• Open Access

Carbon impacts of direct land use change in semiarid woodlands converted to biofuel plantations in India and Brazil


Rob Bailis, tel. +1 203 432 5412, fax +1 203 436 9158, e-mail: robert.bailis@yale.edu


We present an analysis of direct land use change (dLUC) resulting from the conversion of semiarid woodlands in Brazil and India to Jatropha curcas, a perennial biofuel crop. The sites examined include prosopis woodlands, managed for woodfuel production under periodic coppicing, in southern India, and unmanaged caatinga woodlands in the Brazilian state of Minas Gerais. The jatropha plantations under consideration include pruned and unpruned stands and ranged from 2 to 4 years of age. Stocks of carbon in aboveground (AG) pools, including woody biomass, coarse debris, leaf litter, and herbaceous matter, as well as soil organic carbon (SOC) were evaluated. The jatropha plantations store 8–10 tons of carbon per hectare (t C ha−1) in AG biomass and litter when managed with regular pruning in both India and Brazil. Unpruned trees, only examined in Brazil, store less biomass (and carbon), accumulating just 3 t C ha−1 in AG pools. The two woodlands that were replaced with jatropha show substantial differences in carbon pools: prosopis contains ∼11 t C ha−1 in AG stocks of carbon, which was very close to the jatropha stand which replaced it. In contrast, caatinga stores ∼35 t C ha−1 in AG biomass. Moreover, no change in SOC was detected in land that was converted from Prosopis to jatropha. As a result, there is no detectable change in AG carbon stocks at the sites in South India where jatropha replaced prosopis woodlands. In contrast, large losses of AG carbon were detected in Central Brazil where jatropha replaced native caatinga woodlands. These losses represent a carbon debt that would take 10–20 years to repay.


This paper presents an analysis of direct land use change (dLUC) resulting from the conversion of woodlands to biofuel plantations at semiarid sites in both India and Brazil. Each country is currently pursuing biofuel policies that call for rapid expansion of oilseed-based biofuel production in the near future. For example, Brazilian policy calls for a 40% blend of biodiesel by 2035 (Government of Brazil, 2006) and India's biofuel mandate calls for a 20% blend of biodiesel by 2017 (Government of India, 2009). In addition to production for the domestic market, export-oriented production to meet demand in other countries may increase future cultivation. This is particularly true in Brazil, which is already a major exporter of cane-based ethanol and soybeans (a common feedstock for biodiesel production in the US and EU, USDA FAS, 2008) and where future biodiesel export constitutes a major compotabnent of the government's national biofuel policy (see, e.g. Government of Brazil, 2006).

To date, research on carbon balances of jatropha-based biofuel has focused largely on seed production and postharvest processes like oil extraction and conversion to fuel. While land use change (LUC) is acknowledged as having a potentially large impact on life-cycle emissions, there has been little empirical research quantifying the actual effect of land conversion. Instead, LUC research focuses indirect land use change (iLUC), or on dLUC estimated with default values like those published in the IPCCs ‘Guidelines for National Greenhouse Gas Inventories’ (2006, also see Fargione et al., 2008). The range and uncertainty in such default values for both natural vegetation and cropping systems are large and may not reflect actual changes in land cover. In addition, novel biofuel crops, particularly perennial crops like Jatropha curcas (hereafter referred to as jatropha), lack published default values, which impedes any estimation of dLUC impacts.

This paper presents an empirical analysis drawn from two examples of land use change in which semiarid woodlands were cleared for jatropha production. In the following section, we provide a background discussion on jatropha and LUC. The following section reviews site selection and research methods. After that, we present the results of the fieldwork. We conclude by comparing our findings to previous studies and discussing the implications of our findings for biofuel life cycle assessment.



Jatropha is a perennial evergreen shrub native to Central America, which has been naturalized through much of the tropics. Its toxicity, which prevents it from being browsed by livestock, has made it popular as a live fence. Its oil has been used for making soap and for rudimentary lighting since at least the 19th century (FACT, 2009). More recently, jatropha has been popularized as an oilseed crop for biofuel production, and cultivation has proliferated across the tropics as a result (GEXSI, 2008). Jatropha reaches commercial yields in 4–5 years and can grow in a variety of environmental conditions, including regions with poor soils and limited water availability (Achten et al., 2008; Achten et al., 2010). For this reason, jatropha is often promoted as suitable for cultivation on marginal lands, although several accounts note that plant growth suffers in dry and/or nutrient-deficient conditions (Lapola et al., 2009; Behera et al., 2010) and several doubts have been raised about its commercial viability (Biofuels Digest, 2009; Endelevu Energy, 2009).

