometric relations of Eucalyptus globulus Labill. Coppice plantations along a chronosequence in the central highlands of Ethiopia. Biomass Bioenerg 2009;33: 421e8. [21] Rubilar RA, Allen HL, Alvarez JS, Albaugh TJ, Fox TR, Stape JL. Silvicultural manipulation and site effect on above and belowground biomass equations for young Pinus radiata. Biomass Bioenerg 2010;34:1825e37. [22] Keith H, Barrett D, Keenan R. Review of allometric relationships for estimating woody biomass for New South

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

9

Wales, the Australian Capital Territory. Victoria, Tasmania and South Australia. Canberra, Australia: Australian Greenhouse Office; 2000 August. p. 121, Report no.: 5B. Razakamanarivo RH, Razakavololona A, Razafindrakoto M-A, Vieilledent G, Albrecht A. Below-ground biomass production and allometric relationships of eucalyptus coppice plantation in the central highlands of Madagascar. Biomass Bioenerg 2012;45:1e10. Pe´rez-Cruzado C, Rodrı´guez-Soalleiro R. Improvement in accuracy of aboveground biomass estimation in Eucalyptus nitens plantations: effect of bole sampling intensity and explanatory variables. For Ecol Manage 2011;261:2016e28. Chave ME, Rie B, Ge E. Estimation of biomass in a neotropical forest of French Guiana: spatial and temporal variability. J Trop Ecol 2001;17:79e96. Bastien-Henri S, Park A, Ashton M, Messier C. Biomass distribution among tropical tree species grown under differing regional climates. For Ecol Manage 2010;260:403e10. Ketterings Q, Coe R, Van Noordwijk M, Ambagau Y, Palm C. Reducing uncertainty in the use of allometric biomass equations for predicting above-ground tree biomass in mixed secondary forests. For Ecol Manage 2001;146:199e209. Basuki TM, Van Laake PE, Skidmore AK, Hussin YA. Allometric equations for estimating the above-ground biomass in tropical lowland dipterocarp forests. For Ecol Manage 2009;257:1684e94. Nelson BW, Mesquita R, Pereira LG, De Souza SG, Batista T, Couto B. Allometric regressions for improved estimate of secondary forest biomass in the central Amazon. For Ecol Manage 1999;117:149e67. Montagu K, Duttmer K, Barton CV, Cowie A. Developing general allometric relationships for regional estimates of carbon sequestration e an example using Eucalyptus pilularis from seven contrasting sites. For Ecol Manage 2005; 204:113e27. Mun˜oz F, Rubilar RA, Espinosa M, Cancino J, Toro J, Herrera M. The effect of pruning and thinning on above ground aerial biomass of Eucalyptus nitens (Deane & Maiden) Maiden. For Ecol Manage 2008;255:365e73. Cairns MA, Helmer EH, Baumgardner GA. Root biomass allocation in the world’s upland forests. Oecologia 1997;111: 1e11. Eamus D, Chen X, Kelly G, Hutley L. Root biomass and root fractal analyses of an open eucalyptus forest in a savana of north Australia. Aust J Bot 2002;50:31e41.

Please cite this article in press as: Kuyah S, et al., Allometry and partitioning of above- and below-ground biomass in farmed eucalyptus species dominant in Western Kenyan agricultural landscapes, Biomass and Bioenergy (2013), http://dx.doi.org/ 10.1016/j.biombioe.2013.02.011