Genetic parameters of growth and biomass in Leucaena leucocephala for wood energy

Leucaena leucocephala is a potential species for wood-energy production in Indonesia. A study of genetic improvement was initiated with a progeny test of 80 lines from 10 seed sources. Plant height and stem diameter were measured at 6 and 18 months, growth index (GI) calculated at 18 months and wood biomass production measured at 30 months. Differences between seed sources for height and diameter were observed at 6 months but not at 18 months. Significant differences between lines within seed source were observed for height and diameter at 6 months and these differences remained at 18 months. Differences between lines for GI and biomass were significant at 18 and 30 months respectively. At 18 months, line mean heritability for height, diameter and GI were estimated to be moderate, namely 0.62, 0.61 and 0.62 respectively. At 30 months line mean heritability for biomass was moderately low (0.39). Genetic correlations between height and diameter were moderately high at 6 months (0.74) and increased at 18 months (0.82), while correlation between diameters at 6 and 18 months was high (0.93). The expected genetic gain from selecting the 25 and 10 best lines with a high line value for GI was 33.7% and 48.8% respectively, with lines from the Indonesian local seed sources found to be the best performers. Three lines from the newly introduced Tarramba cultivar also had good performance. These results are discussed in relation to the future improvement program of this species for wood energy production.


Introduction
Leucaena leucocephala, a multipurpose legume tree, has been used as fuel wood for decades in Indonesia, due to its high wood energy quality reaching 4,700 cal/g, with a wood density of 0.67 g/cm and lignin content of 31.6% (Toruan-Mathius et al. 1994). It has a combustion index higher than coal (Shrestha et al. 2015). It also has potential for producing charcoal, briquette and wood pellets (Acda and Devera 2014), producing abundant fast regrowth after cutting the main stem, so that annual wood energy can be harvested without replanting for up to 15-20 years (Pagad 2010). Under an annual precipitation of 1200-1300 mm, it was reported to produce 38.8 t/ha of wood in the first year with a calorific value of 4,700 kcal/kg and 5 t/ha of branches and 6.9 t/ha of leaves when harvested at 50 cm above ground level (Rengsirikul et al. 2011). This species thus provides a profitable option for supplying high quality energy from productive wood biomass.
Natural forest is not able to efficiently capture carbon due to reaching growth maturity. However, fast growing L. leucocephala potentially absorbs much more carbon due to its continuing active growth under routine annual harvests. L. leucocephala can act as a green manure, producing nitrogen from the rhizobia in its roots and through leaf drop (Ceccon et al. 2015). It grows well in arid environments, producing roots up to 5 m depth for water access (Brewbaker et al. 1972). Under an annual rainfall of ~900 mm, it produced a litter fall of more than 10 t/ha/ yr which is easily decomposed (Ceccon et al. 2015). The species has the capacity to fix nitrogen amounting to 250 kg/ha/yr (Casanova-Lugo et al. 2014).
Increasing interest in wood biomass as a renewable energy source is due to several advantages such as lower moisture content, higher lignin content, easy handling and storage, and lower ash and nitrogen content compared to other types of biomass (Sims et al. 2006). Compared to other legume species commonly used for fuel wood, L. leucocephala has a higher lignin content (31-33%) than Senna siamea (21-22%) and Gliricidia sepium (26-27%) (Mainoo and Appiah 1996). Higher lignin content boosts the heating value (Günther et al. 2012) making it more efficient in combustion than coal (Demirbas and Demirbas 2009).
The increased interest in growing L. leucocephala for wood energy has led to the initiation of a genetic improvement program of the species in Indonesia, with the objective to identify the best wood biomass alternative for more sustainable energy. Information on the genetic parameters of L. leucocephala is very limited in Indonesia.
Research is needed on assessing the genetic variation of populations and lines and identifying best lines, followed by producing improved seed by converting the better lines from the progeny test to a seed orchard. Genetic parameters such as heritability, line value and genetic gain should be determined and used to formulate efficient selection strategies within the breeding process. Unlike other Leucaena species which are self-incompatible and highly cross-pollinated, L. leucocephala is predominantly self-pollinated (Brewbaker and Styles 1982). This paper reports variation and genetic growth parameters, following progeny testing at the ages of 6 and 18 months and biomass yield measurements at 30 months, which can be used for selecting the better lines for biomass production before individual selection on wood quality for energy is undertaken.

Plant material
Seeds were collected from 80 parent trees located in 10 regions in Indonesia where L. leucocephala is grown (Table 1). Except for Fatuleu, East Nusa Tenggara, the original accession of all seed sources and the number of original trees that contributed seed to these seed sources were unknown, but likely to be cultivars of K8 (Peru) and K28 (Cunningham), which were introduced to Indonesia in the early 1970s (Toruan-Mathius et al. 1994). The Fatuleu seed source was established using Tarramba, a cultivar bred in Hawaii from accession K636 collected from Mexico (Nulik et al. 2013), with the seeds introduced from Australia. Parent trees were selected phenotypically from the populations based on their growth performance: height, diameter and health.