Land use change and biofuels

LUC impacts may be both direct and indirect (Fargione et al., 2008; Searchinger et al., 2008). Direct impacts on LUC are changes occurring within the boundaries of the production system: for example, the conversion of native vegetation to biofuel production. Indirect changes occur outside the system boundaries, but are attributable to actions inside. For example, if market conditions change as a result of biofuel production, this can induce land conversions in other locations, which may lead to a loss of ecosystem carbon. This loss would also negate some of the benefits of activity taking place within the system boundary (Searchinger et al., 2008). However, iLUC methodologies are uncertain and there is no consensus about whether and how to incorporate iLUC into biofuel GHG accounting (Fearnside et al., 2000; Hohne et al., 2007; Kendall et al., 2009; RSB, 2009).

In one analysis, Lapola et al. (2009) modeled the potential productivity of sugarcane and jatropha in Brazil and India and concluded that spatial variation in the productivity of land in the two countries leads to very different land requirements for both jatropha and sugarcane. Land requirements for jatropha in India could be substantially higher than was previously believed due to low productivity in the absence fertilization and irrigation. In contrast, the researchers find jatropha is more productive in Brazil across a larger extent of the country (Lapola et al., 2009).

One component of LUC assessment that often goes overlooked is the quantity of biomass and carbon present in the production zone after conversion from some prior land use. Generally, biofuel assessments use default values to estimate carbon stocks. However, few analyses have quantified biomass in perennial biofuel cropping systems. One recent study developed allometric equations for jatropha in South Africa using 4-year-old unpruned trees (Ghezehei et al., 2009). The authors concluded that aboveground (AG) biomass could be reliably estimated by either basal diameter or crown depth (Ghezehei et al., 2009). However, although there are often similarities among allometric equations developed for the same species in different sites (Tritton, 1981), the authors stress that their results should not be universally applied. This is particularly true with perennial crops subject to pruning, which is a common practice in commercial jatropha production (FACT, 2009), and was the case in the majority of stands sampled for this study.

Materials and methods

The primary goal of this research is quantify changes in carbon stocks in areas recently planted with jatropha by examining carbon stocks in existing stands and comparing them to adjacent stands of natural vegetation similar to the vegetation present before the establishment of the jatropha. Changes in carbon stocks are assessed in AG biomass and soil organic carbon (SOC). Belowground (BG) biomass was not directly measured. Estimations of BG biomass are possible through root–shoot (R : S) ratios that have been published for various types of vegetation (for example IPCC, 2006). These estimates are associated with large uncertainties; however, the suggested defaults for R : S in natural vegetation are very similar to the measured R : S value available for jatropha (IFEU C, University of Hohenheim, 2008). See the supporting information for additional details.

Site selection

Scoping visits took place during January and March of 2009 and fieldwork was conducted from June to August of 2009.

Indian field site

In India, fieldwork was conducted at Saravana Bioventures, which established a 250 ha jatropha plantation in 2005, about 50 km southwest of Madurai, in Tamil Nadu (Fig. 1a). Soils of Virudhunagar District are mainly black loamy soil (locally known as karisal). The region is characterized by low agricultural productivity, although cotton and some other rain-fed crops are grown (Palanisami et al., 2008). Mean precipitation in the region is 812 mm, coming primarily during the northeast monsoon (Palanisami et al., 2008).

Figure 1.

 (a) Location of study site in Tamil Nadu (43°52.31′W latitude, 15°10.30′S longitude). (b) Location of study site in northern Minas Gerais (43°52.31′W latitude, 15°10.30′S longitude).

Trees were planted 4 years before sampling with 2 × 2 m spacing (∼2500 trees ha−1). Trees were pruned at the knee level (∼40 cm) after 6 months and at hip level (∼80 cm) after 12 months. Depending on the architecture of the plant, some trees may have also been pruned at 18 months around neck level (∼150 cm) to increase branching and maximize florescence. The ground was cleared with a small bulldozer and tilled prior to planting. Both organic and inorganic fertilizers were used.

