• Open Access

Comparative life cycle assessment of centralized and distributed biomass processing systems combined with mixed feedstock landscapes



    1. Biomass Conversion Research Laboratory, Great Lakes Bioenergy Research Center, Department of Chemical Engineering and Materials Science, Michigan State University, 3815 Technology Blvd, Lansing, MI 48910, USA
    Search for more papers by this author

    1. Biomass Conversion Research Laboratory, Great Lakes Bioenergy Research Center, Department of Chemical Engineering and Materials Science, Michigan State University, 3815 Technology Blvd, Lansing, MI 48910, USA
    Search for more papers by this author

Pragnya L. Eranki, fax 517 336 4615, e-mail: erankipr@egr.msu.edu


Lignocellulosic biofuels can help fulfill escalating demands for liquid fuels and mitigate the environmental impacts of petroleum-derived fuels. Two key factors in the successful large-scale production of lignocellulosic biofuels are pretreatment (in biological conversion processes) and a consistent supply of feedstock. Cellulosic biomass tends to be bulky and difficult to handle, thereby exacerbating feedstock supply challenges. Currently, large biorefineries face many logistical problems because they are fully integrated, centralized facilities in which all units of the conversion process are present in a single location. The drawbacks of fully integrated biorefineries can potentially be dealt by a network of distributed processing facilities called ‘Regional Biomass Processing Depots’ (RBPDs) which procure, preprocess/pretreat, densify and deliver feedstock to the biorefinery and return by-products such as animal feed to end users. The primary objective of this study is to perform a comparative life cycle assessment (LCA) of distributed and centralized biomass processing systems. Additionally, we assess the effect that apportioning land area to different feedstocks within a landscape has on the energy yields and environmental impacts of the overall systems. To accomplish these objectives, we conducted comparative LCAs of distributed and centralized processing systems combined with farm-scale landscapes of varying acreages allocated to a ‘corn-system’ consisting of corn grain, stover and rye (grown as a winter double crop) and two perennial grasses, switchgrass and miscanthus. The distributed processing system yields practically the same total energy and generates 3.7% lower greenhouse gas emissions than the centralized system. Sensitivity analyses identified perennial grass yields, biomass densification and its corresponding energy requirements, transport energy requirements and carbon sequestration credits for conversion from annual to perennial crops as key parameters that significantly affect the overall results.


Lignocellulosic biofuels are an environmentally superior alternative to petroleum-derived fuels. Abundant and renewable cellulosic feedstocks provide important solutions to fulfill escalating demands for alternate liquid fuels. However, their highly complex physical structure impedes conversion into useful end-products when using biological conversion routes. As a result, pretreatment forms the core of biomass conversion processes (McMillan, 1994). Biorefining is the integrated, industrial-scale production of fuels and higher value products from biomass feedstock, similar to the petroleum refining approach (Kimes, 2007). A centralized, fully integrated biorefinery includes all biomass conversion processes (i.e. size reduction, pretreatment, hydrolysis, fermentation, distillation) in a single location. Production of large quantities of biofuels at optimal scales for efficient capital investment requires biorefineries handling enormous tonnages, probably of mixed feedstocks. This implies contracting with thousands of individual farmers, potentially interrupted feedstock supplies (due to drought, etc.), large transport and storage costs of feedstock and other business and market power issues (Carolan et al., 2007).

This gap between feedstock suppliers and biorefineries can be bridged by a network of smaller scale preprocessing facilities called ‘Regional Biomass Processing Depots’ (RBPDs), or just ‘depots’ in this paper. These strategically distributed depots interact with farms producing feedstock and with animal production operations as well as with the biorefinery and power plants (Carolan et al., 2007). RBPDs can potentially provide benefits in environmental, economic and social sustainability. In one simple configuration, a depot procures, preprocesses/pretreats, densifies and delivers feedstock to the biorefinery and also returns animal feed to farms.

