Agricultural and Forestry Biomass for Meeting the Renewable Fuel Standard: Implications for Land Use and GHG Emissions

Abstract

Agricultural land and forestland are considered as two largest potential biomass sources for meeting the Renewable Fuel Standard (RFS) mandate for cellulosic biofuels. However, the land use change and greenhouse gas (GHG) savings with both agricultural and forest biomass production are yet to be examined systematically. This paper examines the effects of implementing a 16-billion gallon (60 billion liters) cellulosic biofuel mandate by 2035 on the mix of agricultural and forest biomass, land use change and GHG emissions by using a dynamic partial equilibrium model of the agricultural, forestry and transportation sectors in the US. Our results show that crop residues play a significant role in supplying cellulosic ethanol before 2030, while energy crops are the major feedstocks used for meeting the RFS cellulosic mandate after 2030. Milling and logging residues are economically viable supplements to agricultural biomass for cellulosic ethanol production, though their role in total biomass is small. Across different scenarios of cellulosic ethanol mandate that can be met with either agricultural biomass only or with both agricultural and forest biomass, we find GHG savings from displacing the gasoline range from 0.61 to 0.82 B MgCO₂e over the 2015–2035 period. Induced land use change effects associated with expanded feedstock production are modest between and within the agricultural and forestry sectors. We conclude that a mixed feedstock base maximizes the economic and environmental benefits of cellulosic biofuel production. The mitigation potential of cellulosic biofuels is severalfold larger than natural-based solutions such as grassland restoration.

1. Introduction

The production of biofuels from food crops (corn grain ethanol) in the US has grown from 4 billion gallons (15 billion liters) in 2005 to more than 15 billion gallons in 2021 with the help of supportive biofuel policies such as Renewable Fuel Standard (RFS) [1]. Concerns about the impact of diverting land for biofuel feedstocks on food prices, induced land use change and water scarcity have led to US biofuel policy shifting the focus from corn ethanol to second-generation cellulosic biofuel [2,3,4,5]. The RFS sets a goal of 36 billion gallons (136 billion liters) of biofuel production by 2022, of which 21 billion gallons (79.5 billion liters) must be “advanced biofuels”.

Recently, a variety of cellulosic feedstocks from high-yielding dedicated energy crops, crop and forest residues have been promoted with more productive land use and lower GHG intensity [6]. However, there is significant uncertainty about the economic potential and environmental consequences of the supply of agricultural and forest biomass for ethanol production. What would be the economically viable mix of cellulosic feedstocks for meeting the biofuel mandate? Will this lead to the direct and indirect land use change effects? Can the replacement of gasoline with ethanol, particularly cellulosic ethanol, generate greenhouse gas (GHG) benefits? Additionally, how is biofuel mitigation potential compared with natural-based solutions such as grassland restoration or reforestation on formal agricultural land? These controversial questions are among critical issues in evaluating the performance of biofuel to meet transportation fuel needs in a low-carbon economy. Although the actual cellulosic biofuel production has significantly fallen short of the targeted levels in the past decades, the forward-looking analysis of potential economic and GHG implications of adding a significant cellulosic biofuel target to be achieved by 2035 can inform post-2022 RFS volumes.

The main purpose of this paper is to examine the economic and GHG consequences of extending cellulosic ethanol target over the 2015–2035 period. More specifically, we analyze the effects of biofuel policy on

(1)

Optimal mix of agricultural and forest biofuel feedstocks;

(2)

Land use change between and within the agricultural and forestry sectors over time;

(3)

GHG emission implications of cellulosic ethanol production.

2. Literature Review

Decades of technological advances and policy support have resulted in efforts to commercialize cellulosic biofuel pathways using lignocellulosic biomass including agricultural wastes, wood, municipal wastes and dedicated energy crops [7,8,9]. Considerable research has been dedicated to examining the supply potential and regional distribution of feedstock production and its effects on land use change and GHG emissions. Global technical biomass supply potentials are estimated to range from <50 EJ/yr to >1000 EJ/yr [10,11,12,13]. Modeling assumptions and different views about social–ecological constraints leads to varying estimates [14]. The US Department of Energy’s Billion-Ton Report (BT16) has examined the potential of producing one billion tons of biomass in the US and found that there are 991 million dry tons of biomass available at the biomass price of USD 60 per ton by year 2030 [15]. This study only examines the biophysical supply potential and ignores the economic constraints such as land opportunity costs. Since the establishment of national mandates for biofuel use, few existing studies consider the economic supply potential of joint agricultural and forestry sectors under the RFS. Taheripour and Tyner [16] analyzed the land use changes induced by biofuel production with agricultural biomass (including crop residues and energy crops) with the Global Trade Analysis Project (GTAP) model. Their results suggested that producing 26.498 billion liters (BL) miscanthus and switchgrass ethanol would cause 0.22 and 0.78 million hectares (M ha) forest land loss and 0.13 and 0.30 M ha cropland increase in the US, respectively, while corn stover had no significant induced land use change. In their study, biofuel targets were implemented as a one-time shock of biofuel produced from each type of feedstock; therefore, it did not tell the optimal supply mix of various biofuel feedstocks under total targets or mandates as well as the consequent land use change. By applying the Biofuel and Environmental Policy Analysis Model (BEPAM), Hudiburg et al. [17] studied land use change and GHG impacts of various agricultural feedstocks under RFS. They found that cellulosic biofuels can meet a 32-billion-gallon RFS target in 2022 with negligible effects on food crop production and can reduce GHG emissions by about 7%. More recently, Chen et al. [18] examined the effects of RFS mandates on GHG emissions and nitrogen (N)-damages over the 2016–2030 period. They found that cellulosic biofuels can lead to higher GHG benefits and lower N-leakages effects compared to the corn ethanol mandate. The above studies mainly focused their work on the agricultural sector and largely ignored the potential role forestry sector can play in supplying cellulosic biomass.

Apart from agricultural biomass, woody biomass has also been widely discussed as a flexible feedstock for the production of liquid transportation fuels, heat and electricity. In terms of its biophysical potential, the BT 16 [15] assessed that about 116 million dry tons of woody biomass are available at a biomass price of USD 80/dry ton by 2035. A variety of national and regional research has been conducted on estimating the economic and environmental effects of increasing woody biomass demand for renewable energy production with varied assumptions and findings. Sedjo and Sohngen [19] applied the dynamic Timber Supply Model (TSM) to examine how the national forest market might respond to increased demands for woody biomass in meeting cellulosic biofuel mandate. They found that mandated increases in cellulosic biofuels would result in 15–20% higher wood prices and a 60% increase in raw wood consumption by 2022. However, in their work, renewable fuel production was assumed as being sourced only from raw wood (no wastes or residues), which may lead to an overestimation of the consequent market effects. Ince et al. [20] estimated the forest market response to a range of RFS plus RPS (Renewable Portfolio Standard) by applying the US Forest Products Model (USFPM). Fairly small impacts on timber product consumption and prices as well as forest inventories were identified since the logging and milling residues were the primary source of the projected feedstock supply. With the updated version of USFPM, Nepal et al. [21] provided an improved understanding of traditional wood products market impacts of expanded wood energy production. By applying the Sub-Regional Timber Supply (SRTS) model in three southern states (i.e., Alabama, Georgia and Florida), Abt et al. [22] projected an approximate doubling of the 2007 pulpwood price by the year 2037 when a bioenergy demand (including liquid fuels, electricity and pellets) was in place and logging residue recovery rate was only 33%; meanwhile, the increased recovery rate would be crucial to moderating market response to expansion demand of wood biomass. In general, the above studies restrict their examination of expanding bioenergy demand in the context of forest markets without endogenously considering the potential contribution of agricultural biomass and its interactions with the forestry sector through land use change.