Land was previously planted with Prosopis juliflora, a drought-tolerant evergreen tree native to South and Central America (Mwangi & Swallow, 2005). Locally, prosopis is used as a source of fuelwood and charcoal. It is a fast growing, nitrogen-fixing tree that is tolerant to arid conditions and saline soils and was introduced in India in the 1960s as part of the Fuelwood Development Program by the Department of Agriculture (Arnold & Jongma, 1977; Foley et al., 1987). A prosopis stand directly adjacent to the jatropha plantation woodlands was sampled to represent the land cover that preceded jatropha.

Brazilian field site

In Brazil, jatropha trees were sampled at Sada Bioenergia, a sugarcane bioethanol company, and part of Sada Transportes Armazenagens S.A., a transportation conglomerate. Sada started growing sugarcane for ethanol in the 1980s, and currently operates an ethanol plant at their headquarters in Jaiba, Minas Gerais (Fig. 1b). They planted jatropha on 187 acres of land roughly 2 years before sampling took place. As in India, land was cleared with a bulldozer and tilled before planting. Lime was applied to reduce soil pH and inorganic fertilizers were applied annually.

The previous land cover at this site was caatinga woodland, which consists of xeric shrubland and thorn forest common across northeastern Brazil, extending northern Minas Gerais. Caatinga is characterized by a wide variation in land-cover types ranging from low scrub to tall dry forest (da Silva, 2009). Soils in this area are latissols and cambissols according to the Brazilian classification, which are oxisols and inceptisols in the US Soil Taxonomy (Palmieri et al., 2003).

A stand of caatinga woodlands close to the jatropha plantation was sampled to represent land cover that preceded the jatropha. The vegetation was very similar to the vegetation that occupied jatropha plots before their establishment. However, the results of the soil analysis demonstrate that soil conditions differ in ways that cannot be attributed to the establishment of the jatropha plantation alone. Thus, comparisons of land cover change are presented in this analysis, but changes in soil at this site are not included (details are discussed further below).

Sampling and analysis

AG biomass in trees and shrubs, coarse woody debris, leaf litter, and herbaceous matter were quantified along with SOC. The following sections discuss specific approaches used in jatropha plantations and prosopis and caatinga woodlands.

Jatropha biomass

To measure jatropha trees, similar procedures were followed in India and Brazil. Sample plots of 20 × 50 m2 were defined. In each plot, 30 trees were randomly selected and diameter at breast height (DBH) and total height were measured (Jatropha has multiple stems, so DBH of each stem was measured individually. This allows a calculation of basal area and an effective tree diameter (d̂) such that d̂inline image where i represents the ith of n stems.). Diameter at ground level (DGL) was also measured. In Brazil, the sample plots consisted of four unpruned plots, two pruned plots. In India, where managers pruned all trees, the sample consisted of five plots. In addition, one tree in each sample plot was destructively sampled and weighed in the field, making a total of six trees in Brazil (four unpruned and two pruned) and five trees in India (all pruned). For destructive sampling, trees were separated into stems, branches, leaves, and fruits, with each component weighed on site. Subsamples of each component were taken to a lab, oven-dried at 70 C for 4 days, and weighed again to determine the moisture content of the downed tree, and to calculate dry AG biomass in the field. Additionally, five trees per plot were defoliated and defruited. The total leaf and fruit weights were taken in the field, and then subsamples were weighed and dried to obtain dry weights. Figure 2a and b shows jatropha in India and Brazil, respectively. The example from Brazil shows a tree undergoing destructive sampling.

Figure 2.

 Photos of sample sites (a) prosopis plot showing thin stems growing under coppice management; (b) caatinga plot showing high density of shrubs and stems; (c) jatropha in India; and (d) Brazilian jatropha tree selected for destructive sampling.

Leaf litter and herbaceous vegetation in each jatropha plot were also measured. In each 20 × 50 m2 sampling are, two randomly defined 5 × 5 m2 subplots were marked. All litter and herbaceous matter within the subplot was collected and weighed onsite. Subsamples of each were taken to a lab, dried, and weighed in the same manner of the tree components described above to determine moisture content and extrapolate the total mass of leaf litter and herbaceous matter in the field.