Trial Establishment
Seeds of each line were soaked separately in hot water (90 ⁰C) for 5 minutes, drained and replaced with cold Leucaena leucocephala for wood energy water and left for 12 hours. The seeds were then sown in a separate germination box for each line. Two to three weeks after sowing, germinated seeds were transplanted into separate 15x10 cm plastic bags previously filled with media containing a 1:2 mixture of topsoil and compost. After about 4 months, seedlings were large enough to be transplanted into the field.
The progeny test was established in January 2018 in Brebes, Central Java, Indonesia (245 masl). The mean annual rainfall was 1960 mm. The average daily temperature was 29 ⁰C with a mean minimum of 24 ⁰C and maximum of 35 ⁰C. The soil was a Grumosol or Vertisol (Soil Survey Staff 1992). The trial was laid out in a randomized complete block design, consisting of the 80 lines in four-noncontiguous plots, replicated 4 times. Tree planting spacing was 2 x 2 m within plots.
At 6 and 18 months, height and stem diameter (at 20 cm above ground at 6 months and at 130 cm above ground at 18 months) were recorded. Thinning by cutting the whole tree at 5 cm above the ground was conducted after the second measurement at age 18 months by removing the two poorest trees of every line in each block. Thinning was intended to provide wider spacing for the remaining trees to promote early flowering and seed production as the progeny is progressively converted to a seed orchard (Wheeler 1991).
At 30 months, another tree in each plot (leaving one best individual per line to grow in the field) was cut at 100 cm above the ground and separated into stem, branches and twigs and then the components weighed. Leaf was excluded in this study because it has no potential for energy production. Samples of the woody biomass were taken and dried at 105 ⁰C to constant weight to determine total dry weight of biomass. Growth index (GI) was estimated at 18 months using a formula for multipurpose tree species as GI = basal diameter 2 × height (Mullen and Gutteridge 2002).

Data Analyses
Individual tree data were analyzed using the General Linear Model procedure with the following linear model: Sj effect of jth seed source; L(S) jk effect of jkline within jth seed source; B i F(S) jk interaction effect of ith block and jk line; and ε ijk the residual error.
Block was considered fixed while seed source and line within seed source were considered to be random. Variance components were estimated for the random effect by using the Restricted Maximum Likelihood (REML) procedure (Williams et al. 2002). Line means were predicted with the Best Linear Unbiased Prediction (BULP). The use of mixed model procedures (REML and BLUP) was appropriate where data were unbalanced such as found in the progeny test reported here. Data analyses were performed with the software R ver. 4.02.
As L. leucocephala is a self-pollinated species where genetic variation within lines is considered homogeneous, similar to a clone, broad-sense heritability is more relevant to quantify the genetic control of particular traits in the species. The line mean heritability was estimated using the following formula: H 2 mean = σ 2 l / (σ 2 l + σ 2 bl /b + σ 2 e /nb) where: σ 2 l is component of variance due to line; σ 2 bl is the component of variance due to block × line interactions; σ 2 e residual error; b the harmonic mean number of blocks per line; and n the harmonic mean number of trees per line.
The variance component of block (σ 2 b ) is not included in the denominator of the formulae of heritability, implying that the estimated heritabilities are appropriate to selection on block adjusted data. The Genetic Coefficient of Variation (GCV) of measured traits was calculated as follows: GCV (%) = √σ 2 l ) / X × 100% where X is the general mean. Genetic correlation (r g ) between traits, as well as ageage genetic correlation of the same trait were calculated according to Williams et al. (2002): σ l (x,y) = ½ (σ 2 l(x+y) -σ 2 l(x) -σ 2 l(y) ) where σ 2 l(x+y) is the variance component for the sum of the traits. All of the variance components were estimated as previously described.
Genetic gain (G) for line selection was predicted by using the following formula: G = H 2 mean × i × σ p where: H 2 mean = line mean heritability, i = selection intensity, σ p = phenotypic standard deviation.

Results
At 6 months, differences between seed sources (populations) were highly significant for height (p=0.016) and stem diameter (p=0.002). The seed source mean for height ranged from 1.57 to 2.09 m, while that for stem diameter ranged from 1.13 to 1.56 cm. However, at 18 months the differences between seed sources were not significant for height (p=0.37) and stem diameter (p=0.061). The seed source mean for height ranged from 5.17 to 5.72 m and stem diameter ranged from 4.30 to 5.01 cm. At 18 months, the growth index (GI) between seed sources also did not differ significantly (p=0.103) ( Table 2).
At 6 months, differences between lines within seed source were significant for height (p <0.0001) and diameter (p=0.027). Differences between lines within seed source also remained significant for height (p <0.0001) and diameter (p <0.0001) at 18 months. GI between lines also differed significantly (p<0.0001). GCV increased with increasing age, for example GCV for height and diameter at 6 months were 6.5 and 4.8% respectively and the corresponding figures at age 18 months were 7.1 and 8.2% respectively. The estimates of line mean heritability at 18 months for height (0.62) and diameter (0.61) and GI (0.62) were all moderate, while that of biomass at 30 months was low (0.39) ( Table 3).
At 6 months the genetic correlation (rG) between height and diameter was moderately high and positive (0.74), and it increased at 18 months to 0.82. The genetic correlation between 6 and 18 months for height was estimated to be low but positive (0.38); it was very strong and positive (0.93) for diameter. Genetic gain expected from selecting the best 25 lines with high GI was 33.7 %, while selecting the best 10 lines for GI resulted in an expected genetic improvement of 48.8%. The best selected lines based on their line value for growth index (GI) are listed in Table 4.