The Forest and Agricultural Sector Optimization Model (FASOM) is among the very limited models that have comprehensive coverage of both agricultural and forestry sectors. However, most of the existing studies applying FASOM model focused on the impacts of expanded bioelectricity demand but with a simplified biofuel part, as the model itself ignored the demand for fuels emanating from the transportation sector [23,24,25].

Most carbon abatement scenarios rely on substantial contribution from the bioenergy system because of their potential to displace transport sector fossil fuels and mitigate greenhouse gas emissions [26]. However, the climate benefits of large-scale cellulosic biofuels have been challenged due to concerns around land use change and carbon debt generated by the release of carbon stocks in that land. With increasing scarcity of arable land and growing competitive demand for food and feed, bioenergy-induced land use change may cause high GHG emissions and higher N₂O emissions with agricultural intensification. Studies have opposite opinions toward GHG outcomes of producing cellulosic ethanol [26,27,28]. These studies either ignore the economic effects (i.e., opportunity cost of land) or only consider representative energy crops, disregarding the coexistence of feedstocks from multiple sources. Ambiguity in the availability of feedstocks, induced land use change and GHG emissions effects of large-scale development of cellulosic ethanol also results in suggestions on refocusing policy support away from biofuel technology toward natural-based solutions such as grassland restoration [29]. It is essential to conduct a systematic analysis of the supply and mix of feedstocks and consequent GHG emissions effects with a sophisticate model.

Agricultural and forest biomass have been suggested to be a supplement to each other in meeting RFS mandates; however, they also compete for the use of limited private lands that can produce conventional agricultural or forest products [15,30]. Therefore, it is more important to know the economic availability of various feedstocks from both sectors and their joint impacts on direct and indirect land use change and GHG emissions. In the present research, we examine the role of both agricultural and forestry feedstocks in meeting RFS by applying an integrated model of three linked markets: agriculture, forest and transportation fuel. This framework allows for endogenous determination of the relative contribution of fossil fuel, agricultural and woody biomass in meeting transportation fuel demand, land use change and the climate mitigation potential of biofuels. In addition, our model improves over the existing stylized model (e.g., FASOM, SRTS) in several ways. First, land allocation decisions are made on a finer spatial resolution in the agricultural sector. Second, it accounts for a broader source of cellulosic biofuel feedstocks including dedicated energy crops, crop residues, short-rotation woody crops and woody biomass. Third, it enables endogenous assessment of induced land use change and associated emissions. Fourth, the recursive structure facilitates the simulation of future improvement in productivity and cost of biofuels production due to learning-by-doing.

3. Methods

3.1. Model Description

The economic model (BEPAM-F) applied for this study is a dynamic, open-economy, multimarket partial equilibrium model that integrates the agricultural, forestry and transportation sectors in the US [17]. The model endogenously determines the optimal land use and feedstock mix by maximizing the aggregated economic welfare (summation of consumer and producer surplus) subject to various material balance, technological, land availability and policy constraints over the 2010–2035 period represented in five-year timesteps. The agricultural sector in the model includes thirteen conventional crops (corn, soybeans, wheat, rice, sorghum, oats, barley, cotton, peanuts, alfalfa, sugarbeets, sugarcane and corn silage), eight livestock animals (beef, pork, chicken, turkey, wool, lamb, eggs and milk), various processed commodities (vegetable oils from corn, soybeans and peanuts; DDGs; soybean meal; refined sugar and high-fructose corn syrup), five bioenergy crops (miscanthus, switchgrass, energycane, hybrid poplar and willow) and two crop residues (corn stover and wheat straw) at the 295 Crop Reporting District (CRD) level in the US. Five types of agricultural land (irrigated and nonirrigated cropland, idle cropland, cropland pasture and pastureland) are specified at the CRD level. Cropland can move between the production of different crops with no extra cost but is subject to a convex combination of both historical and synthetic crop mixes [31,32]. The cost of converting noncropland to crop production is uncertain due to the absence of empirical data. We assume the one-time conversion costs to be at least as the returns from the least profitable crop in a CRD as in Chen et al. [18]. The same cost is assumed for converting noncropland to energy crops. The opportunity cost of cropland for producing energy crops is determined endogenously as the difference between the per-hectare revenues from the most profitable crop production and the costs of production. More details on the agricultural sector modeling can be found in the previous literature [17,33].

The structure of the forestry sector is similar to FASOM, which was based on the family of timber assessment models that include the Timber Assessment Market Model (TAMM), the North American Pulp and Paper Model (NAPP) and the Aggregate Timber Assessment System (ATLAS). The forestry production occurs in five aggregated regions in the US. The regional definition follows FASOM, and the detailed states included in each aggregated region are displayed in Table S1 the supplementary information (SI). Forestland in the model refers to land privately owned by the forest industry and nonindustrial private land, while public forestland is not explicitly tracked. County-level forestland data were obtained from US Forest Service and aggregated to the regional level (SI Table S2). Timber growth and yield are included for existing, reforested and afforested stands for both softwoods and hardwoods. Timber yields are determined based on the Forest and Rangeland Renewable Resources Planning Act Timber Assessment in 2000 and vary by stand age, regions and management intensity [34,35]. Forest land can be converted within the sector and across sectors based on predefined suitability classes, which constrain what type of land can be converted to forest, forestland pasture or cropland. Softwood and hardwood logs in the form of sawlogs, pulpwood and fuelwood can be harvested from forestland and further be produced to various end products. More than 40 major wood products and associated manufacturing processes as well as products trade between US and the rest of world (ROW) are included in this sector. Optimal decisions regarding forest management intensities, harvesting age and whether to replant or deforest are dynamically determined by the relative returns to alternative actions. Based on current technology, we assume that either logging residues (with a 65% recovery rate) or milling residues (including wood pulp) in this sector can be used to produce biofuel. Considering the use of cellulosic feedstocks in the electricity sector may compete with the use in the production of transportation fuels, we also test the scenario by incorporating the demand for bioelectricity using forest and agricultural biomass in the sensitivity analysis. A complete review of the forestry sector in BEPAM-F model can be found in Wang et al. [36].