Prosopis biomass estimates

Prosopis stands in this region of Tamil Nadu are managed by periodic coppicing, which results in numerous thin stems and few branches (see photo in Fig. 2c). To measure standing biomass, five 10 × 10 m2 plots were randomly established. Each plot contained four to 12 multiple stemmed trees. To estimate total AG biomass, one-third of the trees present per plot were chosen randomly to sample. DBH and height was recorded. DBH was estimated by taking a full count of stems, which ranged from 8 to 54 stems per tree (mean=22). The DBH of a random subsample of 10 stems per tree was measured. Biomass per stem was estimated using published allometric equations derived in semiarid tropical conditions; equations and original sources are shown in supporting information. The range of allometric equations includes sites that receive considerably more precipitation than the study sites. In addition, some equations were derived using larger size classes than were measured in this study. One equation calls for the use of basal area (BA), rather than DBH. In that case, BA per stem was calculated and multiplied by the total number of stems to estimate BA/tree. The allometric equations yield a range of estimates for AG biomass of prosopis, which are described in more detail below. In addition, leaf litter and herbaceous matter was sampled, dried, and weighed in a manner similar to jatropha plots described above, but with smaller 1 × 1 m2 subplots. Coarse woody debris (dead and downed trees) was absent from the prosopis stands.

Caatinga biomass estimates

In the caatinga areas, a similar procedure was followed as in India. Five 10 × 10 m2 plots were randomly established. Trees were identified and counted (the species observed and the range of DBH are listed in the supporting information). Stands were unmanaged and relatively dense, with tree-counts ranging from 23 to 59 individual trees in each 100 m2 plot (see photo in Fig. 2d). DBH and height of all trees were recorded. Litter and vegetation samples were weighed, dried, and reweighed to determine dry mass as described previously. In addition, the caatinga plots were the only site with a large pool of coarse woody debris. To quantify this pool of biomass, two 10 m transects were laid along randomly oriented azimuths and all fallen stems encountered were recorded. An estimate of mass (M) of was obtained using the following equation (based on Pearson et al., 2005):


where ρ is the wood density, d is the diameter at breast height, L is the length of the transect.

Wood densities were unknown, so we developed estimates based on a range of recorded densities of caatinga species between 0.3 and 0.8 ton m−3 (Lima & Rodal, 2010).

Biomass stocks in the caatinga areas were estimated using the same allometric equations used for estimating prosopis biomass. In addition, two equations specific to caatinga woodlands were used. These equations were derived by Sempaio & Silva (2005) based on sampling of trees in caatinga woodlands in Bahia state, roughly 650 km northeast of the field site in northern Minas Gerais. Those authors derived three dozen allometric equations applied to both single species and mixed stands. This analysis uses two equations applicable to the caatinga study site: one for three species with small DBH (<25 cm) and one for mixed species of all size classes that also included height as a determinant of biomass (see supporting information for more details).

The range of allometric equations identified includes some sites that receive considerably more precipitation than the study sites. Further, some equations were derived using larger size classes than observed here. We tested the sensitivity of our results by excluding allometric equations from moister sites and equations based on large-diameter trees and found little change. Thus, our results are based on biomass estimates from the full range of allometric equations. This is discussed in more detail below, as well as in the supporting information.


Soil samples were taken at 0–15 and 15–30 cm from the four corners and center of each plot of jatropha and natural vegetation. Samples were analyzed for SOC and other characteristics at the Federal University of Minas Gerais in Montes Claros, Brazil, and at Tamil Nadu Agricultural University in Coimbatore, India [SOC was determined using the Walkley-Black technique in both locations (Schumacher, 2002)].


Biomass in jatropha plots

Results from Brazil indicate that pruning encourages biomass production, particularly in the stems of the tree: mean AG biomass in unpruned trees was 4.7 kg dry matter, whereas the mean AG biomass of pruned trees was 13.6 kg. Jatropha trees in India were all pruned, with mean AG biomass of 7.8 kg dry matter. Biomass in stems, branches and leaves is shown in Table 1.

Table 1.   Biomass in individual jatropha trees determined from destructive samples
 N (trees)Avg. stems per treeStems (kg)Branches (kg)Leaves (kg)Total (kg)Carbon (kg)
  1. Biomass data are given as means with standard deviations in parentheses.