Discussion
Results of the current study revealed that the best seed source based on GI was Majalengka (West Java), followed by Pamogan (Bali) and Fatuleu (East Nusa Tenggara) ( Table 4). In a previous study, comparison of 9 Indonesian seed sources with cv Tarramba collected from Fatuleu East Nusa Tenggara, indicated that Tarramba performed better in the nursery (Hendrati and Hidayati 2018). After 6 months in the field, the best seed source for height and diameter was found to be the Tarramba cultivar (Hendrati and Nurrohmah 2019). Tarramba is a cultivar released in Australia but selected at the University of Hawaii. It is outstanding for leaf biomass and was introduced to Kupang (Nulik et al. 2013) for cattle fattening in Amarasi, East Nusa Tenggara. This cultivar is said to be tolerant to the psyllid insect (CABI 2017), although not resistant, and better in producing wood for energy compared to the previous outstanding cultivars Peru and Cunningham (Rengsirikul et al. 2011). However, after 18 months in the field, differences among seed sources disappeared for all traits. Nevertheless, the best three seed sources were identified to be from Majalengka (West Java) followed by Pamogan (Bali) and Fatuleu (E. Nusa Tenggara) in terms of GI. Except for the Fatuleu seed source where the genetic material is L. leucocephala cv. Tarramba, the introduction of L. leucocephala ssp. glabrata to Indonesia was poorly documented and earlier introductions presumably originated from K8 and K28 varieties (Toruan-Mathius et al. 1994). It is possible that the genetic material of the remaining 9 seed sources were interrelated and the genetic differentiation between seed sources or populations has not yet occurred. Therefore, selection to identify best seed sources for growth and biomass is not yet feasible. At the line level, however, variation existed at 6 months and remained significant after 18 months in the field and generated moderate broad-sense heritability for growth (Table 3), suggesting that selection to identify the best lines is possible. Similar results were reported from a trial of L. leucocephala at 9, 12 and 15 months after planting in Tamil Nadu, India which found that broadsense heritability estimates for growth were moderate: 0.55, 0.46, and 0.60 respectively for height, diameter and GI (Sangram and Keerthika 2013). In self-pollinated species, broad-sense heritability based on line mean is mostly used for estimating genetic gain from selection, while broad-sense heritability on an individual tree can be used in marker-assisted selection (Xu et al. 2009).
Genetic correlations between height and diameter were moderately high and positive at 6 months (0.74) and even increased at 18 months (0.82), suggesting that selection on height or diameter will not have a substantial effect on the other trait. The very high positive genetic relationship between 6 and 18 months for diameter (0.93) demonstrates that diameter growth at early ages is a good indicator for future performance.
The expected genetic gain from selecting the 10 best lines is substantial. However, the genetic variation in growth of selected lines will obviously decline as number of lines is reduced compared with the initial number of lines. It is worth mentioning that the genetic parameters reported in this study might be overestimated if the linesite interactions of the measured traits were of significant importance as the progeny test was only conducted at one site. For seed production, the reduction in genetic variation should be acceptable considering the species is predominantly self-pollinated. Seed will be produced by selfing in each line without causing inbreeding depression. For future genetic improvement programs, however, selection within pure lines cannot be expected to yield rapid genetic advance. The best selected lines may be used for controlled crosses to produce F 1 progenies and then to self them to generate suitable F 2 genetic variation for the next cycle of selection. This is one of the possible breeding strategies that may be adopted and which is usually carried out in self-pollinated species and is quite common in agricultural crops (Brown et al. 2014). Controlled crosses by introducing genetic material of L. leucocephala from new accessions in addition to crossing the best selected lines may also be used. Developing Leucaena hybrids having hybrid vigor for a specific trait of interest has been done previously (Dalzell 2019).
It is interesting to note that lines from the local seed sources of West Java and Bali had high line (genetic) value for GI. The best lines (Table 4) are all L. leucocephala ssp. glabrata and three lines of Tarramba cultivar (rank 3, 4 and 19) collected from Fatuleu (East Nusa Tenggara) seed source. This inferred that in terms of wood biomass, lines from local seed sources found in Indonesia are very promising, showing good performance and worth further examination for their wood-energy quality.