The on-road transportation sector includes linear demand curves of vehicle kilometer travels (VKT) for four types of vehicles, including conventional gasoline (CV), flex-fuel (FFV), gasoline-hybrid (HV) and diesel vehicles (DV). The projected demand for VKT with each type of vehicle is obtained from recent Annual Energy Outlook. The total supply of biofuels is sourced from four types, first-generation ethanol, cellulosic ethanol, first-generation biodiesel and second-generation biomass to liquids diesel. The model endogenously determines the extent and mix of biofuels blended and implicit cost of VKT based on the energy contents of alternative fuels, the fuel economy of vehicle and the limits of biofuel blend [18].

The agricultural and forestry sectors are linked in the model by competing for the private land use, which can produce either agricultural or forest products. Meanwhile, they link with the transportation sector by providing energy feedstocks to meet the national biofuel demand for VKT. The CRD level agricultural land is mapped to the five regional level forest lands through aggregation. Land moves between sectors until the market equilibrium has been reached, which depends on the net present value of returns to alternative uses plus the investment cost of land transfer (e.g., land clearing, leveling and seedbed preparation, etc.) and any other hurdle cost.

3.2. Algebraic Illustration of the Model

The mathematical representation of main functions in BEPAM-F model is described below. The objective function is the sum of discounted consumers’ and producers’ surplus in the agricultural, transportation fuel and forest sectors over a planning horizon p = 1,…P.

$$\mathrm{Max}: {\displaystyle \sum }_{p=1}^{P \mathrm{or} 9999}\frac{1}{{\left(1+\mathrm{r}\right)}^{5p}}\{{\mathrm{d}}_p·\left(AGRWE{L}_p+TRAWE{L}_p+FORWE{L}_p\right)\}$$

(1)

where p represents a five-year period and r is the discount rate, which is 0.04 in this model. AGRWELp, TRAWELp and FORWELp denote the welfare of agricultural, transportation fuel and forest sector, respectively. $d_p$ is defined as an annuity factor as the time scale is 5 years in the model.

$$d_p=\{\begin{array}{ll}{\displaystyle \sum }_{\mathrm{t}=0}^4\frac{1}{{\left(1+\mathrm{r}\right)}^{\mathrm{t}}}& \mathrm{for} 1\le p\le P\\ {\displaystyle \sum }_{\mathrm{t}=0}^{\infty }\frac{1}{{\left(1+\mathrm{r}\right)}^{\mathrm{t}}}=\frac{1}{\mathrm{r}}& \mathrm{for} p=9999\end{array}$$

(2)

The activities in the last period P are treated as if they continue forever by assuming that market conditions, agricultural yields and technology will remain the same. Terminal conditions in the forestry sector are based on the assumption that returns to various forest products will occur in perpetuity using von Mantel’s formula [37]. For terminal conditions, we define P = 9999.

The algebraic expression for agricultural sector welfare in period p is given in (3):

$$\begin{gathered} AGRWE{L}_p={\displaystyle \sum }_Z{{\displaystyle \int }}_0^{DE{M}_{p,z}}{f}^z(.)d(.)+{\displaystyle \sum }_z{{\displaystyle \int }}_0^{EX{P}_{p,z}}{f}^z(.)d(.)-{\displaystyle \sum }_z{{\displaystyle \int }}_0^{IM{P}_{p,z}}{f}^z(.)d(.) \\ -{\displaystyle \sum }_{R,q}r{c}_{R,q}AC{R}_{p,R,q}-{\displaystyle \sum }_{R,ec,age,type}p{c}_{R,ec,type}ECAC{R}_{p,R,ec,age,type}-{\displaystyle \sum }_{R,q}r{s}_{R,q}AC{R}_{p,R,q} \\ -{\displaystyle \sum }_{k}l{c}_{k}LI{V}_{p,k}-{\displaystyle \sum }_{i}s{c}_{i}PR{O}_{p,i}-{\displaystyle \sum }_{R,type}^{}C\left(AGTOFO{R}_{p,R,type}\right) \forall p \end{gathered}$$

(3)

The first integral term in the first line of (3) represents the areas under the demand function from which consumers’ surplus is derived. $DE{M}_{p,z}$ denotes the endogenous domestic demand variable in year p; $f^z(.)$ denotes the inverse demand function for the associated commodity; and d(.) denotes the integration variable. The next two integral terms account for the areas under the inverse demand functions for exports, $EX{P}_{p,z}$, and the areas under the import supply function, $IM{P}_{p,z}$. The second line in (3) includes production costs of row crops and feedstocks. The land allocated to row crops is denoted by $AC{R}_{p,R,q}$; the acreage of perennial crop ec planted on land type type in region R, age age and period p are denoted by $ECAC{R}_{p,R,ec,age,type}.$ The production practices denoted by q differ by crop rotation, tillage and irrigation. Parameters denoting the production costs of row and energy crops are $r{c}_{R,q}$ and $p{c}_{R,q}$, respectively. Leontief production functions are assumed for both row crops and perennial crops production. The third term in the second line of (3) represents the cost of collection crop residues. $r{s}_{R,q}$ denotes the unit collection cost. The first term in the third line denotes the costs associated with livestock activities, where $l{c}_k$ denotes the cost per unit of livestock category k. Variable $LI{V}_{p,k}$ represents the quantity of livestock k processed in period p. The second term represents the total cost of converting primary crops to secondary commodities (oils, soymeal, refined sugar, etc.). $PR{O}_{p,i}$ and SC_i denote the amount of processed primary crop and per unit processing cost, respectively. The last term represents the cost due to land movement from agriculture to forest sector ($AGTOFO{R}_{p,R,type}$).

$$\begin{gathered} {\displaystyle \sum }_{q}AC{R}_{p,R,q}+{\displaystyle \sum }_{ec,age}ECAC{R}_{p,R,ec,type}\le a{l}_{0,R}+FORTOA{G}_{p,R,type}-AGTOFO{R}_{p,R,type} \\ \forall p,R, type=‘regular’ \end{gathered}$$

(4)

The cropland is subject to the land constraint (4) for each CRD R. The parameter $a{l}_{0,R}$ denotes the regular land availability in the initial period and region R. Total forest land converted to agriculture land in region R period p agriculture land type type is denoted by $FORTOA{G}_{p,R,type}$.

The amount of biomass (BM) used for biofuel production should not exceed the total production, which is expressed in Equation (5):

$$B{M}_{p,R }\le {{\displaystyle \sum }}_{q}b{y}_{R}AC{R}_{p,R,q}+{{\displaystyle \sum }}_{ec,type}r{y}_{R,ec,type}ECAC{R}_{p,R,ec,type } \forall p,R$$

(5)

where $b{y}_R$ and $r{y}_{R,ec,type}$ are the energy crop and crop residue yields in region R for respective perennial and crop production activities.