Brazil – unpruned482.1 (0.3)0.8 (0.4)1.0 (1.2)4.7 (2.9)2.3 (1.5)
Brazil – pruned2612.1 (0.1)0.2 (0.1)0.9 (0.4)13.6 (0.2)6.8 (0.1)
India – pruned555.6 (1.2)1.5 (0.8)0.7 (0.3)7.8 (2.2)3.9 (1.1)

The averages shown in Table 1 are used to estimate the biomass and carbon present the stands of jatropha at each study site. These results are shown in Table 2. A previous study, which was also done in India, examined 3.5-year-old unpruned trees planted with 2 × 3 spacing (1667 trees per ha). In that study, the average AG biomass in individual trees was 3.9 kg dry or 6.5 tons of biomass per hectare, which lies between the pruned and unpruned stands analyzed in this study. These results are also included in Table 2 for comparison.

Table 2.   Average biomass and carbon stored in plots of jatropha trees
SiteStand density (no. per ha)Age (years)Avg biomass per tree (kg)Biomass (t ha−1)Carbon (t C ha−1)
  1. Biomass data are given as mean (standard deviation).

Brazil – unpruned jatropha120024.7 (2.9)5.6 (3.5)2.8 (1.7)
Brazil – pruned jatropha1200213.6 (0.4)16.3 (0.5)8.2 (0.2)
India jatropha250047.8 (2.2)19.5 (5.5)9.8 (2.8)
India (Reinhardt et al., 2008)16673.53.9 (NA)6.5 (NA)3.3 (NA)

Unpruned jatropha stands in Brazil contain 2.0 ± 0.5 tons ha−1 of biomass in litter and herbaceous matter combined. Pruned stands contain 3.2 ± 0.2 tons ha−1 of biomass in litter and herbaceous matter. Jatropha stands in India have 0.7 ± 0.3 tons ha−1 litter, but no measurable herbaceous matter. Additional details are given in the supporting information.

Adding the mass of litter and herbaceous matter to the jatropha trees themselves provides an estimate of the total biomass stored in the plantation. Unpruned Brazilian plots store 7.6 ± 3.5 tons ha−1 of biomass in trees, litter, and herbaceous matter combined, while pruned stands store 19.6 ± 0.6 tons ha−1. In the Indian location, biomass per tree is lower than pruned plots in Brazil, but tree density is higher. Those stands store 20.2 ± 5.5 tons ha−1 in trees, litter, and herbaceous matter.

AG biomass in stands of natural vegetation

In both India and Brazil, the allometric equations yield biomass estimates that vary by a factor of 2.5. In the caatinga woodlands, estimated biomass stocks range between 29 and 71 tons ha−1 with a mean and standard deviation of 48.0 ± 12.8 tons ha−1. In prosopis woodlands the range lies between 10 and 25 tons ha−1 with a mean and standard deviation of 16.5 ± 4.2 tons ha−1. As was mentioned above, the allometric equations used to estimate biomass were derived from a range of studies. Some studies utilized trees with larger DBH than were observed at our study sites and some other studies were situated in moister areas. To determine the sensitivity of our estimates of AG biomass to the inclusion of these allometric equations, we calculated mean AG biomass omitting allometric equations that were derived from moist sites and sites with large diameter trees (we follow Brown (FAO, 1997) by defining dry zones as receiving <1500 mm yr−1. To distinguish large and small diameter samples, we consider the range in our own sample, which was 1–6 cm in prosopis stems and 5–45 cm among the caatinga trees). In each case, the mean AG biomass derived using all of the available equations is statistically similar to the mean derived by excluding both moist sites and sites with large-DBH trees (see supporting information). Thus, we use means derived from all available equations. Comparisons of the range of estimates derived from each combination of equations are shown in Fig. 3.

Figure 3.

 Range of AG biomass estimates in caatinga (blue) and prosopis (orange) using different combinations of allometric equations available in published sources. ‘Only dry’ uses equations derived from sites receiving <1500 mm yr−1. ‘Only small DBH’ uses equations derived from sites where trees sampled both overlapped with DBH of trees sampled in this study (1–50 cm).