This study focuses on this simple depot configuration consisting of feedstock pretreatment for bioethanol production and return of a single by-product (animal feed) to farms. Biomass is procured from farms and undergoes conditioning and size reduction. Using the Ammonia Fiber Expansion Process (AFEX) (Balan et al., 2006; Sendich et al., 2008) in the processing depots offers multiple advantages as described in the literature (Carolan et al., 2007). Transporting the pretreated solids over long distances requires densification to reduce transport costs and its associated environmental impacts as well as to facilitate handling of pretreated biomass. Therefore, a densification step such as pelletization is imperative in the depots. AFEX pretreatment before densification can improve the binding properties of lignocelluloses and enhance pellet characteristics, thereby providing stability during storage and transport (Sokhansanj et al., 2005; Dale, 2009). Part of the pretreated solids is used as animal feed (Sendich & Dale, 2009). A block diagram of this proposed RBPD is shown in Fig. 1. Depots can also be configured to accommodate multiple technologies such as leaf protein concentrate production, thermochemical conversion and stem–leaf separation to deliver additional valuable by-products to end users. For example, if wet biomass is acquired from farms it can be pulped and pressed to extract protein concentrate (Enochian et al., 1980), before being sent to the pretreatment reactor.

Figure 1.

 Block diagram of a simple Regional Biomass Processing Depot (RBPD) used in this study. Lignocellulosic feedstock is procured from the landscape, pretreated, densified and delivered to the biorefinery and a portion of pretreated feedstock is returned to the landscape as animal feed. All processes confined within the dashed line represent operations present in the RBPD. Other technologies such as leaf protein concentrate (LPC) production, stem–leaf separation and anaerobic digestion might also be added to the depot.

The primary objective of this study is to perform a comparative life cycle assessment (LCA) of distributed and centralized processing systems. Additionally, we assess the effect of apportioning land area to different feedstocks within a landscape on the net energy yields and greenhouse gas emissions of the combined landscape-processing systems.

Materials and methods

Terminology and calculations

1. Net energy yield (NEY) is calculated as the difference between total energy outputs from and inputs to the cropping, transport and processing systems. All inputs are considered only in terms of nonrenewable fossil fuel energy used. Energy embodied in the feedstock is included in the outputs only in the form of ethanol and electricity generated.

2. Net carbon emissions reduction (NCER) is calculated as the difference between CO2 equivalent greenhouse gas emission outputs from the transport and processing systems and the carbon sequestration effects of the agricultural systems. In this study, the carbon sequestration effect of the agricultural system is essentially the net carbon (kg CO2 equivalent) resulting from gasoline displacement by ethanol plus the soil organic matter (SOM) sequestration due to crop residue and no-till practices and carbon emissions during cultivation and harvest plus annual SOM losses (based on an yearly SOM maintenance parameter for different farm location). All SOM values are converted to soil organic carbon (SOC).

3. Calculation of relative differences of NEY and NCER between distributed and centralized processing systems:

  • (i)% difference in NEY (%ΔNEY)=[((NEYcent−NEYdist)/NEYcent) × 100].
  • (ii)% difference in NCER (%ΔNCER)=[((NCERcent−NCERdist)/NCERcent) × 100].

Where cent represents the centralized processing system and dist represents the distributed processing system.

All values are in terms of dry tons wherever applicable. Coproduct credit calculations, explained further in the following section, show the conversion of nonenergy coproducts into energy and emission-compatible values which are subsequently included in the NEY and NCER calculations, respectively.

Landscape analysis and LCA description

All energy and emission parameters are calculated on the basis of 1 kg dry feedstock. The functional unit is a 5000 tons day−1 (TPD) biorefinery and the systems for comparative LCA are:

  • System 1: Nine RBPDs, 500 TPD each (with pelletization)+a 5000 TPD biorefinery that contains a single (10th) 500 TPD RBPD (with no pelletization).
  • System 2: Centralized 5000 TPD biorefinery (no pelletization).

Figures 2a and b show the system boundaries of both systems under consideration for LCA.

Figure 2.

 System boundaries for comparative lifecycle assessment (LCA). (a) Distributed processing system, (b) Centralized processing system. Both processing systems are combined with the mixed feedstock landscapes and animal operations for an integrated system-wide analysis.