The formula for forest sector welfare in period p is given in (6):

$$\begin{array}{l}FORWE{L}_p={\displaystyle {\displaystyle \sum }_w}{{\displaystyle \int }}_0^{DE{M}_{p,w}}f(.)d(.)+{\displaystyle {\displaystyle \sum }_w}{{\displaystyle \int }}_0^{EX{P}_{p,w}}f(.)d(.)-{\displaystyle {\displaystyle \sum }_w}{{\displaystyle \int }}_0^{IM{P}_{p,w}}f(.)d(.)\\ -{\displaystyle {\displaystyle \sum }_{R_1,R_2,w}}t{c}_{p,R_{1,}{R}_{2,}{w}_{, }}TRA{N}_{p,R_1,R_{2, }w}-{\displaystyle {\displaystyle \sum }_{R,w}}m{c}_{R,w}MAN{U}_{p,R,w}\\ -{\displaystyle {\displaystyle \sum }_{R,l}}(g{c}_{p,R,l}+h{c}_{p,R,l})(EXIS{T}_{p,R,l}+NE{W}_{p,R,l})-{\displaystyle {\displaystyle \sum }_{R,w}}w{c}_{R,w}WAS{T}_{p,R,w}\\ -{\displaystyle {\displaystyle \sum }_{R,type}}C\left(FORTOA{G}_{p,R,type}\right) \forall p\end{array}$$

(6)

The first integral term in the first line of (6) represents the areas under the domestic demand function from which consumers’ surplus is derived. The term w represents all classes of wood products; DEM_p,w denotes the endogenous domestic demand variable for product w in period p; f(.) denotes the inverse demand function for the associated commodity; and $d(.)$ denotes the integration variable. The next two integral terms represent the inverse demand functions for exports EXP_p,w and the areas under the import supply functions IMP_p,w, respectively.

The second and third lines in (6) include all the costs involved in producing and supplying logs and wood commodities. $t{c}_{p,R_1,R_2,w}$ is the unit cost of interregional transportation of wood products from region R₁ to region R₂; variable $TRA{N}_{p,R_1,R_{2, }w}$ denotes the amount of wood product w being transported from region R₁ to R₂ in period p. The parameter $m{c}_{R,w}$ is the manufacturing cost per unit of wood products; $MAN{U}_{p,R,w}$ denotes the amount of wood product w being produced in region R. The parameter $g{c}_{p,R, l}$ and $h{c}_{p,R, l}$ represents growing and harvesting tree costs, respectively. l indicates a series of attributes of logs including age, ownership (industry only or other private), type (hardwood or softwood), site productivity (20–50 cf, 50–85 cf, 85 cf+) and management intensity. The variable $EXIS{T}_{p,R,l}$ and $NE{W}_{p,R,l}$ represents total biomass of existing tress and newly planted trees (reforested or afforested), respectively. Forestry wastes $WAS{T}_{p,R,w}$ are subject to penalty cost $w{c}_{R,w}$. The last term denotes the cost due to land movement from forest to agriculture sector. When land transfers from forest to agriculture, it requires an investment to clear stumps and level the land and results in foregone profits from forest production.

Equation (7) below shows the wood products supply and demand balance:

$$\begin{gathered} WDDE{M}_{p,R,w}\le \\ {{\displaystyle \sum }}_{R_1}TRA{N}_{p,R_1,R,w}+MAN{U}_{p,R,w}-{{\displaystyle \sum }}_{R_2}TRA{N}_{p,R,R_2,w}-WAS{T}_{p,R,w} \forall p,R,w \end{gathered}$$

(7)

where $WDDE{M}_{p,R,w}$ is demand for regular wood product w in region R period p. $MAN{U}_{p,R,w}$ is manufactured wood product w in region R period p; $TRA{N}_{p,R_1,R,w}$ denotes transportation of wood product w from region R₁ to R, while $TRA{N}_{p,R,R_2,w}$ denotes transportation of wood product w from region R to R₂.

Residue by-product generation is shown in Equation (8). Milling residues are generated primarily at solid wood manufacturing facilities and may be transferred to other manufacturing processes, such as fiber production, combustion for energy generation, or sent to landfills. $rescoe{f}_{w,w1}$ is the coefficient that reflects the amount of logging and milling residues being produced with one unit of manufactured wood product.

$$FORRE{S}_{p,R,w}={\displaystyle \sum }_{w1}MAN{U}_{p,R,w1}·rescoe{f}_{w,w1} \forall p,R,w$$

(8)

$$MILRE{S}_{p,R,w}+LOGRE{S}_{p,R,w}+{\displaystyle \sum }_{w1}MAN{U}_{p,R,w1}·wdcoe{f}_{R,w,w1}\le FORRE{S}_{p,R,w}$$

(9)

Equation (9) presents residue demand and supply balance. The amount of milling residue ($MILRE{S}_{p,R,w}$) and logging residues ($LOGRE{S}_{p,R,w}$) used for bioenergy plus those used as input for manufacturing other timber products should be no more than the total available residues ($FORRE{S}_{p,R,w}$). $wdcoe{f}_{R,w,w1}$ is the input–output coefficient for wood products manufacture, namely the amount of input material w needed for producing one unit of w₁.

Equation (10) restricts the percentage of harvestable logging residue to be no more than 65% based on Thiffault et al. [38].

$$LOGRE{S}_{p,R,w}\le 0.65·FORRE{S}_{p,R,w } \forall p,R,w$$

(10)

The welfare function of transportation fuel sector is given in (11):

$$\begin{gathered} TRAWE{L}_p={{\displaystyle \int }}_0^{MI{L}_{p,v}}{f}^v(.)d(.)+{{\displaystyle \int }}_0^{G_{p,row}}{f}^g(.)d(.) \\ -{{\displaystyle \int }}_0^{GA{S}_p}{f}^0(.)d(.)-{{\displaystyle \int }}_0^{DS{L}_p}{f}^{dsl}(.)d(.)-epcET{H}_p \forall p \end{gathered}$$

(11)

The first integral term in the first line represents the area under the inverse demand function for kilometers traveled with alternative vehicles (MIL_p,v). The second integral term denotes the area under the inverse demand function for gasoline consumer by the ROW (G_p,row). The second line involves the costs accruing to the fuel sector. The first and second integral represents the area under the supply functions for gasoline ($GA{S}_p$) and diesel ($DS{L}_p$), respectively. Ethanol processing cost in the refinery for one unit of ethanol is epc. $ET{H}_p$ is total produced ethanol in period p.

Gasoline kilometers generated for each type of gasoline-based vehicle by blending gasoline and ethanol are formulated in Equation (12) below, where E_p,v and G_p,v represent the consumption of ethanol and gasoline for each type of vehicle, respectively. ${\gamma }_{p,v}$ is the fuel efficiency (kilometers per liter).

$$MI{L}_{p,v}={\gamma }_{p,v}\left(\frac{2}{3}{E}_{p,v}+G_{p,v}\right) \forall p,v$$

(12)

3.3. Data and Parametric Assumptions

CRD-specific crop production and historical planted acres of thirteen conventional crops for the years 2000 to 2010 are obtained from National Agricultural Statistics Service (NASS) [39]. Data on idle cropland, cropland pasture, permanent pastureland and forestland pasture for each CRD are also obtained from NASS. In 2010, the availability of pastureland and forestland pasture was 390 million (M) acres and 38 M acres, respectively, while that of idle cropland was 31 M acres and that of cropland pasture was 30 M acres. Crop yields are computed as CRD-level production divided by the respective planted acres in each CRD. Costs of crop production in 2010 prices for conventional crops are obtained from the crop budgets complied for each state by state extension services. Crop budgets vary by rotation, tillage and irrigation choices. The costs of crop production include the costs of inputs such as seeds, fertilizers and chemicals; the costs of drying and storage; machinery and fuels; and costs of crop insurance. Corn stover and wheat straw yields for each CRD are obtained based on a 1:1 and 1:1.5 grain-to-residue ratio of dry matter of crop grain to dry matter of crop residues, respectively [40]. We assume that 50% of the residue can be removed from fields if no-till is practiced, and 30% can be removed if conventional tillage is used [41]. Corn stover yield ranges 0.5–4.3 Mg ha⁻¹, while wheat straw yield ranges 0.4–2.3 Mg ha⁻¹ in the US.