Coarse debris, litter and herbaceous matter

In stands of natural vegetation, litter, downed trees, and herbaceous matter were also sampled. Dead and downed trees found along random transects in caatinga plots total 7.1 tons of dry matter per ha. There was no analogous material in prosopis woodlands. This is not surprising because those woodlands supply local communities with fuelwood and feedstock for charcoal production: most coarse woody debris is used for that purpose. Litter and coarse woody debris in caatinga and prosopis areas averaged roughly 10.6 and 3.8 tons ha−1, respectively. The herbaceous layer was absent from dense caatinga woodlands, while in prosopis, roughly 1 tons ha−1 was found. Thus, in total, 17.7 tons ha−1 of dead wood and litter was present in caatinga woodlands, and 4.8 tons ha−1 of litter and herbaceous matter was found in prosopis woodlands (see supporting information).

Soil analyses

As stated above, at the Brazilian site, tests revealed that the soil in the two plots differed significantly in structure and other major characteristics (see supporting information for details). Thus, although caatinga vegetation at the two sites was comparable prior to the establishment of jatropha, soil conditions are sufficiently different so that it is not possible to infer a change in SOC. Nevertheless, the baseline data collected during this study will allow us to identify future changes in SOC stocks.

In India, jatropha and prosopis plots were directly adjacent to each other and soil conditions were very similar (supporting information), which allows for a direct comparison. Analysis of soil samples from each site shows SOC in the jatropha plantation was 0.5 tons ha−1 lower than in the prosopis woodland. However, the change is not significantly different from zero. Thus, 4 years after the establishment of the plantation, statistically significant changes in SOC were not detectable with the methods deployed in this study. SOC content in samples from both India and Brazil are shown in Table 3.

Table 3.   SOC in jatropha and woodland vegetation plots in Brazil and India (t C ha−1)
 Brazilian sitesIndian sites
 Pruned JatrophaUnpruned JatrophaAll JatrophaCaatingaJatrophaProsopis
  1. Mean (standard deviation).

0–15 cm20.5 (2.86)17.7 (0.98)18.6 (2.08)38.5 (2.68)13.3 (1.6)13.4 (1.2)
15–30 cm18.1 (2.23)13.3 (2.59)14.9 (3.33)29.3 (4.50)11.1 (2.1)11.6 (1.1)
Total: 0–30 cm38.6 (0.64)31.0 (2.70)33.5 (4.44)67.8 (5.57)24.4 (3.7)24.9 (2.3)


Accounting for each pool of carbon allows for a comparison of net changes in carbon stocks as a result of replacing natural vegetation with jatropha in each location (shown in Fig. 4a and b). Note, Fig. 4a (India) includes SOC and Fig. 4b (Brazil) does not.

Figure 4.

 (a) Stocks and changes in pools of carbon in India (error bars show 1 standard deviation). (b) Stocks and changes in pools of carbon in Brazil (error bars show 1 standard deviation).

In the prosopis woodlands of Tamil Nadu, carbon stocks were relatively low due to periodic coppicing for fuelwood and charcoal production. In addition, no difference was detected in soil carbon. Hence, the total difference between jatropha and the prior land use at this location is statistically indistinguishable from zero (−1.1 ± 3.1 t C ha−1). In contrast, in Brazil, where the caatinga woodlands cleared for jatropha plantations were characterized by higher carbon stocks than prosopis woodlands, there are substantial losses of carbon, ranging from 24.2 ± 7.1 t C ha−1 in pruned jatropha stands to 30.0 ± 7.3 t C ha−1 in unpruned stands. Soils are not included in the Brazilian results. Based on life cycle GHG emissions reported in previous studies of jatropha-based biofuels and the fossil fuel baselines that are replaced, this represents a ‘carbon debt’ that would take roughly 10–20 years to repay (Prueksakorn & Gheewala, 2006; Bailis & Baka, 2010; Hoefnagels et al., 2010).


Four years after the land use change occurred there was no statistically significant change in soil and AG stocks of carbon induced by replacing prosopis woodlands under coppice management with a jatropha plantation in India. However, in Brazil, 2–3 years after replacing native caatinga woodlands with jatropha plantations, a substantial carbon debt exists. This debt varies between pruned and unpruned jatropha sites because pruned stands accumulate biomass more rapidly. The gap may close over time as biomass stocks in both pruned and unpruned plantations increase. Of course, these results represent a single snapshot in a dynamic process. Conditions at either site may change as the jatropha plantations mature.