The feedstocks included in this study are a continuous ‘corn system’ consisting of corn grain, stover and a winter double crop (rye in this case) as well as two perennial grasses – switchgrass and miscanthus. Corn stover removal rates of 70% are based on literature values (Graham et al., 2007). Both perennial grasses are assumed to have average stand lives of 10 years (Scurlock, 1999; Duffy & Nanhou, 2002). Stover is one of the most abundant lignocellulosic resource available and is widely considered to be a primary feedstock for cellulosic biofuels (Aden et al., 2002; Hettenhaus, 2005). Switchgrass and miscanthus have also attracted interest as potentially important feedstocks for biofuel production (Lewandowski et al., 2003; McLaren, 2005; McLaughlin & Adams Kszos, 2005; Murnen et al., 2007). Part of the pretreated perennial grasses and stover is also used as animal feed.

A total landscape area of 280 ha (∼700 acres) (in Barry County, SW Michigan) was selected using the Web Soil Survey (WSS) [a Geographic Information System tool provided online by the United Stated Department of Agriculture Natural Resources Conservation Service (USDA NRCS)]. Although the biorefinery size is fixed at 5000 TPD, for purposes of comparison only a fraction of the biomass generated from the land area under consideration is investigated here. This area is considered as a representative fragment of a larger landscape that would be required to satisfy biorefinery feedstock requirements. It is intended to compare the energy yield and greenhouse gas emissions for this fixed land area but with different acreage distributions among the primary crops of interest. To achieve this objective, different configurations with varying acreages allotted to different feedstocks are combined individually with each processing system and these landscape-processing systems are then compared with each other. The amount of biofuel (bioethanol in this case) and electricity generated from each of these configurations is the same irrespective of the type of feedstock used because of the unchanging scale of biorefinery; therefore they can be compared without ambiguity. Similarly, the basis of 1 kg dry feedstock was chosen since one of the two most important aspects of this study is feedstock allocation within a given landscape. This allows for fair comparison because the land area used remains constant for all configurations.

For this fixed land area, we formulated three configurations to evaluate the effect of decreased acreage in the corn system and increased perennial grasses. It is assumed that this fixed land area is a ‘clean-slate’ where any crops grown on this land area before this analysis are ignored. This assumption is valid based on the impetus that the use of marginal lands to grow cellulosic biofuel feedstocks is gaining (Hill et al., 2006; Fargione et al., 2008). We define marginal land as any land not being used to grow commercial or conventional crops or abandoned and degraded cropland (Dale et al., 2010b) that may be capable of growing low-maintenance biomass feedstocks such as perennial grasses (Schmer et al., 2008). In Configuration 1, 65% of the acreage was dedicated to the corn system and the remaining acreage was divided equally between the two perennial grasses. In Configuration 2, only 15% of the acreage was allotted to the corn system and the remaining 85% was divided equally between the grasses. In Configuration 3, all the acreage was divided equally between these two perennial grasses. There is a subdivision within each configuration acting as an embedded sensitivity analysis to assess the affect of grass yields on the results. A high grass (HG) yield of 10 tons ac−1 (24.7 tons ha−1) yield was chosen for switchgrass and 12 tons ac−1 (29.6 tons ha−1) for miscanthus while a lower grass (LG) yield of 7 tons ac−1 (17.3 tons ha−1) yield was chosen for switchgrass and 8 tons ac−1 (19.8 tons ha−1) for miscanthus. The high and low yields were selected based on literature values from various publications for switchgrass (Powlson et al., 2005; Goddard, 2007) and for miscanthus (Gibson, 2007; Thelen et al., 2009). The perennial grasses were assumed to have an ‘establishment period’ of 3 years in which they have negligible yields but postestablishment they have significantly higher yields and much lower maintenance requirements than the annual crops. Yields are assumed to be on a dry mass basis and with moisture contents of 15% at harvest (Scurlock, 1999; Sokhansanj et al., 2009). Yields of both perennial grasses are averaged over their entire stand-life.