Costs of production of miscanthus and switchgrass are developed for each year of their lifetime for each CRD including the costs of inputs (fertilizer, seed and chemicals), energy and machinery required for establishment and harvest of bioenergy crops, and storage and transportation. Opportunity costs of land for these crops are implicitly determined given a land constraint in the model. The costs of collecting corn stover and wheat straw include the additional cost of fertilizer that needs to be applied to replace the loss of nutrients. Additional costs with harvesting residues are estimated based on the state-specific crop budgets on alfalfa harvesting. Budget data for energycane are based on production cost estimates in Salassi and Deliberto [42]. Willow and poplar production cost is calculated based on Lazarus [43]. National average production costs for feedstocks are presented in Table S3. The conversion efficiency of cellulosic ethanol (biofuel yield per metric ton of feedstock) is estimated as 330.5 Liter Mg⁻¹ for biomass [44]. The processing costs of producing cellulosic ethanol are estimated as USD 0.35 L⁻¹ in 2010 prices. The future costs of conversion of feedstock to biofuel are determined using an experience curve approach following de Wit et al. [45]. The parameters used for specifying these learning-by-doing curves are shown in Table S4.

Cost estimates for stand establishment and intermediate treatments differ by forest type, management intensity and region (see Table S5). Log harvesting and hauling costs vary with forest volume and region (Table S6). The transportation cost for delivering logs from site to mill plants is assumed to be USD 5 per green metric ton for 25 miles one way by truck. Three types of forest biomass used for energy production are considered in the model: logging residues, milling residues and wood pulp. Logging residues are assumed to have no value in the no-cellulosic biofuel case, while milling residues and wood pulp are used extensively in the manufacture of fiberboard and paper.

Linear demand curves for VKT are calibrated for year 2010 using data on VKT by each type of the four types of vehicles, a cost per kilometer and an assumed elasticity of demand for VKT. The VKT data for each type of vehicle are obtained from EIA [46]. Using the average maintenance costs and fuel costs obtained from the Alternative Fuel Data Center (AFDC), we calculate the cost per kilometer traveled as USD 0.08 km⁻¹ for CV, USD 0.08 km⁻¹ for FFV, USD 0.04 km⁻¹ for HV and USD 0.13 km⁻¹ for DV, all in 2010 prices. The price elasticity used for calibrating the demand curve is −0.6 for CV, −0.7 for FFV and −0.2 for the rest of vehicle types [18]. Our assumptions of the elasticities lie in the range of the data from existing literature (−0.13 to −0.96) [47,48,49]. The VKT with each type of vehicle is shifted outward for the 2010–2035 period based on the projection by EIA [46].

3.4. Scenarios Description

To analyze the economic and environmental effects that can be attributed to cellulosic ethanol, we consider the following three alternative scenarios over the 2015–2035 period:

(a)

No-cell EtOH: corn ethanol production is assumed to maintain at the maximum level of 56.78 BL under the RFS, and biodiesel production is maintained at the 2015 level of 4.8 BL.

(b)

Ag-only EtOH: Cellulosic ethanol production is assumed to grow linearly from 0.8 BL in 2015 to 60.56 BL in 2035, while corn ethanol and biodiesel production are the same as in the No-cell EtOH scenario; the cellulosic ethanol target is met by agricultural biomass only.

(c)

Wood & Ag EtOH: Cellulosic ethanol production is the same as in the Ag-only EtOH scenario, and the total target is met by a mix of forest and agricultural feedstocks; corn ethanol and biodiesel production are the same as in the above scenarios.

The RFS target is implemented in the model by establishing blend rates to meet the volumetric goals for ethanol and biodiesel. In all three scenarios, we assume that landowners make decisions within a finite time horizon (2015–2035) and update their decisions every five years based on realized market outcomes, changes in land availability and technological innovation.

3.5. GHG Emissions Estimation

We estimate the direct GHG emissions related to cellulosic feedstock production and conversion to biofuel over the 2015–2035 period by combining life-cycle GHG emissions from all agricultural and forest activities. The aboveground GHG emissions related to agricultural activities, including planting, maintenance and harvest are estimated by multiplying various production inputs with corresponding emission factors extracted from the Greenhouse Gases, Regulated Emissions and Energy Use in Transportation (GREET) model (SI Table S7). These input data are obtained from region-specific crop budgets. The average energy inputs are calculated by multiplying these input quantities with associated energy values (in MJ/unit) and are presented in Table S8. The annual dynamic change of soil carbon for each feedstock is simulated by the ecosystem model (DayCent) over 30 years based on soil characteristics, weather conditions, energy crop types and residue removal scenarios. The summary of average yield and soil organic carbon of each type of energy crop are shown in SI Table S9. We endogenously determine the US domestic indirect land use change (ILUC) related GHG emissions by multiplying the amount of cropland pasture and idle land converted to crop production with the emission factors obtained from the DayCent model. In the benchmark study, international ILUC emissions are not explicitly considered. However, we include these emissions in the sensitivity analysis in which the ROW ILUC emission factors for miscanthus, switchgrass and corn stover are 2.12, 7.92, and −1.06 gCO_2eMJ⁻¹, respectively [16]. Forest sector GHG emissions account for stocks and flows of carbon in forest ecosystem pools, timer production and harvested wood products over the planning horizon. The model explicitly incorporates carbon in standing live and dead tree biomass, forest understory vegetation and organic litter in the forest floor and in forest soils. GHG emissions during harvesting and in harvested wood products are based on Bergman et al. [50].

We then estimate the net GHG savingsfrom replacing gasoline with cellulosic ethanol by deducting life-cycle GHG emissions with cumulative cellulosic ethanol production from GHG emissions with energy equivalent amount of gasoline. The formula used to calculate the GHG saving (GHGs) is listed below, following Chen et al. [18]:

$$GH{G}_s=CE\times EC\times GE\times CI-GH{G}_{cell}$$

where CE is the total amount of cellulosic ethanol production, in liters; EC is energy content of gasoline, 34.8 MJL⁻¹; GE is ethanol-to-gasoline equivalent factor, 2/3; CI is carbon intensity of gasoline, 94 gCO_2eMJ⁻¹; $GH{G}_{cell}$ is the life-cycle GHG emissions with cellulosic ethanol production, in gCO_2e.