For example, 4 years after establishment of jatropha in India, we found no statistical difference in SOC. However, stocks of litter had diminished by 2.0 ± 1.4 t C ha−1. Thus, SOC may not remain equal in the long-term because jatropha plots now have fewer inputs of organic matter to the soil. At the Brazilian site, the difference in carbon stocks present in the leaf litter and herbaceous matter between the caatinga site and the jatropha plantation was more pronounced: 4.8 ± 1.8 t C ha−1 in pruned stands and 5.2 ± 1.8 t C ha−1 in unpruned stands. Jatropha plantations also lacked a stock of dead and downed trees present in caatinga woodlands, leading to an additional loss of 3.6 ± 2.4 t C ha−1 in both pruned and unpruned stands. Pools of litter and coarse woody debris can contribute carbon to the soil; if those pools decline, soil carbon may follow suit. Long-term monitoring would be required to quantify changes over the lifetime of the plantations.

Comparison with other analyses

To our knowledge, this is the first study to conduct an empirical analysis of carbon stock changes induced by dLUC. Several LCAs of jatropha have been published with dLUC impacts incorporated; however, these rely on default values to estimate the carbon stocks present in prior land cover and rough estimates of biomass in jatropha stands. Numerous studies have also quantified biomass or carbon stocks in semiarid woodlands. In addition, two studies have quantified biomass in jatropha plantations. In this section, some brief comparisons are made with these previous studies.

Carbon in semiarid woodlands

The IPCC's ‘Guidelines for National Greenhouse Gas Inventories’ (2006) provides regionally specific ranges or point estimates of AG biomass in tropical dry forest and shrubland, which have become popular default values in dLUC analyses. Several studies have quantified AG biomass in the types of woodlands analyzed by this study. For example, Tiessen and colleagues analyzed caatinga in Brazil, where they found AG stocks ranging from 5 to 10 tons (dry biomass) per hectare in open caatinga grassland and 50–100 tons (dry biomass) per hectare in ‘closed forest formations’ (1998). Fearnside (2000) reports roughly 15 tons of AG biomass per hectare in caatinga and a range of 4–92 tons ha−1 in different types of cerrado.

Several studies have also estimated AG biomass in prosopis woodlands. However, it is difficult to draw comparisons because the site examined here is managed under a short coppice rotation, which reduces biomass stocks relative to an unmanaged stand. Nevertheless, the 16.4 ± 4.2 tons of dry matter per ha in living trees falls close to the measurement of prosopis woodlands in other geographically disparate studies. For example, in the Western US, Ansley and colleagues report a prosopis stand with 22.1 tons of dry matter per ha (2010), and note that their findings are similar to studies from Texas and California, which both found roughly 19 tons of dry matter per ha.

In addition, Romijn (2010) investigated dLUC implications of biofuel production in miombo woodlands, which share several characteristics with caatinga. The author averaged results from two previous studies to estimate initial carbon stocks (∼35 t C ha−1). She also accounted for changes in soil carbon by relying on reported values rather than on-site measurements, assuming soils lose ∼32 t C ha−1. These estimates are summarized in Table 4.

Table 4.   IPCC default values for AG biomass and carbon stocks in tropical dry forests and shrublands (IPCC, 2006)
Type of ecosystem or plantationRegionAG biomass (t ha−1)Source
Tropical dry forestAfrica120–130IPCC (2006)
North and South America200–410IPCC (2006)
Asia (continental)100–160IPCC (2006)
Asia (insular)160IPCC (2006)
Tropical shrublandAfrica20–200IPCC (2006)
North and South America40–90IPCC (2006)
Asia (continental)60IPCC (2006)
Asia (insular)70IPCC (2006)
Miombo woodlandEastern/Southern Africa70Romijn (2010)
Caatinga – open grasslandBrazil5–10Tiessen et al. (1998)
Caatinga – closed forestBrazil50–100Tiessen et al. (1998)
Caatinga – unspecified canopyBrazil15Fearnside (2000)
CerradoBrazil4–92Fearnside (2000)
ProsopisSouthwest US19–22Ansley et al. (2010)
India 10-year plantations50–60Saxena (1993)
15-year plantations75–100Saxena (1993)
Alkaline soils3Saxena (1993)
Alkaline soils with amendments34–43Saxena (1993)
JatrophaIndia6.5Reinhardt et al. (2008)
 Tanzania12–40Romijn (2010)
Results from this study
JatrophaBrazil – unpruned5.6 
Brazil – pruned16.3 