Yields for the corn system were obtained from the WSS for available crops or estimated for unavailable crops based on the yields of similar crops present in the area of interest. For example, winter rye yields were calculated on a weight basis compared with winter wheat and as mentioned previously perennial grass yields were obtained from literature. After obtaining yields crop cultivation and harvest energy and emission values were estimated using crop budgeting spreadsheets (K. Thelen, 2010, personal communication). An extensive literature review was conducted for each crop to obtain inputs including fertilizers, insecticides, fuel used in cultivation and harvest, seeds required or roots transplanted (data available in Supporting Information). The budget spreadsheets provide details such as increases in SOM when using no-till practices and losses in SOM based on farm locations. Figure 3 shows fractions of biomass yields of each crop in each configuration for the fixed land area. The corn system contributes the greatest biomass in Configuration 1 as expected since it occupies the largest portion of land. Moreover, its biomass fraction increases relatively in LG yield scenarios compared with that in HG yield scenarios in all the applicable configurations. Similarly, the fraction of biomass from perennial grasses increases with an increase in their land area allowance. On-farm animals form an integral part of these landscape analyses. It is assumed that ruminant animals are present within the landscape at a stocking rate of two animals per acre (Sendich, 2008).These ruminant animals consume part of the pretreated feedstock and their methane emissions are calculated, converted to CO2 equivalent and included in the NCE impacts.

Figure 3.

 Fractional yields of feedstocks in different configurations. The land area remains the same in each configuration but feedstock acreages vary. Yields of cropping systems vary proportionally with their acreages in each configuration. Corn system acreages as percentage of total land area: Configuration 1=65%, Configuration 2=15%, Configuration 3=0. Combined perennial grass acreage Configuration 1=35%, Configuration 2=85%, Configuration 3=100%, divided equally between switchgrass and miscanthus. HG represents high yields and LG represents low yields of the two perennial grasses.

Animal feed is an important by-product of the AFEX pretreatment method (Carolan et al., 2007). Therefore, coproduct credit calculations for animal feed are an important aspect of this LCA. Here it is necessary to calculate displacement ratios for all the lignocellulosic feedstocks included in this study. The displacement ratio is defined as the amount of conventional animal feeds (corn and soybean meal) that pretreated lignocellulosic feedstock can replace based on animal nutritional requirements. These calculations were performed for stover, switchgrass and miscanthus based on the amount of energy, protein and fiber (in case of grasses replacing hay) replaced by lignocellulosic feedstock compared with the conventional feedstock. The following illustration shows displacement ratio calculation for stover.

Values for digestible fiber, energy and protein for both replacing and replaced animal feeds are shown in Table 1 (B. D. Bals, 2010, personal communication).

Table 1.   Nutritional value of feedstocks
Nutritional value
CornStoverSoybeanSwitchgrassGrass hay
  1. These values are used in displacement ratio calculations.

  2. In this analysis, a ‘displacement ratio’ is the mass of conventional animal feeds that unpretreated perennial grasses can replace based on the nutritional value of crops being replaced. Displacement ratios are used in coproduct credit calculations. Stover is assumed to displace corn and soybean meal whereas the perennial grasses are assumed to displace corn, soybean meal and fiber from grass hay.

Protein0.0940.172 (including nonprotein nitrogen)0.3860.014

Oil content of soybean=0.196, meal=1–0.196=0.804.

Based on equalizing nutritional values:

For corn stover displacing corn and soybean meal:


Where C and S represent corn and soybean meal, respectively

Solving for C, S:


Similar calculations were performed for perennial grasses. Owing to absence of nutritional value data for miscanthus its displacement ratios are assumed to be the same as switchgrass. Displacement ratios were then incorporated in equations [adopted from literature (Edwards & Anex, 2009) and modified for this study] for coproduct credit calculations. The following illustration shows coproduct credit calculation for stover:

Corn stover energy credit for displacing corn (MJ)=[animal feed production (kg feed kg−1 stover) × feed displaced (kg corn displaced kg−1 animal feed)] × corn production energy (MJ kg−1 corn).

Corn stover energy credit for displacing soybean (MJ)=[animal feed production (kg feed kg−1 stover) × feed displaced (kg soybean displaced kg−1 animal feed)] × soybean production energy (MJ kg−1 soybean).

Assumed animal feed production=1 kg 1 kg−1 of stover, energy inputs for corn production is obtained from literature review and displacement ratios (for feed displaced) are calculated as stated previously. Similarly energy and emission credits were obtained for each lignocellulosic feedstock and included in NEY and NCER calculations.