4. Results

We first validate the model by comparing simulated outcomes in the agricultural, forestry and transportation sectors for 2010 and 2015 with observed data. Given that the model is calibrated for a 5-year cohort, the 2010 and 2015 observed data represent an average value over the 2010–2014 and the 2015–2019 periods, respectively. As shown in SI Table S10, the divergence between model results and the observed land use allocations and commodity prices for agricultural, fuel and forest products is typically within ±10%.

4.1. Biomass Supply

We present feedstock mixes used to meet the cellulosic ethanol mandate under different scenarios in Figure 1. RFS cellulosic ethanol mandate results in a mix of various feedstocks. Under the Ag-only EtOH scenario, the primary agricultural biomass provider is projected to be crop residues (corn stover and wheat straw) over the 2015–2025 period. After 2030, nearly half the volume of cellulosic ethanol production is provided by energy crops, in which miscanthus supplies 25.8 BL, followed by switchgrass (1.8 BL) and other energy crops (1.2 BL). Under the Wood & Ag EtOH scenario, the crop and woody residues are projected to be the dominant initial (before 2025) sources of cellulosic ethanol, in which 45% of ethanol is from crop residues, and 20% of ethanol comes from logging and milling residues. With time progressing and increasing biofuel demand, feedstocks from the agricultural sector tend to increase. By 2035, energy crops are the major feedstocks used for cellulosic ethanol production (46%), followed by crop residues (44%) and woody biomass (10%). The amount of ethanol produced from logging residues and milling residues is about 4.3 BL and 1.5 BL, respectively. Little ethanol comes from pulpwood over the projected periods. Although the production cost of pulpwood is relatively low among all feedstocks, it is not projected as a cost-effective provider for biofuels due to the high opportunity cost incurred by conventional pulp and paper products. In summary, woody biomass from logging and milling residues can be an economically viable substitute for part of crop residues in the optimal mix of cellulosic feedstocks.

The supply of biomass differs across regions in the US. Figure 2 shows the regional distribution of different types of biomass in 2035 under alternative scenarios. We find that of the total crop residues produced under both scenarios, 50% is produced in the Midwest and 35% in the Plains. In contrast, of the total energy crops produced, about 52% is sourced from the Southern states and 39% in the Plains. In the case of the Ag-only EtOH scenario, a substantial amount of miscanthus, 70 million metric tons (M Mg), is produced primarily in the South and the Plains, while most of switchgrass is produced in the Plains. In the Wood & Ag EtOH scenario, the majority of logging residues (9 M Mg) are supplied in the southern region, while most of the milling residues (3 M Mg) come from the western states.

4.2. Land Use Change

We estimate the amount and regional distribution of direct and indirect land use changes by tracking land conversion among different uses. Direct land use change refers to diverting existing cropland or idled land to energy crop production; indirect land use change here refers to the conversion of noncropland (i.e., cropland pasture, forestland) to cropland within domestic boundaries in response to the increased demand for cellulosic biomass. RFS cellulosic ethanol mandate leads to 4.6 M ha and 4.3 M ha of agricultural land (including regular cropland, cropland pasture and idle land) being converted to energy crops by 2035 under the Ag-only EtOH and Wood & Ag EtOH scenarios, respectively. The effect of the RFS on cumulative indirect land use change over the 2015–2035 period is determined by comparing a baseline scenario in the absence of the policy (No-cell EtOH) to the scenarios in which RFS is in place (Ag-only EtOH and Wood & Ag EtOH) (Figure 3). Increased crop prices with heightened biofuel demand will consequently create incentives for farmers to convert less productive cropland pasture to regular crop production. The amount of land conversion from cropland pasture to cropland is found to be 0.7 M ha and 0.5 M ha under the Ag-only EtOH scenario and Wood & Ag EtOH scenario, respectively. There is less amount of land conversion from cropland and cropland pasture to forests in both scenarios relative the No-cell EtOH case as biofuel demand increases the land rents for cropland. Under the Ag-only EtOH case, RFS leads to 0.05 M Ha more (relative to the No-cell case) conversion of land from forests to cropland and 0.07 M ha more conversion of land from forestland pasture to forests. With the cellulosic ethanol demand met by both forest and agricultural biomass as in the Wood & Ag EtOH scenario, additional demand for woody biomass leads to reduced conversion of forests to cropland relative to the Ag-only EtOH case, but there is a slightly greater amount of conversion from forestland pasture to forests. As a result, with a mix of agricultural and forest feedstocks, the RFS cellulosic mandate by 2035 leads to the net expansion of 0.9 M ha in regular cropland and a reduction of 0.7 M ha of forestlands. The overall effect on land use change is very modest; this is likely because the large amount of biomass is from crop and woody residues that are harvested as a by-product of existing production without diverting extra land from existing uses.

SI Table S11 shows the distribution of land use change relative to the No-cell EtOH scenario across five major regions in the US. Significant conversion of agricultural land to energy crops is found in the South and the Plains due to the larger amounts of cropland pasture and relative yield advantage of primary energy crops (i.e., miscanthus and switchgrass) in these regions. Cellulosic ethanol scenarios lead to more conversion of cropland pasture to cropland, particularly in the Plains and the Midwest. Land use change from forestland pasture to forestland is mainly found in the southern and western regions, while the largest conversion of forests to cropland is in the northeastern region.

4.3. GHG Emissions Implication

In our model, we use ecosystem-modeled GHG balance for various energy crops and residues combined with life-cycle emissions to calculate the cumulative GHG emissions over the 2015–2035 period for each scenario. The total GHG emissions estimated here include both below- and above-ground GHG emissions in producing various commodities from farm-gate to end-use in each sector.

Table 1. Cumulative GHG emissions and GHG savings due to cellulosic ethanol (2015–2035).

(B Mg CO2e)	Ag-only EtOH	Wood & Ag EtOH
GHG Emissions Due to
Aboveground Agricultural Production	0.091	0.080
Conventional Wood Products Production	0.000	−0.009
Soil Carbon Accumulation	−0.003	−0.032
Carbon Stored in Forest	0.295	0.138
Domestic ILUC	0.126	0.118
Feedstock Transportation and Ethanol Production	0.066	0.068
Coproduct Credits	−0.039	−0.039
Net GHG Emissions	0.536	0.324
Net GHG Savings from Displacing Gasoline	0.611	0.822
LUC emission factors (Mg CO2e ha−1 yr−1)	1.135	0.870
GHG mitigation rate (Mg CO2e ha−1 yr−1)	7.205	10.098

Note: All values are measured as changes relative to the No-cell EtOH case.