Carbon in Jatropha plantations

There are few published estimates of biomass stocks in jatropha plantations. One study, also conducted in India (Reinhardt et al., 2008), destructively sampled 3.5-year-old trees planted with 2 × 3 spacing, and found biomass at the stand level to be ∼6.5 tons of AG dry matter per ha. This estimate, cited by at least one other (Ndong et al., 2009), appears to be the only published estimate of stand-level biomass in jatropha plantations based on empirical measurements of tree biomass. AG biomass in the unpruned Brazilian stand measured in this analysis falls close to the value reported by Reinhardt and colleagues, while pruned stands analyzed here contain considerably more biomass. Romijn (2010) also estimated AG biomass in jatropha; however, she averaged three previous unrelated studies, none of which appear to have made actual measurements of trees (Romijn's estimate is derived by averaging: Struijs' assumption that mature jatropha trees in Tanzania contain 50 kg dry matter (∼40 t C ha−1) (Struijs, 2008);∼40 t C ha−1 from irrigated plots in Egypt (Henning, 2003); and 12.4 t C ha−1 in nonirrigated plots in India (Francis et al., 2005). The latter value is close to the AG stocks measured in young pruned plots in this study). Jatropha estimates are also summarized in Table 4.

Implications for biofuel LCA

Changes in carbon stocks associated with dLUC in biofuel plantations form a critical component of ‘cradle-to-grave’ biofuel life cycle assessments. Many LCAs fail to consider dLUC and those that do consider it typically rely on default values (Dehue & Hettinga, 2008; Reinhardt et al., 2008; Ndong et al., 2009; Bailis & Baka, 2010). This analysis shows the risks of relying on default values to estimate changes in carbon stocks. By comparing dLUC induced through the conversion of two semiarid woodlands into biofuel plantations, we observe substantially different degrees of carbon debt. Relying on the default assumption for the Brazilian case would result in a reasonable estimate of dLUC, while relying on the default for the Indian case would overestimate the carbon debt by roughly 19 ± 5 t C ha−1.

Published biofuel LCAs have shown that default values for GHG emissions from dLUC can negate the GHG emission reductions achieved by replacing fossil fuels with biofuels (Reinhardt et al., 2007; Dehue & Hettinga, 2008; Bailis & Baka, 2010). However, as this analysis shows, dLUC resulting from the conversion of woodlands with low initial stocks of carbon is substantially smaller and has little or no impact on lifecycle GHG emissions.

We must also stress that the plantations examined in both India and Brazil have yet to reach maturity. Trees will continue to add biomass for several years, though this may be reduced by periodic pruning in order to maintain a tree that can be harvested manually. Thus, the data present a snapshot of a dynamic process that must be further elaborated with future research. Nevertheless, the data already reveal a greater complexity than the use of default values can accommodate. We suggest that a system of long-term monitoring is needed to better understand these dynamics, particularly in perennial systems that receive less attention in existing literature than biofuel production using annual crops.

Finally, by focusing on carbon dynamics we acknowledge that there is a risk of overlooking other impacts resulting from a change in land cover from woodlands to perennial biofuel plantations. This includes changes in ecosystem function, biodiversity, and access to woodland resources for nearby communities. Each of these impacts requires study and must be explicitly acknowledged as society weighs the tradeoffs necessary to grow biofuel feedstocks.


This work was supported by a research grant from The Boeing Corporation acting in partnership with the Sustainable Aviation Fuel Users Group (SAFUG). One author (HM) also received valuable support from the Tropical Resources Institute at Yale School of Forestry and Environmental Studies. The authors are grateful for valuable assistance from Jenn Baka who helped set up field research in both locations. We thank our field assistants, Monica Ribas and Anderson Evaristo in Brazil and D. Ramesh and Ayyaner in India. In addition, the authors acknowledge the Sada Bioenergia and Saravana Bioventures for making their plots and areas of natural vegetation available for study. In Brazil, Carlos Alberto and Luciano Cardoso made important introductions, and Rodrigo Meirelles de Azevedo Pimental of EPAMIG provided access to laboratory equipment for drying and weighing samples. Much of this analysis was done while the corresponding author (RB) was in residence at the Copernicus Institute for Sustainable Development and Innovation at Utrecht University and benefited from numerous interactions there. All errors and omissions are the sole responsibility of the authors.