Displacement ratios are based on direct substitution of un-pretreated feedstock for lack of actual data from animal feed trials (i.e. feeding animals with pretreated lignocellulosic feedstock used in this study). However, AFEX-treated rice straw fed to dairy cows has shown higher neutral detergent fiber intake and milk yield (Weimer et al., 2003). Also, preliminary analysis has suggested that 100% of beef cattle nutritional requirements and up to 70% of dairy cattle nutritional requirements (along with a protein supplement and grain silage) can be met with by using AFEX-treated animal feed depending on the age of the cattle (Dale, 2008; Sendich & Dale, 2009).Until large-scale animal feed trials are conducted, these displacement ratios are assumed to be applicable. Animal feed is included in system boundaries of both processing systems (as were transportation and landscape values) since this is an integrated system-wide analysis. Although this does not affect comparative calculations in this case, it is possible in future analyses that only certain RBPDs send back animal feed to farms based on their location, feedstock type processed and technologies included, in which case results from comparative studies may vary.

Processing energy and emissions were obtained from the NREL/Dartmouth Aspen plus biorefinery model (Sendich & Dale, 2009). This is the principal simulation model for US cellulosic ethanol production in a centralized biorefinery. The model contains all the conversion processes for ethanol production namely feedstock handling, pretreatment, biological conversion (hydrolysis and fermentation), product recovery, utilities production and waste treatment (Laser et al., 2009b). The RBPD energy and emissions were calculated by isolating processes applicable to the depots and scaling them to its lower capacity (compared with the fully integrated centralized biorefinery). Processes absent in the depots such as biological conversion and ethanol purification and recovery were excluded from the model. The process energy and emissions for densification were obtained from literature (Sokhansanj & Fenton, 2006a) for all nine depots in the network except for the tenth one which is colocated with the biorefinery. While incorporating densification values from literature, it was ascertained that that the properties of pretreated feedstock are compatible with the conditions required for densification (Mani et al., 2006). (Densification details available in Supporting Information.) Similarly, energy requirements and emissions generated during transportation of densified or nondensified feedstock were added to the distributed and centralized processing system, respectively, using literature data (Sokhansanj & Fenton, 2006a).

The energy and emission inputs and outputs from all the sources discussed above were aggregated and the NEY and NCER values were calculated for each cropping system, combined with distributed or centralized processing and their respective transportation information. Table 2 shows all the modules and inputs included in this LCA.

Table 2.   Unit processes in the life cycle assessment (LCA) and their energy inputs and emission outputs
System moduleInputEnergy input and emissions output per kg dry feedstock
[E=Energy (MJ), C=Emissions (kg CO2eq)]
  • All values are based on kg dry feedstock and represent biomass derived only from the landscape area under consideration.

  • *

    The values for cropping systems are entered as an average of all three configurations and for the perennial grasses as an average of high and low grass yields over all three configurations.

  • All emissions from crops are in terms of sequestered kg CO2 equivalent and hence have a negative sign.

  • ‡All values for combined cropping system of corn grain, stover and rye.

Feedstock production and harvest, animalsEnergy for cultivation and harvestCropping system*EC
Corn system4.274−0.854
RBPD networkProcessing energyProcessEC
Single RBPD (excluding pelletization)0.740.0001
Densified and nondensified biomass transportTransport energy EC
BiorefineryProcessing energy E=13.24C =0.0018

There are two chief assumptions in this study. Firstly, it is assumed that the distributed processing network taken as a whole is as energy self sufficient as the centralized system. Accordingly, all the energy inputs for the distributed system are estimated as a direct scale-down of the integrated biorefinery model. This is true if lignin-rich process residues are burnt for electricity (Laser et al., 2009b) or if other energy sources such as heat/electricity from thermochemical conversion or methane-rich biogas from anaerobic digestion of manure(Laser et al., 2009a) are present in the distributed networks. Second, the transport differences between the two processing systems are accounted for in terms of transporting bales (where no densification is involved – in the centralized biorefinery nor in the tenth depot) vs. pellets (where densification is present – in the nine depots) for an average transportation radius of anywhere between 20 and 100 km for both processing systems as found in literature (Sokhansanj & Fenton, 2006a). For the distributed processing network, it is assumed that the transport distance between the farms and the depots is negligible compared with the transport distance of pellets from depots to the biorefinery. In the case of the centralized biorefinery, it is assumed that all biomass is directly transported from the farms to the refinery in the form of conventional bales.