presents cumulative changes in net GHG emissions in comparison with the No-cell EtOH case. As more land is allocated for producing cellulosic feedstocks, an increasing amount of GHG emissions are associated with planting, cultivation and harvest of the crops; forests and feedstock biomass; feedstock transportation; and ethanol production. Heightened biofuel demand induces carbon losses with ILUC, which plays a significant role in positive GHG emissions. GHG emissions with energy use in agricultural production are responsible for 16% of positive life-cycle emissions, followed by feedstock transportation and biorefinery process (11%). This aligns with findings in the literature [51,52]. Soil carbon accumulation with direct land use change is largely responsible for the GHG gains as the perennial grasses could sequester an average 1.5–2.5 Mg CO₂e per ha per year (see SI Table S9), which offsets the GHG losses due to crop residues harvest. Compared with the Ag-only EtOH scenario, the additional ethanol demand from woody biomass as in the Wood & Ag EtOH scenario reduces GHG losses from aboveground agricultural production by 12% and emissions from change in carbon stored in the forest by 53%. The forest management-induced GHG emission changes are not significantly different between two scenarios, primarily due to the fact that forest residues do not play a big role in the feedstock mix. Estimated domestic ILUC emissions are 0.13 B Mg CO₂e for the Ag-only EtOH scenario, and slightly fewer emissions are found in the Wood & Ag EtOH scenario (0.12 B MgCO₂e); this is mainly due to GHG gains with a greater amount of afforested land relative to the Ag-only EtOH scenario. The overall net GHG emissions are found to be 0.5 B MgCO₂e and 0.3 B MgCO₂e for the Ag-only EtOH and the Wood & Ag EtOH scenarios, respectively.

We also calculated the net cumulative GHG savings from displacing gasoline over 2015–2035. The net cumulative GHG savings is defined as the GHG emissions from gasoline that will be displaced by energy-equivalent biofuels produced net of the life-cycle emissions generated during feedstock production, transportation and conversion to biofuel over the same period. Positive GHG savings indicates that cellulosic ethanol production provides GHG benefits by displacing the energy-equivalent fossil fuels. We find that the RFS mandate could lead to 0.61 B Mg CO₂e of GHG savings by displacing fossil fuels if only agricultural biomass is allowed for ethanol production. If the same amount of cellulosic ethanol can be produced from both agricultural and woody biomass, the total GHG savings is 0.82 B MgCO₂e, a 34% increase relative to the Ag-only scenario. In general, despite the negative GHG emissions associated with feedstock production and ILUC, the cellulosic ethanol mandate met by both agricultural and forest biomass has the big potential to gain GHG savings through the displacement of gasoline in the transportation fuel sector. We further computed annual GHG mitigation potential on a per ha and per year basis by dividing total GHG savings by total induced land use change (including conversion of cropland and marginal land to energy crops, conversion of marginal land to cropland and conversion of forestland to cropland) and 20 years under each scenario. The consequent GHG mitigation rate for Ag-only EtOH and Wood & Ag EtOH scenario is 7.205 Mg CO₂e ha⁻¹ yr⁻¹ and 10.098 Mg CO₂e ha⁻¹ yr⁻¹, respectively.

4.4. Sensitivity Analysis

The model’s ability to predict the economic and environmental effects in the future will depend on the extent of divergence between assumed key parameters and real values in various economic, ecological and environmental factors. We thus examine the sensitivity of GHG savings of cellulosic ethanol to several key parameters assumed here by considering alternative values for energy crop yields (±10%), forest biomass yield of Southern forestland (±10%), the inclusion of domestic bioelectricity demand and trend in demand for conventional forest products, and ILUC in the ROW. We also examine the effects of improvements in energy efficiency in two key stages of the ethanol life-cycle (farming and ethanol production).

Our results indicate that the GHG savings of cellulosic ethanol are sensitive to assumptions about parameters; however, the effect of this on total GHG savings relative to energy-equivalent gasoline is relatively small (Figure 4). We find that GHG savings are most sensitive to the assumption of energy crop yields in both scenarios. A 10% higher yield of energy crop leads to a 12–20% higher GHG savings with cellulosic ethanol relative to the corresponding value in the benchmark case, while a 10% lower yield results in 10–16% lower GHG savings across two scenarios. We also find that a 10% more yield of Southern forestland increases the GHG savings by 8–15% relative to the value in the benchmark case. The interaction of other factors such as an extra demand for bioelectricity and a higher demand for conventional wood products leads to a 4–9% decrease in GHG savings in both scenarios. The effect of inclusion of ROW ILUC-related carbon emissions on the overall GHG savings is negative and small. An amount of 10% lower energy inputs for feedstock planting and harvesting (e.g., fertilizer, chemicals, fuels, machinery, etc.) relative to the benchmark case will lead to 1–1.6% higher in GHG savings across two scenarios. We also examine the effects of lower energy input (2000 MJ/Mg of biomass) in the biorefinery process as reported by Zhu and Zhuang [53]. We find that 50% reduction in energy usage will increase cellulosic ethanol GHG savings by 3.4–4.4% across two scenarios. In general, we find that plausible improvements in biomass yields and biorefining technology would achieve more mitigation potential, which reaffirms the findings in Field et al. [6].

5. Discussion

Biofuel mandates such as the RFS policy enacted with the intention of promoting renewable fuels to reduce dependence on gasoline and mitigate greenhouse gas emissions have received widespread attention. Research examining the land use change effects and carbon intensity of biofuels has largely focused on the corn ethanol mandate [3,4,54]. A few studies have examined the land use effects and associated GHG emissions of producing cellulosic feedstocks using integrated simulation models and found widely varying estimates. The EPA [55] applied the combined FASOM–FAPRI model and found that LUC emissions associated with switchgrass-based cellulosic ethanol production are 1.7 Mg CO₂e ha⁻¹ yr⁻¹. Similarly, Plevin and Mishra [56] estimated LUC effects with switchgrass ethanol production and found that total induced LUC emission factor is 3.3 Mg CO₂e ha⁻¹ yr⁻¹. By applying the Global Biosphere Management Model (GLOBIOM), Valin et al. [57] studied the LUC impacts of cellulosic biofuel from various feedstocks in Europe. They found that the LUC related emission factors are −1.08 Mg CO₂e ha⁻¹ yr⁻¹ for perennial grasses and 0.1 Mg CO₂e ha⁻¹ yr⁻¹ for forest residues. These studies only consider biofuel shock from some specific energy crops and, hence, are limited in their capacity to examine the optimal land use allocation and extent of LUC induced by a biofuel mandate. Our estimate of induced LUC emissions based on a mix of agricultural and forest biomass is 0.87 Mg CO₂e ha⁻¹ yr⁻¹, which is within the range of −1.9 to 3.3 Mg CO₂e ha⁻¹ yr⁻¹ from reviewed literature [58].

A few studies have assessed the optimal mix of feedstocks to meet a cellulosic biofuel mandate. Chen et al. [59] examined RFS impacts on biomass production and land use change by applying BEPAM model and found that miscanthus and forest residues would produce 49% and 22% of advanced biofuels over 2007 to 2022, with switchgrass and crop residues meeting the rest. However, the forest biomass supply curve was exogenously given in this model without connecting to the forest market. The consequent optimal feedstock mix, particularly the role of forest biomass, could be misrepresented. Other studies applied the model with endogenously simulated agricultural and forestry sectors such as FASOM to examine the RFS impacts on the cellulosic feedstock mix [23,34]. They found that corn stover and switchgrass play a significant role in supplying cellulosic biofuel by 2022, while logging residues-based biofuels are in a marginal proportion. The modeling framework used in these studies, however, was not integrated with ecosystem model, which may lead to biased simulation of feedstock yields and associated soil carbon stocks. Hudiburg et al. [17] first applied the integrated BEPAM-F model to assess the potential for the RFS to reduce greenhouse gas emissions through the use of cellulosic biofuels in 2022. They found that cellulosic biofuels can meet a 32-billion-gallon RFS target with negligible effects on food production while reducing emissions by 7% in 2022. However, their study focused on the land use change and GHG emissions within the agricultural sector without explicitly exploring the interaction with forestland and forest biomass. Moreover, the ILUC-related emissions were exogenously determined by applying carbon intensity obtained from the literature. This paper extends that analysis to examine the effects of RFS policy on land use change and GHG emissions over the 2015–2035 period and expands the analysis to study the potential for forest biomass. With newly designed RFS scenarios and integrated modeling framework that combines economic drivers, ecosystem modeling of yields and GHG balance, this study provides a more realistic estimate of direct and indirect land use change, biofuel mix and the associated GHG consequences of optimal fuel pathways. We find that agricultural and forestry residues will be the main contributor to the feedstock supply in the near term, which contradicts with the findings from previous literature focusing on the energy crops [11,12]. In addition, unlike the studies which only consider the forest sector in meeting the biofuel mandate, we find that the proportion of woody biomass is relatively small and in a decreasing trend.

The United States produced 5.5 B Mg CO₂e energy-related emissions from fossil fuels in 2020, of which the transportation sector was the largest component [60]. It is essential to find an effective solution to decarbonize the transportation sector. With decades of research development and policy support, the commercial-scale cellulosic production has become a reality. However, policy makers are cautious in promoting its large-scale deployment due to many uncertainties that still exist regarding the supply potential, induced land use change, and mitigation potential. Some previous studies have considered primarily economic drivers with cellulosic ethanol production without linking ecological features [23,33]. They failed to provide an accurate assessment of GHG potential. Other studies have considered ecological and biophysical drivers but largely ignored the economic outcomes associated with them (e.g., demand of agricultural and forestry products, land opportunity cost, etc.) [6,61]. Our study here goes beyond previous studies by assessing both direct and indirect LUC emissions and total GHG savings with an economically optimal mix of feedstock produced under alternative RFS policy scenarios. We find that the annual average GHG mitigation rate achieved is between 7.2 and 10.1 Mg CO₂e ha⁻¹ yr⁻¹, depending on the woody biomass availability. This rate outperforms the mitigation rate obtained from grassland restoration, which ranges from 1.7 to 3.5 MgCO₂e ha⁻¹ yr⁻¹, and even beats the mitigation rate with reforestation on former agricultural land (3.4 to 11.9 MgCO₂e ha⁻¹ yr⁻¹) in some regions as estimated by [6]. Our assessment of GHG impacts of RFS mandated 60 BL of cellulosic ethanol affirms the important contribution to climate mitigation that cellulosic biofuel production can make.

Our analysis has several policy implications. It shows the condition under which cellulosic feedstock expansion has the potential to offer climate benefits and would warrant further policy support. The mitigation potential with cellulosic ethanol production is several-fold larger than the pure natural-based solutions such as grassland restoration. Our findings inform the policymakers that the conflict between the natural-based solution and bioenergy production can be avoided through a well-designed bioenergy system [62]. Additionally, our analysis can inform policy incentives needed to motivate landowners to grow most appropriate feedstocks in the most appropriate places. Plausible future technology improvements in feedstock yields and energy use efficiency in farming and biorefinery are among effective solutions that can enhance the mitigation benefits of ethanol production. Future policy promoting advances in related fields is essential for achieving greater amount of mitigation potential [14].

6. Conclusions

One of the promising roads to energy independence and carbon neutrality leads away from fossil fuels and into the forests and fields. Biofuel policies such as RFS have been enacted to promote first- and second-generation biofuel feedstock production. However, continued policy support for biofuels is repeatedly challenged due to uncertainties in resource availability and GHG abatement consequences [29,63]. A thorough understanding of land use and GHG effects associated with agricultural and forestry biomass production is essential for future policy design. In this paper, we apply an economic model integrating agricultural, forestry and transportation sectors to examine the economic viability of various cellulosic feedstocks under extended RFS policy and the extent to which biofuel expansion will affect the optimal mix of feedstocks, domestic land use change and GHG emissions. This framework allows us to estimate not only direct life-cycle GHG emissions but also the emissions associated with indirect land use effects induced by changes in the prices of products and land.

Our results show that under the extended RFS mandate with 60 BL of cellulosic ethanol by 2035, mixed use of agricultural and woody biomass is an economic pathway for cellulosic ethanol production. The crop residues (corn stover and wheat straw) and dedicated energy crops (e.g., miscanthus and switchgrass) will be the primary suppliers of feedstocks for ethanol production, providing 90% of total biomass. The forest sector will supplement the remaining 10% of biomass mostly via logging and milling residues.

Increases in cellulosic ethanol production will not cause significant land use changes between and within agricultural and forestry sectors relative to the No-cell EtOH case. We find that in the scenario that restricts ethanol production from agricultural biomass only, there will be 4.4 M ha direct land use change through conversion of the existing cropland and cropland pasture to energy crop production in 2035. Increased cropland rents due to heightened biofuel demand drive 0.7 M ha conversion from cropland pasture to cropland, mainly in the Midwest and the Plains. In the scenario that considers the potential to use mixed agricultural and forest biomass, we find that more moderate direct and indirect LUC occurs within the sector and across the sector. This is mainly due to the fact that logging and milling residues are the main woody biomass in meeting the ethanol demand, which does not require a substantial increase in forest harvests.

Across different scenarios, we find that cellulosic ethanol produced in the US has significant potential to reduce GHG emissions in the transportation sector. The total cumulative GHG savings (over the 2015–2035 period) from displacing gasoline range from 0.61 to 0.82 B MgCO₂e depending on whether the forest biomass is included or not. Across a wider range of parameter assumptions about the biomass yields and future commodity markets, we find that the range of GHG savings is between 0.55 and 1.0 B MgCO₂e. These estimates are most sensitive to assumptions about biomass yields. We model land use change explicitly only in the US, and the ILUCs in the ROW are not endogenously determined given the model boundary. However, incorporating these estimates from other studies indicates a very modest impact on the total GHG savings.

In general, our study illustrates the potential roles of agricultural and forest biomass in meeting the RFS mandate and their implications for land use and GHG emissions. Our results show that with mindful planning and operation, wide-scale deployment of cellulosic biofuel can be achieved without causing significant land use change between and within agricultural and forestry sectors, and it is beneficial for future GHG mitigation. However, in this study, we have not accounted for the climate impacts of carbon storage in soils, vegetation and products. Moreover, the impacts of water usage and biodiversity with wide-scale cellulosic ethanol development are beyond the scope of this study. We leave the analysis of these factors to future research.