Base-case scenario

According to this analysis, Configuration 1, where 65% of the acreage was allotted to the corn system, with HG yields has the greatest NEY because it exhibits the greatest total biomass production from all sources (grain, stover, rye and perennial grasses). Where the corn system dominates the acreage (for both LG and HG yields) the centralized processing system has a greater NEY than the distributed system (in Configuration 1). Rye and grain, which make up approximately 40% of total biomass in this configuration, are not densified prior to transport, thereby eliminating a primary advantage of the depots. As the acreage dedicated to the corn system gradually decreases and the total amount of densified biomass increases in Configurations 2 and 3 (with 85% and 100% of the entire acreage dedicated to perennial grasses, respectively), the NEY of the distributed processing system surpasses that of the centralized system. On average, miscanthus has 15% and 56% greater NEY than the corn system and switchgrass, respectively, because of its greater biomass yield. The corn system on the other hand has 48% greater NEY than switchgrass since the combined biomass from the corn system is greater than that from switchgrass alone. Figure 4 shows the NEY of the two processing systems in different landscape configurations per unit land area.

Figure 4.

 Net Energy Yield (NEY) per hectare of the collective system (landscape, transport, processing) in different configurations. The NEY (MJ) of the collective system comprises differences in energy inputs and outputs of each processing system (distributed or centralized) combined with crops and transport. HG represents high yields and LG represents low yields of the two perennial grasses.

Perennial grasses can contribute large quantities of SOM over time because of their thick root masses and since they do not need to be replanted each year (Post & Kwon, 2000). Moreover, no-till agricultural systems for corn also increase the amount of carbon stored in soil (Bernacchi et al., 2005). According to this analysis, switchgrass has the greatest potential for carbon sequestration among all the cropping systems. The corn system has a comparable carbon sequestration potential to switchgrass. Switchgrass and the corn system both have about 30% greater sequestration potential than miscanthus on average for all configurations. Although both switchgrass and miscanthus are perennial grasses, their different sequestration potential is due to the differences in inputs and planting practices. Switchgrass and the corn system are assumed to be cultivated using no-till whereas miscanthus is planted conventionally since it is a rhizome and must be propagated asexually (Pyter et al., 2007). Moreover, the carbon sequestration potential is related to yields because the total aboveground residue declines with decreasing yields. Consequently, the total residue contributing to SOM increase decreases, thereby reducing carbon sequestration potential. This is the reason for differences in sequestration potential between low and high yielding grasses. It is to be noted that to calculate NCER due to gasoline displaced by ethanol for individual processing systems, a reasonable evaluation would be based on equivalent service provided by the two fuels; however this is a case of relative comparison between sequestration potentials of different cropping systems. Therefore, as stated previously this ‘closed-system’ assumption is valid because the land area remains constant in all configurations and a comparison is made within this land area between different cropping systems

The NCER results for each configuration correspond to the sequestration potentials of each cropping system. The highest NCER occurs when the entire acreage is allotted to high yielding grasses (Configuration 3 HG) mainly because high yielding switchgrass is dominant in this configuration. In contrast, the lowest emission reduction occurs with all acreage allotted to low yielding grasses (in Configuration 3 LG). This is because the sequestration potential of the combined corn system is comparable to that of the grasses and at low yields the sequestration due to grasses does not exceed that of the corn system. Figure 5 shows the NCER of combined landscape and processing systems in different configurations per unit land area.

Figure 5.

 Net Carbon Emission Reductions (NCER) per hectare of the collective system (landscape, transport, processing) in different configurations. The NCER (kg CO2eq) of the collective system comprises differences in green house gas emissions generated by processing system (distributed or centralized) combined with transport, and animal operations and carbon sequestered by feedstocks. HG represents high yields and LG represents low yields of the two perennial grasses.

Sensitivity analysis

Sensitivity analyses were performed to identify significant system variables. Table 3 summarizes the results of variations in NEY and NCER between the two processing systems.

Table 3.   Data sources and results of sensitivity analyses
  1. The effect of various parameters on the relative changes in NEY and NCER of distributed and centralized processing are shown in this table. (% difference in NEY (%ΔNEY)=[((NEYcent−NEYdist)/NEYcent) × 100], % difference in NCER (%ΔNCER)=[((NCERcent−NCERdist)/NCERcent) × 100]).

  2. NA, not applicable; NEY, Net Energy Yield; NCER, Net Carbon Emissions Reduction.

S0 – Base caseSokhansanj & Fenton (2006a)0.09−3.7
S1,1 – PAKs processD. J. Marshall (2010, personal communication)2−4.5
S1,2 – Pelletization from Pro–Xan processEnochian et al. (1980)−5−4.5
S1,3 – Ring/Die processKaliyan et al. (2009)34−4.5
S2 – TransportWang (2001)100
S3 – Credits for conversion to perennial grassesNA0.05−2.4
S4 – Absence of double-cropNA0.14−3.6

Densification. The base-case scenario established the fact that pretreated perennial grass densification is a key contributor to the distributed processing network. Densification reduces both the environmental impacts and economic costs of transportation. Hence choosing the right densification method is essential. Three separate densification processes were considered in this sensitivity: briquettes (‘PAKs’) as produced by Federal Machine (Fargo, ND, USA), pelletizing as performed as part of the Pro–Xan process on dehydrated alfalfa pellets and a ring-die densification process (see Table 3 for references). The energy requirements and emissions for these processes were incorporated into the distributed processing system calculations. The energy requirements for densification in these methods differ by 25%, −67% and 78%, respectively, compared with base-case energy requirements for pelletization. The emissions generated are not significantly different compared with base case. The densification method can be a considerable source of variation in NEY as seen in Table 3 and causes small deviations in the NCER differences between the two processing systems. Selection of a densification method will depend largely on process economics.

Transport. The base-case scenario incorporated pelletized biomass transport energy and emission information obtained from the Integrated Biomass Supply Analysis and Logistics model (IBSAL) (Sokhansanj et al., 2006b). In this sensitivity analysis, we used transportation emissions and energy inputs for nondensified biomass from the Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET) model for comparison. As seen in Sensitivity 1, using different methods of densification alters the NEY values of depots. Similarly, this sensitivity indirectly illustrates the variations in NEY and NCER when densification is not used in distributed processing networks.

Credit for conversion to perennial grasses. Growing perennial grasses instead of annual crops on the same land area can result in environmental improvements. For example, eliminating annual cultivation and monocultures can benefit farmland biodiversity. Perennials can increase SOM content thereby improving carbon sequestration and soil quality (Powlson et al., 2005). Moreover, energy inputs and maintenance costs of annual crops are higher than for perennial crops. The base-case scenario was based on the assumption of a ‘clean-slate land area’ i.e. there were no crops present on the landscape before this analysis. It is also necessary to analyze possible changes resulting from land area conversions from existing agricultural system to perennial grasses (Jensen et al., 2006). We calculated savings in energy inputs and carbon sequestration due to growing each perennial grass instead of the corn system. The corn system has greater energy inputs than both perennial grasses (Table 2); therefore an energy gain on an average of 2.7 MJ kg−1 dry biomass and 2.3 MJ kg−1 dry biomass is predicted for switchgrass and miscanthus, respectively. In Configuration 3, where no corn system was initially present the same energy inputs as corn system in Configuration 1 were assumed. Carbon sequestration increases for all configurations except for miscanthus in Configuration 1. As mentioned in the results section, miscanthus has lower carbon sequestration benefits than the corn system because of different tillage practices. Therefore, in Configuration 1 growing miscanthus instead of the corn system on the allotted acreage is unfavorable for sequestration. On average, carbon sequestration gains of 0.75 kg CO2eq kg−1 dry biomass and 0.36 kg CO2eq kg−1 dry biomass are observed for switchgrass and miscanthus, respectively. The relative differences in NEY and NCER between the two processing systems are nearly the same (Table 3).

Absence of double crop. Double crops are attracting interest as a method to maintain or increase soil carbon content after harvesting agricultural residues, mainly corn stover (Fronning et al., 2008; Dale et al., 2010a). This is the primary reason for including the winter rye double crop in the base-case scenario. In this sensitivity, we removed the double crop from the corn system and reduced it to only corn grain and stover production. Removing the double crop from the system, which undergoes densified transportation, has its primary negative effect on the NEY of the distributed processing system. The NCER values are practically unaltered.

Both sensitivity analyses 3 and 4 involve varying major components of the cropping system. These components are integral to emission reductions and contribute significantly to system energy consumption. Therefore, it is important to assess individual variations in NEY and NCER compared with the base case (further discussed in the following section) because the effects of these changes are not apparent in the relative differences between the two processing systems.


Based on this analysis, to achieve NEY and NCER objectives, the entire acreage should be dedicated to perennial grasses only when their yields are high. But when perennial grass yields are low, it is more advantageous to adopt a landscape configuration containing mostly perennial grasses but including some corn system acreage. The distributed processing system has consistently greater NEY and NCER than the centralized system when combined with perennial grasses. Additionally, different perennial grass yields change NEY values by 15–50% and NCER values by 20–65% in each configuration for both processing systems. On average, the distributed processing system has practically the same NEY as and a 3.7% greater NCER than the centralized processing system.

This study also highlights the fact that distributed processing networks when combined with densified high yielding perennial grasses have consistently greater energy yields as well as larger emission reductions than centralized processing systems. Therefore, dedicated energy feedstock landscapes (using perennial grasses) work best where grass yields are high and some form of densification is involved. However, if most feedstock is trucked as bales and if grass yields are low it is unlikely that the NEY of distributed processing systems will exceed that of centralized processing systems. Evaluating the impacts of landscape conversion from high-maintenance annuals to low-maintenance perennials is also important because of the reduced energy inputs and carbon sequestration benefits of the latter systems. For sensitivity analysis cases 3 and 4, it is more effective to look at the individual differences compared with the base case for NEY and NCER rather than examining relative differences in the two processing systems. The third sensitivity analyses (credit for conversion to perennial grasses) better highlights energy inputs and emission reductions due to growing perennial grasses instead of annuals if each processing system were evaluated individually. It emphasizes the importance of a detailed analysis to assess the energy and carbon sequestration characteristics of each cropping system. Evaluations compared with the base case showed increased NEY values ranging between 13% and 33% and increased NCER values ranging between 8% and 53% for the different configurations averaged over the two processing systems and LG and HG yield cases. Similarly, although in the fourth sensitivity analysis (absence of double crop) the relative differences in emissions and energy yields of the overall systems are nearly the same, individual evaluations averaged over the two processing systems and LG and HG yields shows decreased NEY values ranging from 8% to 22% and decreased NCER figures ranging between 5% and 21%. Analyses such as these can help determine the most sustainable land configurations within mixed feedstock landscapes in the RBPD context.

The economic performance of these depots is an important factor but is outside the scope of this analysis, nor do our data and tools permit us to evaluate other environmental impacts such as water quality or biodiversity. The conclusions from this study probably apply to systems containing similar combinations of crops and land areas. Landscapes with different soil conditions, cropping systems and yields will almost certainly require similar analyses. Hence further research using more advanced tools such as ArcGIS for landscape studies is underway. We are developing flexible models for sustainable landscape configurations combined with distributed processing based on varying yields, soil conditions, landscape sizes and processing technologies. Distributed processing networks using densified biomass may be able to catalyze the formation of commodity cellulosic biomass markets, thereby providing grower incentives and advancing biofuel production. Modeling the logistics and conversion technologies and performing integrated systems investigations is a stepping stone in the successful establishment of large-scale sustainable lignocellulosic biofuel industries.


This work was funded by DOE Great Lakes Bioenergy Research Center (http://www.greatlakesbioenergy.org) supported by the US Department of Energy, Office of Science, Office of Biological and Environmental Research, through Cooperative Agreement DEFC02-07ER64494. We also thank Dr Seungdo Kim and Dr Bryan Bals for their assistance.